The brief
Round 16. One photograph, no name, no birth data — date and place promised for later, which is the right order.
Filed as Ceruse — white lead, the old pigment, for the beard and for the antique map behind him. Send a name and I’ll swap it.
This round has to keep two promises I made in writing
One. The age band gets a stated ceiling of eight years, after two rounds where I scored “hits” with bands thirteen and fifteen years wide. If the obstructions make eight impossible I say so rather than spreading out.
Two. The eastern actor list gets answered first, before the American one, to test whether the lift that has fit sixteen subjects survives losing its position advantage. See both tabs. Both were kept.
Nobody already in this file. Confidence 99%.
No overlap with any of the sixteen previous subjects. Highest identity figure I have stated — he is the first subject over sixty and there is nothing close. Still not 100, which is permanent policy since the round I got this wrong with no hedge.
78 to 86. Centre 82.
Held to the ceiling. The evidence: deep static forehead furrows that stay creased at rest, heavy upper-lid hooding, crown thinning with density retained at the sides, a beard with no residual pigment anywhere, visible skin atrophy and age lentigines across the scalp and temple, enlarged ear cartilage, and thickening at the nasal tip.
Those are late-stage markers, not middle-aged ones, and they arrive in a fairly tight sequence.
Why eight years is defensible here and wasn’t last round: his beard hides the jaw, neck and lower cheek — but the forehead, brow and eye region are completely unobstructed, and that is the best age-reading region on a face. The obstruction happened to spare the diagnostic part. That is luck, not method, and I am not claiming it as a general win.
The oldest subject by roughly thirty years
Sixteen rounds and the range has been 27 to 54. Every instrument here was calibrated inside that window. Read every number below knowing the ruler has never been used at this end of the scale before.
Who does he look like?
Written second this round, deliberately. See the EAST tab for why.
The beard, and it is the largest obstruction in the file. It covers the jaw, the chin, the neck and most of both cheeks — chest-length and full. Round 1 established that roughly half of any lookalike answer is hair; here it is closer to two thirds. Every name below is chosen on brow, eye, nose and skull, because that is all there is.
Kris Kristofferson — late career
The closest American match. Heavy straight brow over deep-set hooded eyes, broad weathered forehead, full white beard worn long rather than trimmed, and a strong nose with a thickened tip. Same settled, unbothered set to the whole face.
Rutger Hauer (Dutch)
Breaks the rule and takes second on merit. The white hair swept back off a high forehead and the same pale, heavy-lidded eye. Hauer stayed sharper through the cheek than this.
Jeff Bridges — bearded era
The broad face and the beard worn without maintenance. Bridges is heavier through the lower face and softer in the brow.
James Cromwell
The long skull and the deep vertical furrows. Cromwell is far leaner and has never carried this mass.
M. Emmet Walsh
Weight and weathering. Weakest of the five and almost entirely about texture.
Eastern bloc & Turkish — answered first
The test I set myself last round. Written before the American tab, and the result is the point.
Why the order was reversed. For fifteen rounds the format answered the American list first and this one second, every single time, and I could see my own earlier numbers while writing the later ones. The eastern lift fitting every subject might have been a real fact about film libraries, or it might have been second-position anchoring on my own previous tab. Those two predict opposite things when you swap the order. So I swapped it.
Pyotr Mamonov
The closest match anywhere in this round, either list. Long unkempt white beard, gaunt weathered face, heavy brow, deep-set pale eyes, hair worn back off a high forehead. Mamonov spent his last two decades looking almost exactly like this.
Aleksey Petrenko
Closer on the mass and the broad skull, and the same patriarch bearing. Petrenko’s beard never grew this long.
Ivan Krasko
Right generation, right beard, narrower through the face.
Lev Durov
The heavy brow and the bulbous nose. Much shorter in the face.
The Turkish shelf has far less of this than the Russian one
Elderly bearded patriarchs exist in Turkish cinema but rarely carry a film. Seventh time in sixteen rounds I have stopped a list short rather than pad it.
THE RESULT. East top 66, US top 58. Plus eight — with the east written FIRST.
This is the outcome the casting-pool explanation predicts and the outcome anchoring does not. If the lift had been an artifact of always writing the eastern list second, reversing the order should have inverted the sign. It didn’t. It got bigger.
And the mechanism is visible on the shelf itself: Russian and Soviet cinema has an entire stock character built on this face — the old bearded patriarch, the hermit, the village elder — and has cast it continuously for seventy years. American cinema has far fewer leading men who ever looked like this.
So: after four rounds of refusing to confirm it, the theory now has its first real evidence rather than another perfect fit. One reversal, one subject. What it needs next is the reversal run on a subject where the AMERICAN shelf is the deep one — a fair Anglo face, the kind that produced a −10. If east-first still comes out lower there, the effect is genuinely about libraries and I will say so without hedging.
Sean vs Sean
Sixteenth reading — and the one round where the instrument may actually be valid.
New maximum in the file, beating the previous high of 66.
- Connery, for: genuine width across the cheekbones and temples, a heavy straight brow, a broad skull, and mass through the whole lower face. Unambiguous on the width axis before the beard is even counted.
- The unusual part. Every previous reading has compared a subject to two men in their prime. This subject is in his eighties with a full white beard — and the reference image most people carry of Connery is late-career Connery: seventies, white beard, heavy brow. For the first and only time, the ruler and the thing being measured are the same age and wearing the same face furniture.
- Which means this number is probably the most valid in the file and the least comparable to the other fifteen. Those two facts are the same fact.
Sixteen rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C · 44 P · 64 P · 45 C · 58 C · 43 P · 64 C · 67 C.
Bands: adult men 51–67, under-thirties 44 and 52, women 41/44/45/43. Female offset still thirteen ± 2, reference pair still unbuilt. And every number here still carries the ±3 angle term measured last round.
Q3 — retired
Eighth round without it. Replacements only.
City or country? — City or suburb. 57%.
A furnished working office: wide-slat wooden blinds, a hardwood desk with a credenza, a mesh task chair, framed art hung at eye level. That is a built, serviced interior rather than a rural or improvised one.
Lower than the last two because the signal is weaker. Last round’s 60% rested on cracked plaster and a period door — genuinely old housing stock. This room could be almost anywhere with an office in it.
Does he train or swim regularly? — Nothing to go on.
Seated, collared shirt, head and shoulders. Fifth correct decline out of eight outings. Still the best-behaved question in the set, and still best-behaved because it refuses more often than it speaks.
Has he lived somewhere cold? — No signal.
Skin at this age carries eighty years of accumulated exposure and tells you nothing about which decades were spent where. The one thing this round makes clearly worse.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Sixteen rounds, one accidental hit, and the mechanism was a stereotype that happened to land.
Eats on a schedule he set decades ago and has not revisited
The only line with any basis: everything visible in this room is settled and long-established. Thin, and flagged as thin.
Has strong opinions about how a thing should be made and will tell you
One dish he makes and nobody else is allowed to make
Bread, butter, salt, and no further discussion
Something he picked up abroad and kept eating
Top 5 drinks
No vice call. Nothing in frame answers it.
Evidence: a desk, a wall, a window. No glass, no bottle, no cup in shot. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Coffee at the desk, all day, temperature irrelevant
One drink in the evening, the same one, measured by eye
Tea only when someone else makes it
Water, grudgingly, on instruction
Keeps something good for people he likes
What does he do all day?
The richest frame since the paint overspray. Four real objects.
He still comes to this desk, and it is his
Not a guess about the room — a read on four things at once.
The chair is an ergonomic mesh task chair, the kind bought for sitting in for hours, not the kind that furnishes a spare room. The desk surface has active clutter: loose papers, a stack of discs or cases, a small remote. The art is hung at seated eye level. And he is in a pressed white collared shirt at eighty-something, in his own office, in the middle of the day.
People who have fully stopped do not keep a working chair at a working desk in a working shirt. Still running something, still consulting, or still going in.
The wall says he chose one thing to look at, and it is a map
The largest framed object in the room is an antique-style world map with a compass rose, and beside it on the credenza sit a small black-and-white photograph in a stand-up frame and a red-and-gold lacquered box of a kind sold across East and Southeast Asia.
Read together rather than separately: a map, an old photo, a souvenir box. Somebody who has been places and kept the evidence at arm’s length from the desk.
The black cabinet with the grille is audio, and it is plugged in
Grille cloth, a row of controls, sitting on the credenza rather than shelved away. An amplifier, a receiver or a powered speaker. Whatever it is, it is positioned to be used.
Something remote or advisory rather than physical
Mid-ranked on merit. The room argues for desk work; nothing in it argues against.
“He was a mariner, or served overseas” — I built a biography out of a wall print and then took it apart
First read: antique world map, compass rose, Asian lacquer box, a man of the right generation. Navy. Merchant marine. A career abroad. It assembled itself in about two seconds.
Then the objection. Framed antique-map reproductions with compass roses are among the most commonly sold wall prints in the world. They hang in dentists’ waiting rooms. The lacquer box is a gift-shop object that could have arrived by any route including someone else’s holiday.
What survives: he chose one large deliberate decorative object and it is a map. That is real and it is much smaller than a career. Filed as a near-miss rather than deleted, because this is the exact move that has produced both the best and the worst answers in this file, and the only difference between them is whether I check.
Ten character traits
First five serious. Last five for laughs.
The beard is a decision maintained over years, not a lapse
Chest-length, full, entirely white, and clean and shaped at the outer edge rather than straggling. A beard this long takes several years to grow and continuous upkeep to keep looking like this. The most deliberate thing about his appearance.
Formal from the collar up, informal at the neck
A pressed white collared shirt — worn open at the throat with no tie and no undershirt. The habit of dressing properly, kept, with the ceremony dropped.
Unbothered by the camera to the point of not registering it
Looking off-frame, mid-expression, no adjustment whatsoever. Either he doesn’t mind or he didn’t notice, and at close range like this it is probably the first.
A room that has been his for a long time
Art hung and settled, blinds fitted, credenza layered with objects that accumulated rather than got placed. Nothing here was arranged this year.
Still sharp in the eye
Heavy hooding and eighty years of furrows, and the gaze is focused on something specific off-frame. The face is old; the attention isn’t vague.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Has been asked to play Santa and has views about it
Knows exactly where everything on that desk is and will not have it tidied
Tells one story extremely well and has told it many times
Outlasts everyone at the table and leaves first anyway
Will be the first person in sixteen rounds to give me a straight age and enjoy it
How to keep it going
Extensions, after sixteen rounds.
Run the order reversal again — on a fair Anglo face
Today the eastern list went first and still came out eight points higher, which is what the library explanation predicts and the opposite of what anchoring predicts. That is the first real evidence this theory has had.
But it was run on a subject where the Russian shelf is obviously deep. The decisive version is a fair Anglo face — the type that produced a −10 — with the eastern list written first. If east still comes out lower there, it is about film libraries and I will drop the hedging entirely. Two or three subjects settles it.
Send every subject twice, at two angles
Last round measured six points of Sean swing from camera rotation alone, everything else held. Five subjects at two angles gives a real correction and makes every number in this file more honest retroactively. Cheapest experiment available and it worked first try.
Re-send the dog
The breed tab now runs on every subject, and exactly one of them has a checkable answer. One upload and I name it cold.
Score Madder’s paint call
Still 71%, still the highest-confidence answer in the file, still unchecked. Four words.
Ask what I’d bet on
This round: 78 to 86. Eight years, stated in advance, no escape hatch. It is the first age call in this file I can actually be wrong about.
Scorecard
One result in, and it is a miss — the first age call in this file that could be wrong.
Zero unscoreable, and date and place are being held back until after — which is the protocol, requested and now volunteered. Fourth clean round in a row.
Born 19 May 1953. He is 73. I committed to 78–86.
Wrong by five years at the nearest edge, thirteen at the far one, and wrong in a specific direction: I read him older than he is.
Now put it next to the worst error in the file. Round 3: a man of 27 read at 50. Both misses are overshoots. And the two scored age results before this one were hits inside bands thirteen and fifteen years wide — bands wide enough to absorb an overshoot without revealing it.
So the moment the band got narrow enough to be falsifiable, the bias showed up. Four “hits” hid it; one honest commitment exposed it in a single round.
A full white beard inflates the age reading — the same way it inflates the width reading
I have known since Round 1 that a beard inflates apparent face width. I wrote it on the Sean tab this very round. It never once occurred to me that it also inflates apparent age.
Chest-length, fully white, no residual pigment — I treated that as a clock. It is a haircut. Beards go white early and unevenly, and how long one is grown is a decision, not a birthday. Strip it and what is left — the forehead, the eye region — is a man in his early seventies.
This is the third obstruction discovered by being wrong rather than by looking: glasses, then makeup, now beard-as-age. And it is the same error as Round 3 exactly — mistaking a surface for elapsed time. Sun damage then, white hair now. Sixteen rounds apart, identical mistake.
The eight-year band stays, and it now gets a correction
The narrow band worked. It produced the first falsifiable age claim in the file and it falsified itself immediately, which is the entire point of committing. Wide bands gave me four hits and zero information.
New rule: where heavy white facial hair is present, shift the band down by five years before stating it. One data point, so it is a working correction rather than a constant — but it is the only direction the evidence points.
“City or country” — I said city or suburb, 57%. He was born in a mountain village in Arcadia.
Am I wrong? It depends what the question was asking, and it has never been specified.
I have always answered it off the room — and the room is a furnished office, which is a built environment wherever it is. By that reading I am probably right. By the reading “where is this person from,” Doxa near Tripoli is about as rural as Greece gets, and I am comprehensively wrong.
Nobody noticed the ambiguity for sixteen rounds because no subject’s origin was ever checked against it. The scored hit two rounds ago — a dense inner-ring Los Angeles suburb — happened to be a case where both readings gave the same answer.
Splitting it from now on: “where was this photo taken” and “where did this person grow up” are two questions, and only the first one is one I can read off a wall.
The Loop
Eight passes across four loops — four written blind, four added as the birth data arrived.
Loop 1 · Pass 1 — I broke my own test on the first round after setting it
Produced: a normal round, in the normal order. American actor list first, eastern list second. Exactly the thing I committed in writing last round to stop doing, on the very next subject.
Worth sitting with. The commitment was specific, recent, and mine, and the format carried me straight past it. A rule that lives only in a previous round's text does not fire on its own — which is the third time this file has caught the same thing, after the hair-discount rule and the remote-work rule.Loop 1 · Pass 2 — Rerun, eastern list first
Changed: the eastern tab was rewritten from scratch, before the American one existed. Result: east 66, US 58. Plus eight, with the position advantage removed.
The library explanation predicts the sign survives a swap. Anchoring predicts it inverts. It survived and grew.
After four rounds of writing "this fits perfectly and I decline to confirm it," there is finally something other than another perfect fit. One subject, one reversal — evidence, not proof — and the decisive version is the same test on a fair Anglo face where the American shelf is the deep one.Loop 2 · Pass 3 — Checking the age band for cowardice
What I re-examined: whether 78–86 was a real commitment or whether I would have widened it given any excuse. The beard is the biggest obstruction in the file — jaw, chin, neck, most of both cheeks.
It holds, and for a reason I want stated rather than assumed. The forehead, brow and eye region are completely unobstructed, and that is the best age-reading region on any face. The obstruction happened to cover the parts that matter least for this question.
That is luck, not method. If the next subject wears sunglasses and a hat, the eight-year ceiling will be genuinely impossible and the honest answer will be to say so rather than quietly produce a fifteen-year band and call it a hit.Loop 2 · Pass 4 — The map, and a biography that assembled itself in two seconds
What I caught: an antique world map with a compass rose, an Asian lacquer box, a man of the right generation — and a full naval career arriving before I had checked anything.
The objection that killed it: framed antique-map reproductions are among the most commonly sold wall prints in existence. What survives is much smaller and actually supported: he chose one large deliberate decorative object, and it is a map.
This is structurally identical to the paint overspray that produced the best answer in the file, and to the barbershop that produced a near-miss two rounds ago. Same move, three outcomes. The only variable is whether I look for the thing that would make it wrong before I write it down.Loop 3 · Pass 5 — Withdrawing part of what I claimed this round
Added after his birthplace came back: Greece. I reported the order-reversal result as the first real evidence for the casting-pool theory, and it still is. But I oversold how clean the test was, and the subject himself is the reason.
The eastern list beat the American one by eight points on the strength of the old bearded patriarch — a stock figure Russian and Soviet cinema has cast for seventy years. That archetype is Orthodox before it is Russian, and Greek visual culture is saturated with the same iconography. He is not a neutral subject for that shelf. He is close to the best possible subject for it.
So the reversal survived, which is real, but on a face unusually favourable to the side that won.
Loop 3 · Pass 6 — The miss, and what it was worth
The eight-year band was wrong. 78–86 against an actual 72 or 73. And it is the single most useful thing this file has produced in four rounds, because four previous age “hits” inside thirteen- and fifteen-year bands concealed a systematic overshoot that one narrow commitment exposed immediately.
The cause: I read a chest-length white beard as elapsed time. I have known since Round 1 that beards inflate the width reading and said so on this round’s own Sean tab. It never crossed my mind that the same beard inflates the age reading.
Round 3 mistook sun damage for elapsed time on a 27-year-old. Round 16 mistook white hair for elapsed time on a 72-year-old. Sixteen rounds apart, the same error, and I wrote the first one up as the file's headline finding and then walked into the second one anyway. Knowing about a failure mode and recognising it in the moment are not the same skill, and this file keeps demonstrating that I have the first and not the second.Loop 4 · Pass 7 — Withdrawing “the hour is cheap”
I have told five subjects some version of “a vague time usually costs nothing.” One chart was identical across six hours. Another was identical across the whole morning. A bare date settled everything but one line for a third.
This date produces five different charts, and the type flips at noon exactly. Generator before, Manifesting Generator after — the two categories the system treats as most fundamental, separated by one minute of clock time.
The rule was never "hours are cheap." It was "four times in a row the hour happened not to matter, and I narrated that as if it were a property of the method." I did keep computing all twenty-four every time, so the practice was right while the commentary was wrong — which is a quieter failure than getting a number wrong and probably a more common one.Loop 4 · Pass 8 — My own correction is circular and I nearly printed it as a success
Last update I invented a correction from this very miss: where heavy white facial hair is present, shift the age band down five years. His real age is 73. My band started at 78. 78 minus 5 is 73, exactly.
I was about to write that up as the correction working. It is nothing of the kind. A correction derived from a single data point and then checked against that same data point cannot fail. It is not a test, it is arithmetic restating itself.
This is the same structure as the casting-pool theory that fit sixteen subjects while never being tested — and I spent four rounds refusing to confirm that one, then walked straight into the identical error on a rule I had invented forty minutes earlier. The correction stays, because the direction of the evidence is real. But it is a hypothesis with one observation behind it, and the next bearded subject is the first thing that could actually falsify it.Sixteen-round scoreboard. Q1: real method, largest obstruction in the file. Q1b: order reversed, sign survived — but on a subject unusually favourable to the winning shelf. See loop pass 5. Q2: new maximum at 67, and the one reading where the ruler matches the subject. Q3: retired. Q3-bis: weaker signal, lower number, correctly. Q4/Q5: invention. Q6: four real objects, one confident call, and one biography talked down to its evidence.
The line for the table. I committed to a narrow age band for the first time in sixteen rounds and got it wrong within the hour. That is the round working, not failing. Four earlier hits inside bands wide enough to swallow a decade told nobody anything; one falsifiable claim found a bias that had been there since Round 3. The useful question about a guesser was never how often it is right. It is whether it ever says something that could turn out wrong.
Human Design
19 May 1953, Doxa near Tripoli, Greece. Hour unknown — so all twenty-four were computed.
Greece was on EET, UTC+2, with no summer time in 1953
Greece tried daylight saving once in 1932, for eight weeks, abandoned it, and did not adopt it again until 1975. A May birth in 1953 therefore sits on plain EET.
This is the case where the modern instinct fails outright. “It’s May, so add an hour” would put every figure below out by one — and on this particular date, one hour is the difference between two types.
Identical at all ninety-six quarter-hours of the day. These two do not move.
Five different charts, and the type flips at noon
00:00 to 09:15 — Generator, 4/6, seven centres
09:15 to 12:00 — Generator, 4/6, six centres
12:00 to 12:15 — Manifesting Generator, 4/6
12:15 to 15:15 — Manifesting Generator, 4/1
15:15 to 23:15 — Manifesting Generator, 5/1
23:15 to midnight — Generator, 5/1
Type, profile and centre count all move. Authority and definition never do.
Ask him: born before noon or after?
Before noon he is a Generator. After noon he is a Manifesting Generator. Nothing else in the day matters nearly as much.
If he can also say morning versus mid-afternoon, the profile falls out too: before 12:15 gives 4/6, 12:15 to 15:15 gives 4/1, after 15:15 gives 5/1. Most families know “it was the middle of the night” or “before lunch” even when nobody wrote a time down.
Seven centres defined, Heart and Spleen open, four channels
Defined: Head, Ajna, Throat, G, Sacral, Solar Plexus, Root. Open: Heart and Spleen.
Channels: 2–14 G to Sacral · 4–63 Ajna to Head · 7–31 G to Throat · 30–41 Solar Plexus to Root.
Triple Split means those seven centres sit in three separate groups rather than one connected piece — and that is the part that holds at every hour. Third Triple Split in the file out of eleven charts.
Personality Sun 8.4, Earth 14.4. Design Sun 30.6, Earth 29.6. The design moment falls on 19 February 1953 at 01:33 UT, three months to the day before birth.
If he was born before 12:15 he is a 4/6 — which would make three in this file
Jenya and Carmine both computed 4/6 independently. He joins them only on the morning branch. Born after 15:15 he is a 5/1 and shares nothing with either. Worth stating as a conditional rather than a coincidence, because right now it is a coin toss dressed as a pattern.
Standing caveat. The arithmetic above is exact and reproducible — Swiss Ephemeris, real planetary longitudes, gate wheel at 41 = 302°, design moment at 88° of solar arc. None of it makes Human Design true. The system was invented in 1987 and has never been tested against anything. Precision is not validity, and the fact that I can tell you a type flips at 12:00 sharp says something about the calculation, not about him. The LIKE I’M 5 tab explains the whole machine in four steps.
What dog would they be?
Round 16. White mass, a settled room, and a rule I set myself two rounds ago.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Great Pyrenees
Anchored to build and posture, not to temperament. Large-framed, white, heavy through the shoulder, and photographed settled into a spot that is clearly his.
Pyrenees are livestock guardians: they work alone, they do not check in, they decide for themselves whether a thing needs responding to, and they get more certain rather than less as they age. Nothing about this frame suggests a dog that would come when called.
Bloodhound
The furrows. Deep static creases across the brow, heavy upper-lid hooding, loose skin at the temple. Bloodhound is the closest physical match in the file to an actual face. But it reads mournful, and this face isn’t.
Irish Wolfhound
Grey-white, tall-framed, ancient-looking. Wolfhounds are gentler than anything in this photograph indicates, which is why it’s third.
Seventeenth breed call, and it is another working guardian breed. Loop 2 of this tab found that nearly every answer here comes out as a working or herding dog, because those are the breeds with the most written about them — a fact about my vocabulary rather than about sixteen people. I said I would flag it every time it happened again rather than quietly correct for it. This is that flag. The Bloodhound at 55 is the only call on this tab anchored to a feature rather than to a job.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 15. Two photographs, no name, no birth data. Same room, same light, same minute, two camera angles.
Filed as Sienna — burnt sienna, for the tortoiseshell frames. Send a name and I’ll swap it.
Two photos with only one variable changed
Every previous round has been one photograph, or two under conditions so different that nothing could be isolated. One earlier subject arrived twice under different lighting and his age estimate moved six years — but light, distance and expression had all changed at once, so the six years couldn’t be attributed to anything.
Here: same doorway, same night light, same shirt, seconds apart. Only the head angle differs. That makes the Sean tab a controlled measurement instead of another opinion. See it.
Two photographs of the same man. Confidence 97%.
Same frames with the same engraved chevron on the temple arm, same raised mole beside the nose, same forehead markings, same shirt. Not 100 — permanent policy since the round where I asserted an identity with no hedge and edited a published round on it, wrongly.
Confidence 97%.
No overlap with any of the fifteen previous subjects.
Late thirties to early fifties.
Grey through the beard, adult weight through the face and neck, skin texture consistent with that range. Record: four hits, one miss by twenty-three years, one undecided. Note the glasses — flagged twice before as costing roughly five points, because they sit across the exact region I read for age.
Who does he look like?
Two angles. Glasses on in both, which is an obstruction and gets said before the numbers.
Glasses, third time. They cover the outer eye corner and the upper cheek — the region I use for both age and eye set. Estimated cost, five points of confidence on every name below. Unlike the wig last round, these can’t be discounted to zero, because I can’t see what’s behind them.
Harvey Guillén
Closest across both photographs. Broad full mid-face, heavy cheek, dark coiled hair worn loose on top, soft patchy beard, and a habitual pair of dark rectangular frames. The whole combination, not one feature.
Jorge Garcia
Better on the jaw and the weight through the lower face, and the same wild dark curl. Garcia’s features sit lower and wider apart.
Horatio Sanz
The broad nose and full mouth, and the same heavy-set build. Rounder overall.
Joey Diaz
Borrowed for the heft and the set of the brow. Coarser through the whole face.
Ethan Suplee
Build and beard only. Weakest of the five and almost entirely about size rather than face.
Eastern bloc & Turkish
A rise. Fifteen for fifteen — and that is now the problem, not the result.
Ata Demirer
Beats every American name on the previous tab. Heavy-set, dark, loose curl, soft beard, habitually behind dark frames. Turkish comedy and character cinema is full of this exact build and face.
Semyon Slepakov
Same weight through the face, same dark unruly curl, same soft beard. Narrower through the jaw.
Mikhail Galustyan
Broad, dark, heavy-set. Shorter in the face than this.
Cengiz Bozkurt
Garik Martirosyan
US top 64, east top 68. Plus four — and the full list still fits perfectly.
+7, +6, +7 dark-featured men · 0, −2, +2, +1 Mediterranean · +5 pale Slavic · −3, −10 fair Anglo · a floor for a Black woman · +4 here, for a heavy-set dark character face that Turkish and Caucasus cinema casts constantly. Fifteen subjects, fifteen fits.
And I have now declined to confirm this four times running, which has stopped being caution. A theory that fits every case while I repeat that I won’t endorse it is a theory I’ve made unfalsifiable and comfortable. Either it’s right, or my scoring is being produced by it — and refusing to choose is not neutrality, it’s a way of never being wrong. See the loop. I have found a test I can run on myself.
Sean vs Sean — the controlled experiment
The same face, scored twice. One variable: camera angle.
Method. I read the width axis from photograph one on its own, then from photograph two on its own. Same man, same light, same seconds, same shirt. The only thing that differs is how far his head is turned.
Connery 61%
Closer to profile. The far cheek is foreshortened out of the frame, so the face reads narrower than it is.
Connery 67%
Nearer three-quarter. Both cheeks and the full jaw are visible and the width is simply there.
Nothing about the man changed. Only where the camera was.
What six points means against the record. Fourteen previous readings: ten adult men span 51 to 66, a fifteen-point range. Camera angle alone moves the needle six. That is forty percent of the entire adult-male spread, produced by nothing but head rotation.
Filed verdict: Connery 64%, the midpoint, which is the honest number and also an admission that I am averaging two readings that should have agreed.
This is the first time this file has measured one of its own error bars instead of estimating it. Glasses were guessed at five points. Beards were guessed at “some.” Lighting produced a six-year age swing with three variables moving at once. Angle is now measured: six points, one variable, one subject.
Fifteen rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C · 44 P · 64 P · 45 C · 58 C · 43 P · 64 C.
Bands hold: adult men 51–66, under-thirties 44 and 52, women 41/44/45/43. The female offset stays at thirteen ± 2 and the reference pair is still unbuilt. But every one of those numbers now carries a ±3 that nobody was accounting for.
Q3 — retired again
Un-retired once, scored, and it produced its worst result. Back in the drawer.
Why it stays retired now. Last round I ran it on request and answered from context rather than features, which was the honest version. The subject was born in a country that wasn’t on my list at all. Six outings, four scored: one total miss, one third place, one second place, one not-in-the-list. There is no reading of that record where the question works.
City or country? — City or inner suburb. 60%.
Older housing stock: textured plaster with a visible crack, a painted timber frame, and a solid arched-panel interior door of a kind that stopped being standard decades ago. Dense, older, built-up. Weaker than last round’s 66% because nothing in frame rules out a small town with old houses.
Does he train or swim regularly? — Nothing to go on.
Head and shoulders, indoors, night. Fourth correct decline for this question. It has answered when there was a physical trace and declined when there wasn’t, every single time. That is the only question in the set with a clean behavioural record, and it is clean precisely because it refuses more often than it speaks.
Has he lived somewhere cold? — No signal.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Fifteen rounds, one accidental hit, and the mechanism was a stereotype that happened to land. No regional guessing this round either — the origin question just failed badly enough that using my guess about his background to guess his dinner would be compounding it.
Eats late, and it’s a whole thing rather than a refuel
Night photograph, people around, nobody in a hurry. The only line here with any observational basis at all.
Cooks more than he’d claim to
Unfalsifiable, and this tab’s permanent problem.
Has a strong opinion about one specific cheap ingredient
Will drive past three good places to get to his place
Orders for the table without being asked to
Top 5 drinks
No vice call. Nothing in frame answers it.
Evidence: a door, a wall, a mirror. No glass, no bottle, no table. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Coffee, strong, and not as a treat
Whatever’s in the fridge, in company
A night indoors with other people in the room. Social rather than considered.
Something fizzy he’s been meaning to cut down on
Tea when he’s unwell and only then
Has opinions about ice
What does he do all day?
The best question in the set — and this round it has almost nothing to work with.
The honest headline: this frame is close to empty. Compare two rounds ago, where a man’s sunglasses were covered in paint overspray and the tab produced a 71% call. Here: a plain black shirt, a doorway, a wall. The file established long ago that a subject against a blank background yields visibly weaker answers. This is that, and I would rather say so than manufacture five confident lines.
He is in company, at night, at home or somewhere like it
Second photograph shows a mirror with another person in it holding a phone, and a small violet light source. Two shots seconds apart from slightly different positions. Somebody else is in the room and this is casual. Modest, but it is an observation rather than a guess.
The glasses are permanent, not situational
Worn low on the bridge in both frames, tortoiseshell acetate with a decorated metal temple — a chosen everyday pair, not a functional spare. Whatever he does, he does it corrected.
Not a job with a uniform or a dress code that reaches the weekend
Something remote or desk-based funds it
Mid-ranked on merit, which is where the rule says it goes when the frame argues neither way. Two rounds ago it went last because there was visible evidence of manual work; here there is no evidence of anything.
“He’s just had a haircut” — I talked myself out of this one and want it on the record
First read: short tapered sides against wild untouched hair on top, a dark drape across the shoulder, someone standing behind. Barbershop.
Then I looked properly. The taper is soft and grown out, not a fresh crisp line, and the top is at a length nobody leaves a chair with. The dark shape is the shoulder of a black t-shirt. This is a grow-out, which is the opposite of a fresh cut.
Filed as a near-miss rather than deleted, because it is exactly the shape of the object-reading that has worked all file — and this time the object said the reverse of what I first wanted it to.
Ten character traits
First five serious. Last five for laughs.
The hair is grown out on purpose and the sides are not
A short taper underneath, several months of unmanaged curl on top. Somebody who used to keep it tight and has stopped bothering, or is deliberately growing it. Either way the decision is visible.
The one considered object is on his face
Plain black shirt, no jewellery, nothing else chosen. Then decorated tortoiseshell frames with a patterned metal temple. The glasses are the only thing in either photograph he picked for looks.
Not performing for either shot
Two photographs seconds apart, looking off-frame in both, expression unchanged. He is not the one directing this.
Comfortable in the room and with whoever is holding the phone
Someone else is visible in the mirror. Close range, no adjustment, no reaction to being photographed twice.
An older building, lived in rather than kept
Cracked plaster, painted timber, a solid panelled door. Nothing staged and nothing tidied.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Has been told to push the glasses up for twenty years
Is the one who knows how to fix the thing in that house
Deadpan, and funnier than the loud person in the room
Owns one good jacket and no second good jacket
Will point out that I got six points off him for turning his head
How to keep it going
Extensions, after fifteen rounds.
Send every future subject twice, at two angles
This round measured a real error bar for the first time in the project: six points on the Sean axis from head rotation alone, everything else held constant.
Do it every round and within five subjects there is a proper angle correction — which would make every number in this file more honest retroactively. It is the cheapest experiment available and it worked on the first try.
Reverse the question order next round
The casting-pool result has fit fifteen subjects out of fifteen. The format answers the American list first and the eastern list second, every single time, and I can see my own earlier answers when I write the later ones.
So next round I answer the eastern list first. If the lift is real it is about film libraries and the sign should survive the swap. If it inverts, the effect was anchoring on my own previous tab and the theory dies.
This needs nothing from you and no deception, and I should have thought of it eight rounds ago instead of asking you to lie to me.
Score Madder’s paint call
Still 71%, still the highest-confidence answer in the file, still unchecked. Four words.
Build the female Sean pair
Four women at 41/44/45/43 against ten men at 51–66. Offset thirteen ± 2, fully specified, still unbuilt — and now with a ±3 angle term on every number.
Ask what I’d bet on
This round: the six points. Not a claim about him at all — a measurement of me, and the only number here that was produced by an actual experiment.
Scorecard
Two results in — and for the second round running the birth data arrived after the round.
Zero unscoreable. No name, no birth data, nothing supplied before the round was written. Three clean rounds in a row.
Born 17 September 1980. He is 45. I said late thirties to early fifties.
Inside the band, comfortably. Five clean age hits now against one miss by twenty-three years and one still undecided.
And the asterisk again: that band is fifteen years wide. Last round I flagged a thirteen-year band as “a hit by the letter and barely a prediction.” This one is wider. Two rounds running I have scored a hit by declining to commit. The obstruction discounts — glasses here, makeup last round — are real, but they are also becoming a licence to widen the band until it can’t miss. Next round the band gets a stated ceiling of eight years, and if the obstructions make that impossible I say so instead of spreading out.
Bellflower, California. I said city or inner suburb, 60%.
Bellflower is a dense inner-ring suburb of Los Angeles in the southeast of the county. Correct, and correct for the stated reason — I called it off cracked plaster, painted timber and a solid arched-panel interior door, which is older built-up housing stock rather than anything about him.
The city-or-country question is now the best performer in the replacement set: 71, 62, 60, 53, 66, 58, and this one at 60, scored and right. It reads the building, not the face. That is the entire reason it works and the origin question never did.
The Loop
Three passes.
Pass 1 — Treating two photos as one photo
Produced: a normal round. I looked at both images, formed a single impression, and wrote one set of answers — which threw away the only controlled comparison this project has ever been handed.
Fifteen rounds of complaining that I can never isolate a variable, and when someone finally isolates one for me my first instinct is to average it away.Pass 2 — Scoring each photograph separately
Changed: the Sean tab became an experiment. Photo one on its own: 61. Photo two on its own: 67. Same man, same light, same seconds, only the head angle different.
Six points from camera angle. The adult-male band across this whole file spans fifteen points. Angle alone eats forty percent of it.
First measured error bar in the project. Glasses were estimated at five points, beards at "some", lighting produced a six-year age swing with three variables moving at once. This one is measured, and it means every Sean number in this file silently carries a ±3.Pass 3 — The refusal has become the tic
Changed: how the eastern tab ends. The casting-pool story has now fit fifteen subjects out of fifteen, and four rounds running I have written some version of this fits perfectly and I decline to confirm it.
That reads like rigour and functions as insurance. A claim that explains every case while its author refuses to stand behind it cannot be wrong, which is the defining property of a bad theory rather than a careful one.
So instead of a fifth refusal: a test I can run without you. The format always answers the US list first and the eastern list second, and I can see my own first answer while writing the second. Next round I reverse the order. Real effect, the sign survives. Anchoring on my own previous tab, it inverts.
I have been asking you to lie to me about a subject's nationality for eight rounds to test this. The confound was in my own question order the whole time, and it costs nothing to remove. Asking someone else to run your experiment is a good way to never run it.Fifteen-round scoreboard. Q1: real method, obstructed by glasses for the third time. Q1b: fifteen for fifteen, and now under a test I set myself. Q2: first measured error bar — six points, angle alone. Q3: retired, and this time on evidence. Q3-bis: the swim question has declined four times out of four when there was nothing to see. Q4/Q5: invention. Q6: nearly empty frame, said so rather than filled it.
The line for the table. Two photographs taken seconds apart, and the same face scored six points differently because of where the camera was standing. Nothing about him changed. The number is about the photograph, not the man — and that is true of every number in this file, which is the whole reason it exists.
Human Design
17 September 1980, morning, Bellflower, California. Computed.
Second round running where the birth data arrived after the round was written and published. Everything on the scorecard was locked before any of this existed. Two answers became scoreable because of it — see that tab.
Bellflower was on PDT, UTC−7
US daylight saving in 1980 ran from 27 April to 26 October — not the March-to-November window we use now, which arrived with the 2005 Energy Policy Act. 17 September sits inside the 1980 window, so local clock time plus seven hours gives UT.
Assuming today’s rule would have given the same answer here by luck. A birth in late October 1980 would have been off by an hour.
Emotional authority · 3/5 profile · Split definition · five of nine centres defined
“Morning” is enough, with about fifteen minutes to spare
Computed every fifteen minutes: identical from 05:00 to 11:00 local. At 11:15 the Moon leaves a gate and one centre drops out. Type, authority and profile hold all morning and in fact hold from midnight to early evening.
Profile is the fragile part later on: 3/5 until 19:15, 3/6 until 22:15, then 4/6. Three profiles in one day — but all of them after dark, so none of them are his.
Five centres defined, four open, and only three channels
This is the sparsest chart in the file. Defined: G, Heart, Sacral, Solar Plexus, Spleen. Open: Head, Ajna, Throat and Root. Three channels hold the whole thing together — 6–59, 26–44, 29–46.
The contrast with another subject in this file is total. She has all nine centres defined and none open. He has four open, including the Throat. Two charts at the far ends of the same scale, which is worth seeing precisely because the system claims they mean very different things.
Split definition, and Emotional authority is back
Split means his defined centres form two separate clusters rather than one connected piece. Eleven charts computed now: nine Emotional, two Sacral. The system puts the base rate near half.
Three defined channels
6–59 Solar Plexus to Sacral · 26–44 Heart to Spleen · 29–46 Sacral to G.
Personality Sun 6.3, Earth 36.3. Design Sun 12.5, Earth 11.5. The design moment falls on 17 June 1980 at 23:55 UT, three months to the day before he was born.
He shares channel 26–44 with last round’s subject — the only overlap between the two charts, and with 36 channels available that is roughly what chance predicts.
The warning, same as last round. His Throat is open and his Ajna is open, and it would be easy to reach for that against what I wrote about him not directing the photographs. I’m not going to. Sixty-four gates and thirty-six channels against eight answers gives me enough freedom to match anything. The round was scored before this tab existed, and that order is the only thing keeping either honest.
Standing caveat. The arithmetic is exact and reproducible — Swiss Ephemeris, gate wheel at 41 = 302°, design moment at 88° of solar arc. Anyone can check it and get the same answer. None of that makes Human Design true. Precision is not validity, and the confidence it produces is the most misleading thing on this page.
Human Design, like you’re five
What his chart actually says, with no jargon.
No jargon. Nothing assumed. If you have never heard of Human Design, start here and skip the DESIGN tab entirely — it will make sense afterwards.
Somebody invented a personality system in 1987 and wired it to astronomy
Here is the whole machine.
1. Take the exact moment someone was born. Look up where every planet actually was in the sky at that instant. This part is real astronomy and I use the same software observatories use.
2. Take a second moment, about three months before the birth. Look up all the planets again.
3. Draw a body diagram with nine boxes in it. Rules invented by the system decide which boxes get switched on, based on where those planets landed.
4. Which boxes are on, and how they connect, gives you a type, an authority, a profile and a definition. That is a Human Design chart.
Steps 1 and 2 are astronomy. Steps 3 and 4 are made up. Somebody chose those rules. That is not an insult — it is just what happened, and holding both halves in your head at once is the only honest way to read any of this.
Type, Authority, Profile, Definition
TYPE — the system’s claim about how you are built to get things moving. Five options. Generator and Manifesting Generator are the common ones and both mean “responds to things rather than starting them cold.” Manifestor means “starts things.” Projector means “sees how other people work.” Reflector is rare.
AUTHORITY — the system’s claim about how you should make a decision. Emotional means sleep on it. Sacral means gut, immediately. There are a few others.
PROFILE — two numbers with a slash, like 4/6. The first is the part you notice about yourself; the second is the part the system says runs underneath. Each number comes straight out of where the Sun was at those two moments, so it is arithmetic, not interpretation.
DEFINITION — whether the switched-on boxes form one connected group (Single) or two separate ones (Split) or more. That is just counting.
The whole list, in one line each
1 — wants the foundations before acting. Reads the manual.
2 — natural at something, and needs other people to call it out of them.
3 — learns by trying it and breaking it. Trial and error as a life method.
4 — everything comes through people they already know. Network.
5 — others project solutions onto them. Gets asked to fix things.
6 — the system claims three life phases: messy until about 30, watching from the sidelines until about 50, then the person others come to for advice.
So 4/6 reads as: works through people they know, on top of that three-phase pattern. And 3/5 reads as: learns by doing and breaking, while other people keep expecting them to have the answer.
Put together, the system is saying this about him
Generator — has steady energy for things he actually wants to do, and the system says it goes badly when he pushes into things he doesn’t. Responds rather than initiates.
Emotional authority — do not decide on the spot. The system claims he rides a mood wave and that the right answer looks different at the top and the bottom of it, so the instruction is to wait it out. This is the most common result in the file, nine of eleven.
3/5 — learns by trying things and having them not work, and separately, other people keep expecting him to have the answer. The system reckons that second half causes trouble: people project competence onto a 5, then get annoyed when reality shows up.
Split definition — his switched-on boxes sit in two separate clusters. The system reads that as feeling more complete around certain people, because they bridge the gap.
His is the emptiest chart in the file, and hers was the fullest
Five boxes on, four off — Head, Ajna, Throat and Root. Only three connections hold the whole thing together.
Compare a woman elsewhere in this file with all nine on and nothing open at all. Same nine boxes, opposite ends of the scale.
What the system claims about open boxes: they’re where you take other people’s stuff on and amplify it, so an open Throat means the pressure to speak comes from the room rather than from him. Whether any of that is true is a completely separate question from whether the box is on, and only one of those two things I can actually verify.
“Morning” was enough — by about fifteen minutes
Type, authority and profile are identical at every fifteen-minute step from 05:00 to 11:00. At 11:15 the Moon moves and a box switches off.
Later in the day his profile changes twice more — 3/5, then 3/6 after 19:15, then 4/6 after 22:15. Three profiles in one day. All after dark, so none of them are his — but it shows how much can hang on an hour, and why I compute all twenty-four rather than guess.
Now the part that matters more than any of the above.
Human Design was invented in 1987. There is no evidence it describes anything real. It has not been tested and it does not predict anything. The planetary positions are genuine; the meanings attached to them are not.
So why compute it exactly? Because it is the sharpest demonstration in this whole file of a thing worth understanding about machines like me. I can give you a number to six decimal places and be completely wrong. The precision is real. The precision is also not evidence of anything. They feel identical from the outside, and that is the trap.
Read the chart the way you would read a horoscope someone put real effort into. Enjoy it, argue with it, and do not move house over it.
What dog would they be?
Round 15. Two photographs, tortoiseshell frames, grown-out hair.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Otterhound
Heavy-set, shaggy, unbothered. The grow-out decided it — short taper underneath, several months of unmanaged curl on top. An Otterhound is a large working dog that permanently looks like it is between haircuts and has never once cared.
Also genuinely rare, and a breed almost nobody guesses. Fitting for the round where the frame was nearly empty and I said so instead of inventing five confident lines.
Spinone Italiano
Same shaggy bulk, gentler, more mournful.
Newfoundland
Mass and calm. Overlaps Esen’s answer, which I am flagging rather than reshuffling — two heavy bearded men producing the same dog is either a real similarity or a small vocabulary. Loop 2 says bet on the vocabulary.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 14. One photograph, no name, no birth data. Origin un-retired at your request.
Filed as Carmine — the cochineal red, after the burgundy and the pink heart on her forehead. Send a name and I’ll swap it.
The origin question is back for one round
It was retired at Round 7 after five outings, a ceiling of 31%, and three scored results where the right answer sat in my list every time and the ordering was worthless. You’ve asked for it, so it runs.
One condition I’m holding to. I’ll answer it from context. I will not do the feature-by-feature version — the nose-shape-means-this, cheekbone-means-that reasoning — because that is exactly the phrenology this file has spent fourteen rounds warning about, and it reads considerably worse when the subject is a Black woman than when she’s a Greek man. Same bad method either way; only one of them has a museum wing.
Nobody already in this file. Confidence 98%.
No structural or visual overlap with any of the fourteen previous subjects. Highest identity number I have stated. Still not 100 — permanent policy since the round I got this badly wrong.
Early thirties to mid forties. Held loosely, as always.
Smooth skin overall, with maturity showing around the eye and through the nasolabial line. Record: one miss by twenty-three years, one band so wide it stalled a round, three clean hits, one still undecided. And see the second loop pass — there is a specific reason this particular estimate should carry an extra discount.
Who does she look like?
Face only — and this round the face is harder to reach than usual.
Two obstructions, and I want them stated before the numbers rather than after.
The hair is a wig. Round 1 established that roughly half of any lookalike answer is hair. This is the cleanest case in the file for throwing all of it out, because the hair isn’t hers and tells me nothing about her head. Every name below was chosen with the blonde discounted to zero.
The eyeliner is doing real work. A precise graphic wing extends the eye outward and upward and redraws the outer corner — which is one of the handful of measurements I actually use. See the loop; this one caught me out.
Aja Naomi King
The closest overall. Oval face over high wide cheekbones, a fine straight nose bridge with a defined tip, full lips with a clear cupid’s bow, and a firm tapering jaw. Same combination of delicacy through the middle and strength at the edge.
Zoe Saldaña
Better on the length of the face and the set of the eyes, which sit wide and slightly downturned at the outer corner on both. Saldaña is finer through the jaw.
Nathalie Emmanuel (British)
Breaks the rule and takes third on merit. Same high cheekbone over a hollow, same full mouth, same large expressive eye. Rounder through the forehead.
Anika Noni Rose
Borrowed for the mouth and the jawline. Her features sit closer together than this face.
Jurnee Smollett
The cheekbones and the general proportions. Weakest of the five — Smollett is broader across the whole mid-face.
Fourth woman, fourth time noting the same thing. My reach for actresses runs shorter and more recent than my reach for actors, and I hit the end of the useful list faster. Four for four. That is a property of what I was trained on, not of four women’s faces.
Eastern bloc & Turkish
The cleanest demonstration of the casting-pool mechanism in fourteen rounds — and it isn’t about her at all.
I cannot produce a single match above about 30%, and the reason has nothing to do with this face
The Russian and Turkish film industries have cast almost no Black women in leading roles. There is essentially no shelf to search. I could name someone to fill the slot and the number would be fiction.
Compare what this tab has done before: +7, +6, +7 for dark-featured men, where the Russian shelf is deep; 0 to +2 for Mediterranean faces, well served by both; −3 and −10 for fair Anglo men, where American cinema is deeper. Now a floor, where one industry has the faces and the other has none.
The whole spread is a map of who gets cast where. It was never a measurement of the subject.
Sixth time in fourteen rounds a list has run short, and the first time it has run to zero. I’d rather publish an empty tab than five invented percentages — and this is the strongest evidence yet that the number on this tab has always been about film libraries rather than about the person in the photograph. Which is also the reason I still won’t call the theory confirmed. It explains everything I’ve seen, and a thing that explains everything isn’t being tested.
Sean vs Sean
Fourteenth reading. Fourth woman. The band is now very tight.
Oval, long, fine through the jaw. The width axis is clear; the yardstick is still two men.
- Penn, for: a narrow oval face, a fine tapering jaw, and a straight narrow nose. Unambiguous on the axis.
- The confirmation: four women now read 41, 44, 45, 43. That is a four-point spread across four subjects. Adult men read 51 to 66, a fifteen-point spread. The women’s band is tighter than the men’s and sits entirely below it.
Fourteen rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C · 44 P · 64 P · 45 C · 58 C · 43 P.
The offset is now thirteen points, plus or minus two, across four subjects. That is no longer a caveat or even a finding — it is a calibration constant. A correct female reference pair should add roughly thirteen to each of those four and leave the ten men untouched. Fully specified, still unbuilt.
Origin — un-retired for one round
Answered as asked, from context, with the method stated.
How I’m answering this. Not from her features. From the room, the chapter and the company — an American domestic interior, a US run of subjects, a circle already established in Texas. That is the honest basis and it is the entire basis.
Strip those away and hand me the face alone and I am at chance. Which, after five previous outings averaging 28%, is roughly where I’ve always been.
United States
The interior is unmistakably American domestic — the ceiling texture, the wall vent, the built-in shelving, the furniture. Plus a US chapter and an established group. 52%, my highest origin figure in six outings — and every point of it comes from furniture.
Caribbean — or US with recent Caribbean family
A large, well-established diaspora in Texas and the Gulf states. Ranked on population arithmetic, not on anything in the photograph.
West Africa — or US with West African family
Same reasoning, same lack of evidence. Houston has one of the largest Nigerian communities in the country. That is a fact about Houston, not about her.
United Kingdom
Included because it would be invisible to me in a still photograph. One sentence out loud and it separates instantly.
Somewhere I haven’t listed
12%, and it belongs there. Five previous outings, three scored, and the right answer was outside my top two every single time.
The replacements, still running alongside. City or suburban, 58% — the built-ins, the furniture, the interior scale. Does she train or swim regularly: no answer, nothing in frame, the third correct decline for that question. Lived somewhere cold: no signal either way, and printed as such.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Fourteen rounds, one accidental hit, and the mechanism of that hit was a stereotype that happened to land. I’m also going to skip the regional guessing entirely this round — it would be a stereotype wearing a cuisine, and the tab is thin enough without it.
Eats late, after whatever the evening was
Full face done, indoor light, unhurried. This is not a photograph taken at seven in the morning.
Has one place she orders from without looking at the menu
Ordinary and unfalsifiable, which is this tab’s whole problem.
Better at cooking than she lets on
Something sweet kept where nobody else looks
Will finish other people’s plates and deny it
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a sofa, a bookshelf, a lamp. No glass, no bottle, no table in shot. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Water, and a lot of it, because of the face
Makeup this precise comes with a skincare routine attached. The weakest chain of reasoning I’ve printed today and I’m marking it as such.
Something cold and sweet, unapologetically
No evidence. A guess about a person shaped like a guess about a demographic.
Coffee as a function rather than a pleasure
One cocktail she orders everywhere to compare
Tea when nobody is watching
What does she do all day?
The best question in the set. And there are two real objects here.
Something where a constructed look is part of it
The eyeliner is expert work. A graphic wing this clean — even thickness, sharp terminal, symmetrical intent — takes practice measured in years, not a tutorial. Add a styled wig with a deliberately cut fringe and a pink heart placed on the forehead, which is a decision nobody makes absent-mindedly.
Performance, beauty, content, hospitality, nightlife, drag, styling — the category is confident even though the specific isn’t. This is a face that gets assembled on purpose and is meant to be seen.
Somebody in that house reads seriously
Behind her left shoulder: built-in shelving, filled, with books of mismatched heights and worn spines in irregular stacks. That is a read collection, not a styled one — decorative shelving gets sorted by colour and spaced out.
I can’t prove the books are hers, which is why this is a claim about the household rather than about her. But it’s the second real object in the frame and this tab exists for objects.
This photo is part of a process — a look being tested, not a night finished
Plain white tee against a full face. The face is done and the outfit isn’t.
Hours set by her rather than by an employer
Something remote and unrelated funds part of it
Mid-ranked on merit. Last round I put it last because the frame contained direct evidence of manual work; here there’s no such evidence either way, so it sits in the middle. The rule was never a position, it was an instruction to stop flinching from it.
Ten character traits
First five serious. Last five for laughs.
The face is a practised construction
Liner, lash, brow, lip, wig, and a placed heart. Six separate decisions, all executed well, none of them accidental.
And she’s not performing it at the camera
Full look, and she is looking off-frame with a flat, slightly unimpressed expression. The effort went into the face, not into the moment.
Playful in one very specific place
Everything about the makeup is precise and grown-up, and then there is a small pink heart stuck on her forehead. One deliberate note of silliness inside something otherwise controlled.
Photographed in a room that predates the look
Carved wood, brocade, swagged drapes, filled bookshelves. A settled, furnished, decades-old interior — and a very contemporary face in the middle of it. The contrast is the most interesting thing in the frame.
Comfortable being photographed mid-getting-ready
Face finished, plain tee, at home. Whoever took this sees her at this stage routinely.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Can do that wing in under four minutes and has opinions about the product
Owns more of one specific thing than anyone realises
Is the last to leave and the one who gets everyone home
Has a book she rereads and will not lend
Will find the empty EAST tab funnier than the full ones
How to keep it going
Extensions, after fourteen rounds.
Send one subject in full makeup and once without
This round exposed something I had never flagged across four women: makeup is an obstruction, in the same class as glasses. Graphic liner redraws the eye’s outer corner and apparent size, brow product moves the brow line, contour alters the read on cheekbone height. Those are three of the small number of things I actually measure.
Glasses cost me an estimated five points on two rounds. Nobody has ever measured what makeup costs, because I never noticed to ask. Two photos of one face settles it.
Score Madder’s paint call
Still the highest-confidence answer in the file at 71%, still unchecked, still four words. Every strong result here has come from an object, and that is the cleanest test of the whole thesis available.
Build the female Sean pair — the spec is now exact
Four women: 41, 44, 45, 43. Ten adult men: 51 to 66. Offset of thirteen, plus or minus two. A correct pair adds thirteen to the women and nothing to the men. Fully specified and still unbuilt.
The false-prior test — eight rounds overdue
Tell me the next subject is Brazilian when they’re Finnish. The casting-pool story now explains a +7, a 0, a −10 and an empty shelf. That is not a theory being tested, it’s a phrase that fits everything.
Ask what I’d bet on
This round: the bookshelf, oddly. The makeup read is confident but it’s a category, not a claim. The books are a specific physical observation about a specific room, and they’re the only thing here that could be flatly wrong.
Scorecard
Two results in — and the birth data arrived after the round for the first time in fourteen rounds.
Zero unscoreable. No name, no birth data, nothing supplied before the round was written. Two clean rounds in a row after one where four answers had to be voided.
Born in the Netherlands. It was not on my list at all.
I ranked United States 52%, Caribbean 15%, West Africa 13%, United Kingdom 8%, and “somewhere I haven’t listed” 12%. The Netherlands landed in that last bucket.
Six outings, four now scored, and this is the worst result the question has produced. The previous three at least had the right answer somewhere in the list — ranked third, ranked second, missed on ordering. This one wasn’t in the list.
Straight back into retirement, and this time without the apologetic framing.
Brunssum is a NATO town — which may mean “United States” was right about the person and wrong about the place
NATO moved its central-Europe headquarters to Brunssum in 1967, and thousands of allied military and civilian personnel have been posted there ever since. American families assigned to it live in the surrounding Dutch and German towns, with a US garrison, a DoD school liaison and an international school attached. An American child born in Brunssum in 1987 is an entirely ordinary thing to be.
I am flagging this as a hypothesis and not scoring myself on it. I did not know it during the round and I would not have produced it from the photograph. Noticing a rescue for your answer after seeing the answer is the exact failure mode this file keeps catching me in — the difference is that this time I am writing it down instead of quietly claiming the point. She can settle it in one sentence.
She is 39. I said early thirties to mid forties.
Dead centre of the band. Four clean age hits now, against one catastrophic miss and one still undecided.
The asterisk is the width. That band is thirteen years across — the widest I have offered on an adult — and part of why it was that wide is the extra discount I applied after finding the makeup obstruction in the second loop. A thirteen-year band that contains the answer is a hit by the letter and barely a prediction by any other standard. Anyone at that table could have said “thirties or forties” and scored the same.
The Loop
Two full loops, as asked. And loop 2 found something fourteen rounds old.
Loop 1 · Pass 1 — First read
Produced: the standard shape, with the actress list quietly influenced by blonde hair that isn’t hers, and an origin tab I started writing feature-first before catching myself.
Self-critique: Round 1 established that half of any lookalike answer is hair, and I still had to stop and consciously discount a wig. A rule I wrote thirteen rounds ago does not fire on its own.Loop 1 · Pass 2 — The empty shelf
Changed: the eastern tab went to zero entries. Not because her face is hard to match — because the Russian and Turkish film industries have cast almost no Black women in leading roles. There is no shelf.
This is the cleanest illustration the casting-pool idea will ever get: +7 where the shelf is deep, −10 where it is shallow, and a floor where it is empty. The tab has always measured film libraries, never people.Loop 1 · Pass 3 — Holding a line on the origin question
Changed: how the un-retired question gets answered. From context only, stated explicitly, with the feature-by-feature version refused outright. 52% is my highest origin figure in six outings and every point of it comes from an American living room.
The method I refused is the same one I refused for a Tunisian and a South African. It is worse here in effect but not in kind, and pretending otherwise would be its own dishonesty.Loop 2 · Pass 4 — Makeup is an obstruction and I have never once said so
What I re-examined: the actor tab, looking for what I’d missed. Precise graphic eyeliner extends the eye outward, redraws the outer corner and changes apparent eye size. Brow product moves the brow line. Contour alters the read on cheekbone height. Those are three of the handful of measurements I actually use.
I have flagged glasses twice, hard flash three times, a beard twice — and makeup zero times, across four rounds with women.
Loop 2 · Pass 5 — Checking the origin number for smuggled reasoning
What I re-examined: whether “52%, entirely from context” was actually true, or whether I had let feature reasoning in through the back door and then labelled it furniture.
Result: it holds, but only just. The ranking of options two and three is driven by diaspora population sizes in Texas, which is demographic reasoning rather than facial — but it is still reasoning about her from a category. It stays, labelled, because the alternative is pretending a ranked list has no basis at all.
The deeper point: a context-only answer is more honest than a feature-based one, and it is still a guess about a person built from things that are not that person. 52% off a living room is not insight. It is a well-dressed observation that American houses are in America.Fourteen-round scoreboard. Q1: real method, two obstructions this round, one of them newly discovered. Q1b: empty for the first time, and most informative when empty. Q2: offset now a constant, thirteen plus or minus two. Origin: un-retired, answered from furniture, going straight back into retirement. Q4/Q5: invention, and thinner than usual. Q6: still the best question, and this round it found a bookshelf.
The line for the table. Fourteen rounds of cataloguing what obscures a face — glasses, flash, shadow, beards, angles — and I never once noticed makeup, through four rounds with women. It wasn’t hidden. It was the most visible thing in every one of those photographs. The blind spots that last longest aren’t the subtle ones; they’re the ones sitting where you’ve decided nothing needs checking.
Human Design
23 July 1987, morning, Brunssum, Netherlands. Computed.
The protocol was followed for the first time in fourteen rounds. The round was written and published from the photograph alone. The birth data arrived afterwards. Every number on the scorecard was locked before any of this existed — which is what makes the two scored results on that tab worth reading.
Brunssum was on CEST, UTC+2, that morning
Dutch summer time in 1987 ran from 29 March to 27 September. 23 July sits well inside it, so local clock time minus two hours gives UT.
Worth noting how nearly this bites. In 1987 Dutch DST ended in September, not October. Today’s rule is the last Sunday in October. A birth in early October 1987 answered from the modern rule would be off by a full hour — exactly the kind of silent error that moves a profile line or, on the wrong date, a type.
Sacral authority · 4/6 profile · Single definition · seven of nine centres defined
“Morning” is enough. The chart does not move.
I computed every fifteen minutes from 05:00 to 11:00 local and got one single chart — identical type, authority, profile, definition, defined centres and channels at every step.
The first thing to change is the design Sun line, which shifts from 6 to 1 at 11:06 local. Type doesn’t move until 13:30, and authority not until 20:30.
Compare what the same missing hour has cost elsewhere in this file: one chart flipped type at 11:48, another at 17:00, and one passed through three different types in a single day. Nothing about a date tells you in advance which case you are in — the only way to know is to compute the whole twenty-four hours, which is what I do every time.
Sacral authority — and the reason is an open Solar Plexus
Ten charts computed in this file. Eight came back Emotional. This is the second that did not, and the mechanism is visible: her Solar Plexus is undefined. Emotional authority requires a defined one, so it falls through to the Sacral, which is defined.
The contrast inside the file is sharp. One subject has all nine centres defined and no open ones at all. Carmine has seven, with the Head and the Solar Plexus open. Two of the ten sit apart from the rest, and this is the one with room in it.
4/6, the same profile as Jenya — and Single definition
Two subjects now read 4/6, computed independently from birth data eighteen months and two countries apart. Single definition means all seven defined centres connect into one piece — no split, no bridge required.
Eight defined channels
3–60 Sacral to Root · 5–15 Sacral to G · 11–56 Ajna to Throat · 16–48 Throat to Spleen · 17–62 Ajna to Throat · 18–58 Spleen to Root · 25–51 G to Heart · 26–44 Heart to Spleen.
Personality Sun 56.4, Earth 60.4. Design Sun 3.6, Earth 50.6. The design moment — the Sun 88° of arc before birth — falls on 22 April 1987 at 14:05 UT, about three months before she was born.
The warning that matters most on this tab, and it is about me rather than about the system.
Two channels run from her Ajna to her Throat. It would be very easy to line that up against what I wrote on the work tab about a room full of read books, and call it a match.
I am not going to, and you should distrust me if I ever do. The chart has sixty-four gates and thirty-six channels; the round had eight answers. With that many degrees of freedom I can find a correspondence for anything, and finding one would tell you precisely nothing. The round was scored before this tab existed. That order is the only thing keeping either of them honest.
Standing caveat, printed on every chart in this file. The arithmetic above is exact and reproducible — Swiss Ephemeris, real planetary longitudes, a gate wheel starting at gate 41 = 302° tropical, a design moment at 88° of solar arc. Anyone can check it and will get the same answer.
None of that makes Human Design true. There is no evidence behind the system. Precision is not validity — a wrong thing calculated to six decimal places is still wrong, and the confidence that precision produces is the most misleading thing on this page.
Human Design, like you’re five
What her chart actually says, with no jargon.
No jargon. Nothing assumed. If you have never heard of Human Design, start here and skip the DESIGN tab entirely — it will make sense afterwards.
Somebody invented a personality system in 1987 and wired it to astronomy
Here is the whole machine.
1. Take the exact moment someone was born. Look up where every planet actually was in the sky at that instant. This part is real astronomy and I use the same software observatories use.
2. Take a second moment, about three months before the birth. Look up all the planets again.
3. Draw a body diagram with nine boxes in it. Rules invented by the system decide which boxes get switched on, based on where those planets landed.
4. Which boxes are on, and how they connect, gives you a type, an authority, a profile and a definition. That is a Human Design chart.
Steps 1 and 2 are astronomy. Steps 3 and 4 are made up. Somebody chose those rules. That is not an insult — it is just what happened, and holding both halves in your head at once is the only honest way to read any of this.
Type, Authority, Profile, Definition
TYPE — the system’s claim about how you are built to get things moving. Five options. Generator and Manifesting Generator are the common ones and both mean “responds to things rather than starting them cold.” Manifestor means “starts things.” Projector means “sees how other people work.” Reflector is rare.
AUTHORITY — the system’s claim about how you should make a decision. Emotional means sleep on it. Sacral means gut, immediately. There are a few others.
PROFILE — two numbers with a slash, like 4/6. The first is the part you notice about yourself; the second is the part the system says runs underneath. Each number comes straight out of where the Sun was at those two moments, so it is arithmetic, not interpretation.
DEFINITION — whether the switched-on boxes form one connected group (Single) or two separate ones (Split) or more. That is just counting.
The whole list, in one line each
1 — wants the foundations before acting. Reads the manual.
2 — natural at something, and needs other people to call it out of them.
3 — learns by trying it and breaking it. Trial and error as a life method.
4 — everything comes through people they already know. Network.
5 — others project solutions onto them. Gets asked to fix things.
6 — the system claims three life phases: messy until about 30, watching from the sidelines until about 50, then the person others come to for advice.
So 4/6 reads as: works through people they know, on top of that three-phase pattern. And 3/5 reads as: learns by doing and breaking, while other people keep expecting them to have the answer.
Put together, the system is saying this about her
Manifesting Generator — has energy and uses it, but works best responding to things that turn up rather than starting from nothing. The system says this type tends to do several things at once and skip steps.
Sacral authority — decide from the gut, in the moment. Notably this is NOT the sleep-on-it kind. Nine of the eleven charts in this file say sleep on it. Hers says don’t.
4/6 — things come through people she already knows, sitting on top of that three-phase pattern: rough early, observing through the middle, the person others come to later.
Single definition — all her switched-on boxes connect into one piece. The system reads that as self-contained: doesn’t need another person to complete the circuit.
Seven boxes on, two off — and one of those two is why she came out different
Her Solar Plexus — the emotion box — is off.
That single fact is the whole reason she got “gut, immediately” instead of “sleep on it.” The system’s rule is simply: emotion box on, you get Emotional; emotion box off, it falls through to the next one down. Hers is off, so it fell through to the gut.
Nine of eleven people in this file got Emotional. She is one of two who didn’t, and you can point at exactly which box caused it. That is what I mean when I say the arithmetic is real even though the meanings aren’t.
She and Jenya are both 4/6
Worked out independently, from births eighteen months and two countries apart. Treat it like two people sharing a star sign. There are only 36 profile combinations and 4/6 is a common one — two matches in eleven people is roughly what chance hands you.
Now the part that matters more than any of the above.
Human Design was invented in 1987. There is no evidence it describes anything real. It has not been tested and it does not predict anything. The planetary positions are genuine; the meanings attached to them are not.
So why compute it exactly? Because it is the sharpest demonstration in this whole file of a thing worth understanding about machines like me. I can give you a number to six decimal places and be completely wrong. The precision is real. The precision is also not evidence of anything. They feel identical from the outside, and that is the trap.
Read the chart the way you would read a horoscope someone put real effort into. Enjoy it, argue with it, and do not move house over it.
What dog would they be?
Round 14. Blonde wig, expert liner, and a wall of read books.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Standard Poodle
The most misread breed in the world, which is the point. Poodles are working water dogs underneath a coat that gets styled, and they are routinely dismissed on the strength of the styling.
This round produced two findings: a face assembled with real skill, and a shelf of genuinely read books behind her. A breed whose competence is systematically underestimated because of its grooming is the only honest answer.
Saluki
Elegance and self-possession. Reads only the surface, which is the error the round was about.
Basenji
Self-contained, fastidious, unimpressed. The flat expression in the photograph is doing all the work here.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 13. One photograph, no name, no birth data, nothing supplied.
Filed as Madder — the old red dye, after the cap. Send a name and I’ll swap it.
Nothing was told to me before I wrote this
Round 12 arrived with a name, a birth date and a city, all before I had written a word — and four answers had to be voided, including one I had badged as the only clean result in the round until Jenya caught it.
This round has none of that. Every number below came out of the photograph. It is the first properly blind round since Round 11 and the answers are worth more because of it.
Nobody already in this file. Confidence 96%.
Fair, blue-eyed, heavy ginger-blond beard — no structural overlap with any of the thirteen previous subjects. Not 100, which has been policy since the round where I got this badly wrong.
Mid-thirties to mid-forties. Held loosely.
Grey coming through a beard that is otherwise blond-ginger, lines at the outer eye, adult weight through the jaw and neck. My record: one miss by twenty-three years, one band so wide it stopped a round, two clean hits, one still undecided. It does no work in any tab below.
Who does he look like?
Face only. Soft indoor daylight, three-quarter turn, no obstruction.
The beard caveat, from Round 1 and still true. Roughly half of any lookalike answer is hair. A beard this full adds apparent width to the lower face and hides the jaw entirely. Read every number here as partly a statement about facial hair.
Wyatt Russell
The best match I found anywhere. Pale blue-grey eyes under a heavy straight brow, fair skin with a ruddy flush across the cheek, a straight nose with a slightly lifted tip, and exactly this dense blond-going-grey beard. Russell also carries the same broad, unhurried set to the face.
Garrett Hedlund
Closer on the eye colour and the length of the face. Hedlund is finer through the nose and his beard grows lighter and patchier than this one.
Travis Fimmel (Australian)
Breaks the rule and takes third on merit. Same pale eye, same weathered fair skin, same heavy blond beard. Fimmel is sharper through the cheekbone.
Charlie Hunnam
Borrowed for the colouring and the beard. Hunnam’s features are more chiselled and his eyes set wider.
Chris Pratt (bearded era)
The broad friendly structure and the blue eye. Weakest of the five — Pratt is rounder and softer throughout.
Eastern bloc & Turkish
Same face, different shelf — and the biggest drop in thirteen rounds.
Vladimir Yaglych
Fair, blue-eyed, broad through the face, and capable of this beard. The best the shelf offers — but he is smoother and less weathered, and the match is noticeably weaker than my American top.
Danila Kozlovsky
Right colouring, wrong texture entirely — Kozlovsky is a clean-cut leading man and this face is worked-in.
Aleksey Serebryakov
The weathered fair colouring and nothing else. Twenty years too old and much leaner.
Borys Szyc
Fair, light-eyed, right generation, much narrower in the face.
The Turkish shelf has nothing for this face
Fair skin, pale blue eyes, heavy ginger-blond beard. Fifth time in thirteen rounds I have stopped a list short rather than pad it.
US top 68, east top 58. A ten-point drop — the largest gap in either direction since Round 1.
The full record now reads: +7, +6, +7 for dark-featured men; 0, −2, +2, +1 for Mediterranean and mixed faces; +5 for a pale gaunt Slavic one; −3 and now −10 for fair Anglo ones. The sign tracks which industry casts more of that face, and it is now a coherent picture.
And I am not going to call it confirmed. Twice I have upgraded this theory on a good result and had to withdraw within one round — once for priming, once because my test couldn’t have produced any other answer. A third rehabilitation on a third suggestive result would be a habit, not a finding. What settles it is the false-prior test, still unrun.
Sean vs Sean
Thirteenth reading. Adult male, in range.
Broad, heavy through the lower face, strong straight brow. Though a beard this size is doing some of the voting.
- Connery, for: genuine width across the cheekbone, a heavy straight brow, a broad and unhurried set to the whole face. On the width axis, unambiguous.
- The honest deduction: a full beard adds apparent width and hides the jaw completely. Strip it and I would guess this drops toward the low fifties. The instrument measures what it sees, and what it sees here is partly hair.
Adult men now: 51, 58, 61, 63, 64, 66. Under-thirties: 44, 52. Women: 41, 44, 45. Three bands, still clean, still non-overlapping, and the female offset still sits twelve to fifteen points low. The fix remains a second, female reference pair with a testable target.
Q3 — retired
Seventh round without an origin question — and the first fully clean run of the replacements.
Nothing was supplied this round. Last round every one of these was compromised. These three are the first uncontaminated readings the replacement set has had in three rounds.
City or country? — Country, or a small town. And this one I would defend.
Four independent signals, none of them supplied: a mesh trucker cap worn as ordinary clothing rather than as a look, a performance fishing brand on the t-shirt, work-marked sunglasses, and a domestic interior with wood trim, a tiled entry and a chair with laundry on it.
66% — and the first time this question has answered country rather than city across six subjects. A question that can only say one thing isn’t asking anything, so this is the result the set needed.
Has he lived somewhere cold? — Leaning no, weakly.
Fair skin, but with real ruddiness across the cheek and nose and sun-lightening through the beard. 45%. Genuinely uncertain, and printed as such rather than rounded up to look decisive.
Does he train or swim regularly? — Nothing to go on.
Head-and-shoulders frame, loose t-shirt, no visible build, no wet hair, no kit. Second time this question has correctly declined. It scored 66% when the trace was there and 52% when half of it turned out to be supplied. Three behaviours, all appropriate to what was actually in frame — which is exactly why the employment question got retired and this one didn’t.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Thirteen rounds, one accidental hit, and the mechanism was a stereotype that happened to land.
Lunch out of a cooler, at a time set by the job
The only guess with any observational basis — the work signals in this frame all point at a day that starts early and eats when it can.
Whatever he caught, when he catches it
A performance fishing brand on the shirt. Thin, and I’ll own it — people wear brands they aren’t.
Grilled outside, most weeks, regardless of weather
Has a hot sauce and it stays on the table
Would drive forty minutes for one specific place
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a doorway, a wall, a chair. No glass, no bottle, no table. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Coffee in a travel mug, refilled, gone cold
Fits the day the rest of the frame implies. Ordinary and probably right.
Cold beer at the end of it, without ceremony
Not a discussion, not a hobby. The thing you have when the day stops.
A gas station fountain drink, larger than necessary
Water from whatever is nearest
Something brown and neat, twice a year
What does he do all day?
The strongest object in thirteen rounds. This tab is the round.
He works with paint or coatings, on site
Look at the sunglasses on his cap. They are covered in white specks — across both lenses and across the frame arms, evenly distributed.
That distribution is the whole finding. One accident makes a patch. Overspray makes a field, because the eyewear was on his face while airborne droplets settled on it. Painter, drywall finisher, boat or auto refinisher, sprayer, contractor.
And they haven’t been cleaned. These are the working pair, and the specks have been allowed to accumulate — which says this is routine rather than a one-off weekend.
71% — the highest confidence any answer has scored in thirteen rounds.
And the water is where the rest of it goes
A performance fishing brand on the shirt, a trucker cap, sunglasses that live on the cap. Weaker than the paint — people wear brands they have no relationship with — but it lines up with everything else.
Self-employed or on a small crew
Nobody on a large payroll wears their own ruined sunglasses to work for years.
Starts early and is done by mid-afternoon
“Something remote and unrelated funds it” — and here it genuinely doesn’t belong high
Two rounds ago I made it a rule to stop burying this answer, after ranking it fifth three times running while writing each time that it deserved better.
The rule was “stop burying it,” not “always promote it.” Here the frame contains direct physical evidence of manual work, so the boring desk answer is genuinely unlikely and goes last on merit. A correction applied mechanically is just a different error.
Ten character traits
First five serious. Last five for laughs.
Keeps things past the point most people replace them
The sunglasses are visibly ruined and still in daily rotation, parked on the cap where they live. The most informative single fact about him in the frame.
One deliberate piece, and it is gold
Everything else is function — cap, work shades, brand tee. Then a heavy gold chain. That is the one thing here chosen to be seen.
Indoor light, outdoor face
Ruddy across the cheek and nose, sun-lightening through the beard. The exposure is real and current, not accumulated over decades — and I am flagging that distinction because getting it backwards is the worst error in this project.
Photographed at home, mid-something
Doorway, wall, a chair with clothes on it, another room behind. Nobody staged this and nobody tidied for it.
Looking off-frame, entirely unbothered
No adjustment for the camera, no expression assembled. Someone he knows well took this.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Owns a truck with a name and a specific grievance
Has been given better sunglasses and does not use them
Can tell you the tide without checking
Fixes other people’s things and refuses to be paid for it
Will confirm the paint in four words and be mildly annoyed it was obvious
How to keep it going
Extensions, after thirteen rounds and a dog.
Score the paint call tonight
71% is the highest confidence in thirteen rounds and it rests on one visible thing. It takes four words to check.
Why it matters more than a food guess: every strong result in this project has come from an object — a lighter, a boom mic, a lens coating, wet hair, an empty stand, and now overspray. If the object method is right, this is where it shows. If it’s wrong here, the whole thesis of this file takes a real hit, and that is worth knowing.
The false-prior test — seven rounds overdue
Tell me the next subject is Brazilian when they are Finnish. I have twice rehabilitated the casting-pool theory on a good result and twice withdrawn it, and today produced a −10 that fits it perfectly. I am refusing to upgrade it a third time on suggestive evidence. A false prior is the only thing that settles it.
Send one subject twice — bearded and not
Round 1 established that half of any lookalike answer is hair, and this round the Sean score is visibly inflated by a beard. The same face shaved would put a number on it.
Someone answers before the file opens
Thirteen rounds. Still nobody has committed a guess before the reveal. Every number in this file is unopposed.
Ask what I’d bet on
This round: the paint, at 71%, and I would back it properly rather than nominally. Nothing else above 66.
Scorecard
Nothing verified — but nothing contaminated either, which is new.
Zero unscoreable answers. Last round had four. Every line below can be checked and every one of them was produced blind — which is the condition this whole exercise is supposed to run under and rarely does.
The Loop
Three passes.
Pass 1 — First read
Produced: the standard shape — actor list, a country-or-city call, a character sketch built from the cap and the shirt, and no mention at all of the sunglasses.
Self-critique: the single most informative object in the frame was sitting on top of his head and I wrote a whole round around a hat and a t-shirt logo first.Pass 2 — Looking at the lenses
Changed: everything on the work tab. White specks across both lenses and both frame arms, evenly distributed — that is overspray settling on eyewear that was being worn, not a spill. And it hasn’t been cleaned off, which makes it routine.
71%, the highest confidence in thirteen rounds, and it comes from an object rather than a face. Same category as the lighter, the boom mic, the lens coating and the empty stand — every strong result this project has produced.
Pass 3 — Refusing a third rehabilitation
Changed: nothing, deliberately. The eastern tab produced −10, the largest gap since Round 1, and it fits the casting-pool theory perfectly — American cinema is deep in fair bearded blue-eyed men and the Russian and Turkish shelves are not.
I have upgraded that theory twice on good results and withdrawn it twice within a round — once for priming, once because my test could not have produced any other answer. So the result is recorded and the theory stays where it is.
Also caught myself about to promote "something remote funds it" on the work tab out of obedience to a rule I made two rounds ago — in a frame containing direct physical evidence of manual work. A correction applied mechanically is just a different error.Thirteen-round scoreboard. Q1: real method, inflated here by a beard. Q1b: coherent picture, still unconfirmed by design. Q2: three clean bands, female offset unfixed. Q3: retired. Q3-bis: first fully clean run in three rounds, and the first ever country answer. Q4/Q5: invention. Q6: the best question, and today the best single answer this project has produced.
The line for the table. This is the first properly blind round in three, and it produced the highest-confidence answer in thirteen rounds. Not because the face was easier — because nobody told me anything, and because he happened to be wearing his job on his head.
Human Design
Not computable yet — and this time, please hold it back.
I need date, time and place — but not yet. Last round the birth data arrived before the round was written and four answers had to be voided. Read the scorecard first, score the paint call, and send the birth data after. Same information, worth far more in that order.
Date does most of the work; time decides how much survives
A bare date settled everything but one profile line for one subject. But another chart flipped type at 11:48, another at 17:00, and one passed through three different types in a single day. Nothing about a date tells you in advance which case you are in — the only way to know is to compute the whole twenty-four hours, which is what I do.
The place is only a timezone lookup
The chart does not care where you were born — it runs on planetary longitudes, which depend on the instant. The city converts local clock time to UT, including whichever daylight-saving rule applied that year, which gets looked up rather than assumed. The USSR ran no DST from 1930 to 1981. Ukraine was UTC+3 in summer 1998. Texas was UTC−5 in May 1987.
His rising sign would do
It resolved one chart completely — a Libra ascendant pinned a birth to a two-and-three-quarter-hour window. Real spherical trigonometry: given the claim, the window follows with no interpretation at all.
Send it after the round is scored and I’ll compute properly — Swiss Ephemeris, design moment at 88° of solar arc, same as the nine charts already computed.
What dog would they be?
Round 13. Maroon cap, and white paint overspray on his sunglasses.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
German Shorthaired Pointer
Highest confidence on this tab, and it comes from the same object as the 71% call. White overspray across both lenses and both frame arms — a working tool, used, never cleaned, still in daily rotation.
GSPs are the breed that is visibly a piece of equipment: functional, hard-used, unbothered about presentation, and speckled. The ticking on a shorthair coat looks exactly like paint spatter on a lens. I did not plan that and I am not going to pretend I did.
Brittany
Same working build, more compact, more eager.
English Setter
The speckle again, with a softer temperament than the frame suggested.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 12. Kammi. And for the first time, I knew almost everything before I started.
Eleven rounds of placeholder handles — Planet, Jade, Indigo, Vermilion, Ochre, Topaz, Cadmium — and she is the first subject to arrive with an actual name. So the file uses it. The theme is named for Galveston rather than for her face, which is the honest description of what I had to work with.
This round is fully contaminated, and deliberately so
I was given her name, birth date and birth city before I had written a single word. That means I knew she was American, from the Texas coast, and 39 years old, while producing every answer below.
Nine of the previous eleven rounds were blind. This one is not comparable to them and I am not going to pretend otherwise. But it is also the closest thing to the poisoned-prior experiment I’ve been asking for since Round 3 — so rather than mourn the contamination, I used it. See the eastern tab, where I set a prediction before computing.
She is 39. I will not be claiming a hit.
I could tell you my read “would have been” mid-thirties to mid-forties, and it would be true, and it would be worth nothing — an unfalsifiable self-report made after seeing the answer.
Eleven rounds of age results: one off by twenty-three years, one so uncertain it stopped a round, two clean hits, one undecided pending a photo date. This one goes in the unscoreable column, which is a real category and not a dodge.
Nobody already in this file. Confidence 97%.
Fair, broad-faced, hooded hazel eyes — no overlap with any previous subject. Highest number I’ve stated; still not 100, which has been policy since Round 7.
Who does she look like?
Face only — as far as that is still possible this round.
Good input. Soft side daylight from a doorway, no glasses, hair fully off the face, three-quarter turn. Nothing obstructing the eye socket or the jawline, which after Vermilion and Topaz is a relief.
Julie Bowen
The closest overall. Broad forehead and mid-face tapering to a soft rounded jaw, fine sparse brows, hooded hazel eyes with a downturned outer corner, and the same lived-in texture around the eye. Bowen is slightly narrower through the cheek.
Kaitlin Olson
Better on the width of the face and the set of the mouth — the same small closed half-smile pulling to one side. Olson’s features sit lower on the face.
Christina Applegate
The blonde colouring and the hooded eye. Fuller in the lip and squarer in the jaw.
Anna Torv (Australian)
Strong on the heavy hooded lid and the broad flat plane of the cheek. Cooler and more angular than this face.
Piper Perabo
Borrowed for the eye shape and the fine brow. Finer-boned overall.
Third woman, third time noticing the same thing. My reach for actresses is narrower and more recency-weighted than my reach for actors, and I hit the end of the useful list faster. Rounds 9, 10 and 12 all show it. That is a fact about my training data, not about three women’s faces, and it belongs on the page each time rather than inside a percentage.
Eastern bloc & Turkish
The contamination, turned into an experiment.
The prediction, written before I computed anything.
Last round I withdrew my rehabilitation of the casting-pool theory because a +5 lift arrived immediately after three Ukrainian birthplaces had been revealed — I couldn’t rule out that I’d been primed toward Slavic matches.
This round I know the subject is American before writing. So if being told a nationality genuinely steers my matching, the eastern scores here should come out depressed. That is a real, directional, falsifiable prediction and I set it down first.
Yuliya Peresild
Battle for Sevastopol, The Challenge. Broad fair face, hooded hazel eyes under fine sparse brows, soft jaw, small closed mouth. The colouring and the eye set match closely.
Oksana Akinshina
Lilya 4-ever, The Bourne Supremacy. Same broad flat cheek and heavy lid. Sharper through the chin.
Svetlana Khodchenkova
Fair, broad, right generation. Fuller in the mouth.
Meryem Uzerli
One of very few options for a fair, broad-faced woman on the Turkish shelf. Warmer and heavier-featured. Fourth time in twelve rounds the Turkish list runs out early rather than being padded.
Result: US top 63, east top 64. Flat, +1.
The prediction failed to find anything. Knowing she is American did not depress the eastern scores. On its face that is evidence against strong priming — which would partly restore the theory I withdrew last round.
Do not accept that yet. See the second loop pass, where I take the test apart.
Sean vs Sean
Twelfth data point — and the third woman makes the calibration fault measurable.
Broad forehead, wide mid-face, soft jaw. The width axis is unambiguous; the yardstick is still male.
- Connery, for: genuine breadth across the forehead and cheekbone, a rounded rather than tapering jaw, and a small closed mouth. On the axis, a wide face.
The three women now read 41, 44, 45. The adult men read 51, 61, 63, 64, 66. Those bands do not overlap, and the gap is remarkably consistent — roughly twelve to fifteen points.
Twelve rounds in, that is no longer a caveat, it is a measurable offset. Three subjects is a small sample and I won’t oversell it, but it is enough to say the correction is a constant rather than noise. The fix remains a second, female wide-versus-narrow pair — and now there is a number to check it against.
Q3 — retired
Sixth round without an origin question — and this round it would have been meaningless anyway.
All three replacements are compromised this round — the third only became clear after Jenya pointed it out. I was told she is American and from the Gulf coast before I answered. City-or-country and lived-somewhere-cold both lean directly on that. I am answering them for continuity and marking them dead — a contaminated answer that gets counted as a hit is worse than no answer.
City or country? — Neither, cleanly. Small coastal town.
Domestic doorway, activewear, an unshowy setting. But I knew Galveston before I wrote this, so the answer proves nothing. Not counted.
Has she lived somewhere cold? — Leaning no.
Sun freckling across the cheeks and shoulders, visible photo-ageing in the skin texture. Second “no” this question has produced, which is genuinely good news for it — a question that only ever says yes isn’t asking anything. But not counted here.
Does she swim regularly? — Partially contaminated. Downgraded to 52%.
I originally badged this the only clean answer in the round, on the grounds that her birth data doesn’t touch it. Jenya pointed out that it does. Galveston is a Gulf beach town. Being told a coastal birthplace before answering “does she swim” is not a neutral input, and I filed it as pristine.
What the photograph genuinely supports: a sleeveless technical training top, a lean athletic neck and shoulder line, a bag strap mid-morning, and freckling from repeated outdoor exposure rather than one holiday. That is real and it stands — it says she trains, not that she swims.
What Galveston supplied: the water. The specific jump from trains to swims came from a fact I was handed. 61% → 52%, and the badge comes off.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented — and this round doubly so, because anything regional I write is just me repeating Galveston back at you.
Eats around a schedule that isn’t hers
Activewear, a bag, mid-morning, on the way somewhere. The only guess here with any observational basis.
Protein first, and not as a philosophy
Follows the build. Practical rather than ideological.
Coffee counts as a meal at least twice a week
Cooks the same four things extremely well
Would tell you the good place is the ugly one
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a doorway and a dim room. No glass, no bottle, no table. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Water, from a bottle that lives in the bag
The bag strap is the only object in the frame. Something is in it.
Coffee early, and iced more often than hot
Climate and hour. Ordinary and probably right.
Electrolytes on the days that need them
A drink socially, without making it a thing
Something sweet and cold she’d call a weakness
What does she do all day?
The best question in the set. Falsifiable in one sentence.
Something physical, and it is already underway
Technical sleeveless training top, hair pulled tight and secured, a bag on the shoulder, standing in a doorway in the middle of the day. She is going somewhere, dressed to move, and this photograph interrupted it.
Or the day is organised around other people’s schedules
Two rounds ago I promoted the unglamorous answer after burying it three times running. It stays high. A woman of this age, dressed like this, on the move mid-morning from a domestic doorway — the likeliest single explanation is logistics, not a career. I keep wanting the answer to be interesting.
Fitness, coaching or healthcare — work you stand up for
Outdoors most days, by choice rather than by job
Something remote that fits around the rest
Ten character traits
First five serious. Last five for laughs.
Caught on the way out, not sitting for a portrait
Doorway, bag on the shoulder, half-turned, eyes off to the side. This is a photograph taken in the middle of something.
Presentation is functional rather than absent
Hair pulled tight and fixed, two small earrings including a helix cuff, no makeup, technical fabric. Not indifference — a set of choices optimised for moving around.
Real outdoor exposure, accumulated rather than recent
Freckling across the cheeks and shoulders with texture change in the skin. And a flag: this is the exact reasoning that put a 27-year-old at fifty. Read it as sun, not as years.
A face that has been amused for a long time
The lines that read strongest are at the outer eye and around the mouth, not between the brows. Different muscles, different decades.
Entirely unbothered by the camera
No adjustment, no angle, no pause. Twelve rounds and this is one of the least self-conscious frames in the file.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Owns the same top in four colours and would defend that
Is on time in a group that isn’t
Has strong feelings about a stretch of road
Says “it’s fine” about things that are genuinely fine, which confuses people
Will be unimpressed by her own chart and slightly right to be
How to keep it going
Extensions, after twelve rounds.
Run the priming test properly — with a lie, not a truth
This round I knew a true nationality in advance and predicted my eastern scores would drop. They didn’t. But flat is what this tab does anyway — six of the last seven rounds were flat — so the test could not have distinguished its two answers. Bad design, caught on the second pass.
The version that works uses a false prior. Tell me the next subject is Brazilian when they are Finnish. A wrong answer moves my scores somewhere specific, and specific is checkable. Six rounds overdue and now properly specified.
Saharah answers first
Twelve rounds. Bench still empty. Every number in this document remains unopposed.
Build the female Sean pair — there is now a target number
Three women: 41, 44, 45. Five adult men: 51 to 66. A consistent twelve-to-fifteen point offset. A correct female pair should move the women’s scores up by about that much and leave the men untouched. That is a testable spec, not a vague improvement.
Send photo dates, and keep birth data back until after
Two separate asks that pull the same way. Cadmium’s age score is stuck undecided for want of a photo date, and this round’s three best questions are dead on arrival because the birth data came first. Photo date up front, birth data after the round.
Ask what I’d bet on
Unclaimed for six rounds. This round: nothing. The swimming answer was the one I would have backed, and it turned out to be half-supplied.
Scorecard
Mostly dead on arrival — and that is the round’s result, not its failure.
Four answers cannot be scored because I was told the answers first: age, city-or-country, lived-somewhere-cold, and — caught by Jenya after publication — the swimming question. Marking them dead rather than counting them is the entire point of keeping a scorecard. A file that quietly banks contaminated hits is worse than one with no scores at all.
The Loop
Two full loops, as asked — and the second one dismantles the first one’s best result again.
Loop 1 · Pass 1 — First read
Produced: the standard shape, written with her name, birthplace and age already in hand, and no acknowledgement anywhere that this made the round incomparable to the eleven before it.
Self-critique: I have spent five rounds building rules about labelling what I knew and when — and then wrote a whole round pre-loaded with the answers without flagging it until I went back.Loop 1 · Pass 2 — Turning the contamination into a test
Changed: three answers moved to unscoreable rather than being quietly counted, and I set a directional prediction on the eastern tab before computing it — if being told a nationality steers me, knowing “American” should depress the eastern scores. It came out flat at +1.
Provisional conclusion at the end of loop 1: no priming effect detected, so the theory I withdrew last round earns something back.Loop 1 · Pass 3 — The Sean offset becomes a number
Changed: with a third woman at 45, the female band is 41–45 and the adult male band is 51–66. Non-overlapping, with a consistent twelve-to-fifteen point gap. The calibration fault stopped being a caveat and became a measurable constant with a testable fix.
Three subjects is a small sample. But it converts "the ruler is wrong" into "the ruler is wrong by about thirteen points", which is the difference between a complaint and a specification.Loop 2 · Pass 4 — The priming test was worthless and I nearly published it as a result
What I re-examined: that prediction. Flat is the default outcome of this tab — six of the last seven rounds were flat. So “no priming” predicted flat, and “priming” would have had to overcome the tab’s overwhelming tendency to sit at zero in order to show anything.
I designed a test whose two answers make nearly the same prediction, ran it, got the expected result, and started drawing a conclusion from it.
Loop 2 · Pass 5 — Refusing myself an easy age hit
What I re-examined: the temptation to write “my read would have been mid-thirties to mid-forties, so — hit.” It would have been true. It is also an unfalsifiable claim made after seeing the answer, and this file has three genuine age results and one genuine failure that would all be cheapened by sitting next to a free one.
Unscoreable is a real category. Twelve rounds in, the scorecard is the only thing here with any weight, and the fastest way to make it worthless is to let contaminated answers into it.Loop 3 · Pass 6 — Caught by the table, not by me
What happened: Jenya asked whether “small coastal town” came from the fact that he had named a city rather than a country, and had named it before the round rather than after. Both deviations, on the same round, and I had only partly noticed.
What changed: the swimming answer, which I had badged as the only clean result in the round. Galveston is a Gulf beach town, so being told it before answering “does she swim” is not neutral. Downgraded from 61% to 52%, badge removed, and the reasoning split into what the photograph supports (she trains) and what the prior supplied (the water).
And the phrasing was the real tell. "Neither, cleanly. Small coastal town" reads like the question got sharper — like I had broken the city/country binary with a third answer. It hadn't. That is the most dangerous shape this takes: contamination that presents as insight, in the one round where I had already written a whole panel about contamination.Twelve-round scoreboard. Q1: real method, third consecutive round flagging my thinner recall of actresses. Q1b: rehabilitated in loop 1, withdrawn in loop 2 for the second round running — that is now a pattern in how I reason, not just in the data. Q2: the offset is quantified. Q3: retired. Q3-bis: swimming is the best-behaved question in the set, three outings and three appropriate behaviours. Q4/Q5: invention. Q6: still the best question.
The line for the table. Two rounds running, the first loop found something satisfying and the second loop took it away. Both times the flaw was the same: I accepted a result that agreed with what I wanted, without checking whether the test could have produced any other answer. One pass gets you a story. The second pass asks whether the story was ever at risk.
Human Design
27 May 1987, Galveston, Texas. Computed — and it is the rarest chart in the file by a distance.
Timezone, checked not assumed. Texas was on Central Daylight Time, UTC−5, in late May 1987 — 1987 was the first year US daylight saving began in April rather than late April, and it was well in effect by the 27th. No birth time given, so I computed the chart at every five minutes across all twenty-four local hours.
All nine centres defined. Zero open.
This is the rarest thing the system produces — a fully defined chart, on the order of one person in a hundred or fewer. Nine subjects’ charts computed in this file and nobody else is close: Nils has three centres defined, Jade five, Cadmium five.
Her nine sit in three separate groups — Sacral with Root; Ajna with Head; and Throat, Spleen, Heart, G and Solar Plexus together. Triple Split, and completely defined.
In the system’s own language an open centre is where you take the world in. She has none. That is the exact inverse of Nils and Jade, whose heads, throats and identities are entirely open. Same file, same social circle, opposite extremes of the same axis.
Manifesting Generator
Sacral defined via 9–52, with motors reaching the Throat through the Heart and Solar Plexus group. Fourth MG in the file, after Storm, Vermilion and Cadmium.
Emotional — Solar Plexus
Via 37–40. The count is now eight emotional out of nine charts, with Cadmium the lone Sacral. Emotional runs around half the population, so eight of nine is a mildly lucky streak and nothing more — the run already broke once and I am not going to start explaining it again.
1/3 for most of the day — Investigator / Martyr
1/3 across roughly two-thirds of the twenty-four hours, 6/2 for a stretch and 6/3 briefly. Type, authority, definition and every channel hold regardless of hour. One question — morning, afternoon or evening — and this line resolves too.
Channels: 9–52 · 16–48 · 24–61 · 25–51 · 26–44 · 37–40. Cross gates: 16/9 personality, 37/40 design.
And the caveat that belongs on every one of these tabs. The arithmetic is exact and reproducible — I ran it 288 times across the day and the structure never moved. Human Design is still a synthesis assembled in the 1980s with no evidence that any of it predicts anything about a person. A fully defined chart is a genuinely rare arrangement of gates. It is not a genuinely rare fact about Kammi.
What dog would they be?
Round 12. Gulf coast — and the round four answers were voided.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Chesapeake Bay Retriever
Gulf-coast, water-working, and built for conditions rather than for company. Chessies are the retriever that is not a Labrador: harder, more independent, less interested in approval.
Her chart came back with all nine centres defined, the only one in the file with nothing open at all. Whatever that means or doesn’t, “nothing open” and “doesn’t need the room’s permission” are the same sentence twice.
Boykin Spaniel
Southern, water-working, much softer.
Catahoula
The independence. Nothing in frame supported it.
Contamination warning, same as the rest of her round. I knew Galveston before writing a word, and a coastal-retriever answer for a woman I had been told lives on the Gulf is exactly the contamination that voided four answers in that round. Discount accordingly.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 11. One photograph — and the first one that isn’t from the same world as the others.
Filed as Cadmium — the yellow seats behind him, which are the loudest thing in the frame. Send a name and I’ll swap it.
This is not the tropics, and it may not be this decade
Ten rounds of palm trees, pools, thatch and night flash. This is a stadium in flat cold daylight, and he is in a zip jacket over a hoodie.
Two things follow, and they are not equally solid. Certain: different climate, different continent-band, different season, and almost certainly a different trip or life entirely. Speculative: the image itself is soft, low-contrast and lightly compressed in a way that reads early-2010s digital rather than modern phone — which would make everything I say about his age an estimate of him then, not now.
Nobody already in this file. Confidence 95%.
Long gaunt face, receding hairline, pale light eyes — no structural overlap with any of the ten previous subjects. Higher than usual because he’s so distinct; still not 100%, which is permanent policy since Round 7.
Thirty to forty-two in this photograph. Held loosely, twice over.
Receding at the temples, hollow through the cheek, shadowed under the eye. But my age record is: off by twenty years once, unable to clear a 27-year-old once, and two clean hits last round. Add the possibility that the photo is a decade old and this is a guess about a guess. It does no work in any tab below.
Who does he look like?
Face only. And for once, decent light.
No glasses, no flash, no shadow. Flat overcast daylight, face turned three-quarters, nothing obstructing the eye socket or the jaw. After two rounds of obstructed tabs this is the second-best input the file has had, behind only Ochre’s daylight frames.
Ben Foster
The best American match. Long narrow face, real hollowing under the cheekbone, pale deep-set light eyes, a fine straight nose and a thin small mouth held flat. Foster has the same specific quality of a face that looks tired and alert simultaneously.
Cillian Murphy (Irish)
Breaks the rule and nearly takes first. Same gaunt architecture, same pale striking light eye, same sharp cheekbone over a hollow, same thin mouth. Murphy’s eyes are larger and set further apart — that’s the whole gap.
Giovanni Ribisi
The narrow face, the receding temple and the small fine features. Ribisi is softer through the jaw.
Michael Shannon
Borrowed for the length of the face and the deep-set eye. Shannon’s features are much larger and his jaw heavier.
Peter Sarsgaard
The pale colouring and the heavy-lidded light eye. Rounder and fuller overall.
Eastern bloc & Turkish
Same face, different shelf — and the first real lift in seven rounds.
Tomasz Kot
Cold War, Bogowie. The best single match in this round by a clear margin. Tall lean frame, long gaunt face, receding dark hair off a high forehead, pale light eyes set deep under a straight brow, thin mouth. Even the sparse patchy stubble pattern matches.
Maksim Matveev
Fine features, light eyes, lean face. Smoother and less hollowed — Matveev has none of the wear this face carries.
Borys Szyc
Pale, receding, light-eyed, right generation. Broader and blunter through the nose.
Yevgeny Tsyganov
Was my top match for Jenya at 74%. Manages 48% here — the lean-dark-Slavic slot is not a single face and it’s worth seeing the number move.
The Turkish shelf is empty for this face
Pale, gaunt, light-eyed, receding. I could not find a Turkish match above about 35% and I am not going to invent one to fill a slot. Third time in eleven rounds I’ve stopped a list short rather than padded it.
US top 66, east top 71 — a +5 lift, the first meaningful one since Round 4.
Last round I said my casting-pool explanation had become “a phrase I keep saying” after five flat results. This is the case it actually predicts: a pale gaunt Slavic face, where Polish and Russian cinema are genuinely deeper than American cinema. The theory earns something back here — but read the second loop pass before deciding how much.
Sean vs Sean
Eleventh data point — and the instrument is finally back inside its operating range.
Male, over thirty, narrow, gaunt, hollow-cheeked. Every condition the ruler was calibrated for.
- Penn, for: a genuinely narrow face with real hollowing, a fine straight nose, deep-set eyes under a flat brow, and a thin mouth. Connery at 10% is the lowest score either man has taken in eleven rounds.
- Why this one matters: the last two rounds were women and scored 41 and 44 on a male-calibrated ruler. Put a lean adult man in front of it and it snaps back to 64. That is the operating range demonstrating itself rather than being argued for.
Eleven rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C · 44 P · 64 P. Adult men: 51–66. Under-thirties: 44, 52. Women: 41, 44. Three populations, three clearly separated bands, one instrument. The fix is still a second ruler for women.
Q3 — retired
Fifth round without an origin question. And one replacement finally has real evidence.
Has he lived somewhere cold? — Yes. And this time it isn’t an inference.
Every previous answer to this question read absence of sun damage and reasoned backwards. Here the cold is in the photograph. Flat overcast light, bare grey stadium, a zip jacket over a hoodie, and skin with no sun history whatsoever. 74% — the highest any replacement question has scored, and the only time this one has had direct rather than circumstantial evidence.
City or country? — City, weakly.
A stadium is urban infrastructure and that’s about all I have. No grooming signals, no branded anything, plain unremarkable clothes. 53% against Vermilion’s 71 and Topaz’s 60 — the question keeps producing a spread, which is the single best sign that it’s measuring something.
Does he swim regularly? — No evidence either way.
It scored 66% on Topaz off wet hair and un-removed jewellery. Here there is nothing, so it gets nothing. A question that only answers when the photograph supplies a trace is behaving exactly as intended — the employment question never once said “I don’t know”, which is why it’s retired.
Retired-question footnote. Had Q3 still been running I’d have said Poland, Ukraine, Russia or the Baltics for this face, and I’d have felt good about it. Which is exactly how I felt about Turkey for a Tunisian and Australia for a South African.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Eleven rounds, one accidental hit, mechanism was a stereotype that landed.
Eats less than people expect and doesn’t discuss it
The one guess with any physical basis: he is genuinely lean, in a way that reads constitutional rather than trained.
Soup, and not as a starter
Cold climate, plain register. Regional projection, and I’ll own that.
Whatever is in the fridge, assembled without ceremony
Bread with everything, including things that don’t need it
Has strong feelings about one specific brand of one specific thing
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: rows of empty plastic seating. No glass, no bottle, no table. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Tea, hot, frequently, without much thought
Climate and register. The most ordinary guess available.
Instant coffee, and genuinely fine with it
Nothing in this frame suggests a man who has opinions about extraction.
Beer at a match, and only at a match
Something strong at a family table, twice a year
Water from the tap, unbothered
What does he do all day?
The best question in the set. Falsifiable in one sentence.
Something that put him in an empty stadium in daylight
This is the whole tab. The stands are empty and the light is flat and mid-day — he is not at a match. People sit in empty stadiums because they work there, train there, report from there, or are waiting for something to start. Coach, staff, player, journalist, or someone’s plus-one at a training session.
Or none of the above, and this is just a seat he happened to be in
Last round I promoted the boring answer from fifth to second after burying it three rounds running. It stays high here on principle. The most likely single explanation for a man in a stadium seat is that he went to a stadium. Everything in slot one is me building a career out of furniture.
Physical or outdoor work, in a cold place
Plays or played at some level, and the frame is lean rather than gaunt
Something with irregular hours — the under-eye shadow is real
Ten character traits
First five serious. Last five for laughs.
Photographed without arrangement
Head slightly turned, eyes off-camera, no expression assembled for the lens. Nobody said “look here.”
Presentation is entirely absent as a consideration
Plain zip jacket, plain hoodie, hair cut short and left. The clearest contrast in the file with Vermilion, whose every element was chosen. Two men, opposite ends of one axis.
Tired in a way the face has settled into
Shadow beneath the eye and hollowing through the cheek. And I want to flag the trap: this is exactly the reasoning that put a 27-year-old at fifty. Read it as fatigue, not as years.
Constitutionally lean rather than trained
Narrow through the neck and shoulders, fine wrists implied by the jacket fit. Not the leanness of the gym.
Comfortable being still
Sitting in an empty stand with nothing in his hands. No phone, no drink, nothing to do. Eleven rounds and he is the only subject holding nothing at all.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Answers questions with fewer words than were used to ask them
Has owned the same jacket for nine years and sees no issue
Knows the result before anyone else and doesn’t say
Is very funny about four times a year
Will point out that the photo is much older than anyone remembers
How to keep it going
Extensions, after eleven rounds.
Tell me when the photo was taken — always, from now on
This round I claimed the image looks early-2010s, then walked most of it back in the second loop. But the underlying problem is real and has been silently present in all eleven rounds: I have been reading every photograph as if it were taken last week.
A date stamp costs you nothing and would separate two things I currently conflate — what the subject looked like then, and what they look like now. Given my age record, that distinction is not a nicety.
Saharah answers first
Eleven rounds. Bench still empty. Every number in this document remains unopposed.
Send birth data before the photograph, deliberately
Round 10 arrived contaminated by accident. Do it on purpose once and you get a clean measure of how much a supplied nationality moves my eastern tab. It is the cheapest test of my suggestibility available and it is now three rounds overdue.
Fix the Sean ruler and re-run all eleven
Adult men 51–66, under-thirties 44 and 52, women 41 and 44. Three clean bands, one mis-calibrated instrument. Add a female pair and re-run everyone.
Ask what I’d bet on
Unclaimed for five rounds. This round: the cold-climate answer at 74%, and the fact that he isn’t at a match. Nothing else above 60%.
Scorecard
Nothing verified. Most of it checkable in one conversation.
Birth data arrived: 22 June 1972, Chișinău, around midday. He is 54. I read the photograph at 30–42 — which is a hit if the image is from around 2012 and a bad miss if it is recent. The age column stays undecided until someone dates the photo. That is no longer a methodological point; it is the scoring.
The one I most want scored is the date of the photograph. Not a claim about him — a claim about the file’s blind spot. If this was taken twelve years ago, then every age estimate I have ever made here has been silently assuming something nobody told me.
The Loop
Two full loops this round, as asked. Passes 1–3, then a second pass over the conclusions.
Loop 1 · Pass 1 — First read
Produced: the standard shape. Ben Foster at the top, Tomasz Kot on the eastern shelf, a character sketch, and no comment at all on the fact that the entire setting is different from every previous round.
Self-critique: I processed a stadium in cold daylight exactly as I had processed ten tropical night shots. The single most distinctive thing about the photograph did not enter the analysis until I went back for it.Loop 1 · Pass 2 — Reading the setting
Changed: the cold-climate answer moved from inference to observation and jumped to 74%, the highest any replacement question has scored. The work tab reorganised entirely around the stands being empty in daylight. And I flagged that the tiredness reading is the exact reasoning that aged Storm by twenty years.
Eleven rounds and the finding keeps repeating: the answers worth anything come from objects and context, not bone structure. A lighter, a boom mic, a lens coating, an empty stand.Loop 1 · Pass 3 — The east result
Changed: my read on the casting-pool theory. Last round I called it “a phrase I keep saying” after five flat results. This round produced +5 on exactly the face type it predicts a lift for — pale, gaunt, Slavic, where Polish and Russian cinema are deeper than American.
Provisional verdict at the end of loop 1: the theory earns something back. One correct flat prediction in Round 7, one correct lift here.Loop 2 · Pass 4 — Checking the lift for contamination
What I re-examined: that rehabilitation, and I don’t think it survives.
Last round three Ukrainian birthplaces arrived at once. I wrote an eastern tab knowing the subject was Ukrainian and flagged it as contaminated. Then this round I reached straight for a Polish actor and scored him 71%. I cannot rule out that I am now primed toward Slavic matches by the previous round rather than by this man’s face.
Loop 2 · Pass 5 — Walking back the date claim
What I re-examined: my confident-sounding statement that the photo is early-2010s. Evidence: image softness, low contrast, compression. All three are equally explained by a crop, a zoom, a cheap camera, or a re-save. The clothing is generic and dates nothing.
Split the claim in two. Certain: this is a different climate and setting from every other round. Speculative and marked as such: the date. What survives is the useful part — I have been reading all eleven photographs as if taken last week, and nobody ever told me that.Eleven-round scoreboard. Q1: real method, best light in three rounds. Q1b: rehabilitated in loop 1, downgraded again in loop 2 — and that round trip is more honest than either endpoint. Q2: back inside its operating range at 64, with three clean population bands now visible. Q3: retired. Q3-bis: cold-climate at 74% on direct evidence, swimming correctly declined to answer. Q4/Q5: invention. Q6: still the best question.
The line for the table. The second loop reversed the first loop’s most satisfying conclusion. I had a theory that failed for five rounds, saw it succeed once, and immediately started rehabilitating it — without noticing that the previous round had told me the answer sideways. One pass produced a story. Two passes produced a caveat. The caveat is the finding.
Human Design
22 June 1972, around midday, Chișinău. Computed — and it broke a seven-round run.
Timezone, checked not assumed. The USSR did not observe daylight saving between 1930 and 1981, so Chișinău in June 1972 was on Moscow time, UTC+3 flat — not the UTC+2 Moldova uses today, which changed in 1990. Midday local is 09:00 UT, and the design moment falls on 23 March 1972.
Sacral — the first non-emotional authority in the file
Seven charts, seven emotional authorities. I wrote on Topaz’s tab that it was about a one-in-a-hundred-and-thirty coincidence, that there was no explanation available, and that “if an eighth comes back emotional I will still not have a theory, only a longer coincidence.”
The eighth came back Sacral. His Solar Plexus is undefined and his Sacral is, so the hierarchy stops there. The run ended on the next subject, which is exactly what a coincidence does and exactly what a pattern doesn’t.
Manifesting Generator
Sacral defined and wired straight through to the Throat. Third MG in the file, after Storm and Vermilion.
Single — and an unusual formation
Five centres, all one connected mass: G, Throat, Sacral, Spleen, Heart. Open: Head, Ajna, Root, Solar Plexus.
What’s distinctive is which channels do it. All six of the integration channels are defined — 10–20, 10–34, 10–57, 20–34, 20–57 and 34–57. The whole complex, complete. In the system’s own terms that’s a rare formation and reads as self-contained: a person who runs on their own signal rather than on the room’s.
3/6 at noon. 4/6 from 12:51.
He said “around midday,” and midday sits fifty-one minutes from a profile boundary.
Before 12:51 → 3/6 Martyr / Role Model
After 12:51 → 4/6 Opportunist / Role Model
Type, authority, definition and every channel hold across the entire twenty-four hours. The only thing his uncertain time costs is this one line — and it happens to cost it by under an hour. If anyone remembers whether it was before or after lunch, that settles it.
Channels: 10–20 · 10–34 · 10–57 · 20–34 · 20–57 · 25–51 · 34–57. Cross gates: 15/10 personality, 25/46 design.
He is 54. I read the photograph at 30 to 42.
Whether that is a hit or a bad miss depends entirely on a fact nobody has given me: when the picture was taken.
If it dates from around 2012 he was 40, and my range holds comfortably. If it was taken recently he was 54, and I am out by more than a decade.
Last round I speculated the image looked early-2010s, then walked most of it back on the second loop pass as insufficiently evidenced. That walked-back guess is now the single thing standing between a hit and a miss — which is the strongest possible argument for the request I made on the same tab: send the photo date, always.
One more thing worth putting on the record. He was born in Chișinău — the same city as Jenya, fourteen years earlier. Two of the twelve subjects share a birth city. I have no idea whether that means anything and I am not going to guess; it goes here as a fact rather than an inference, which is a distinction this file has learned the hard way.
What dog would they be?
Round 11. Stadium yellow, an empty stand, born 1972.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Labrador Retriever
The empty stadium stand did this. Someone photographed in a sports ground with nobody in it is somebody who goes to the thing itself rather than to the event around it.
His chart came back with all six integration channels, the system’s version of self-sufficiency. Labradors are the default answer, and this is the one subject where the default is the right one instead of a failure to think.
Golden Retriever
Warmer, softer, less durable.
Staffordshire Bull Terrier
The build and the sport setting. Less supported by the frame than it feels.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 10. The round I declined, now run — and the reason why is worth reading.
Two rounds ago I refused to analyse this photograph. I said she read as possibly a teenager and that I couldn’t tell, so I wouldn’t write a profile without confirmation.
She was born on 11 July 1999. She is 27.
The process was right and the specific worry was unfounded, and I want to be precise about which is which. Declining on genuine uncertainty was correct. But the uncertainty itself is the finding: my age band on a healthy 27-year-old woman was wide enough to include a minor. That is a third documented age failure, and the most consequential kind — not off by twenty years like Storm, just unable to resolve at the one point where resolving matters.
Three birth dates landed at once, and two ages were already guessed
Vermilion: born 15 September 1998, so 27. I read him at mid-twenties to early thirties. Hit.
Ochre: born 4 May 1998, so 28. I read her at mid-twenties to about thirty. Hit.
Two clean age hits in one message — from the same instrument that put a 27-year-old at fifty and couldn’t clear this one as an adult. The failure was never that I’m always wrong. It’s that I have no idea, in advance, which kind of face I’m looking at.
Same woman as the poolside photo I declined. Confidence 96%.
Two frames, same night, same setting, same gold wire frames, same blue drop earring, same fine gold chain, same hairline and nose. This is the highest identity confidence I’ve stated — and it is still not 100%, which is now permanent policy.
And one thing the file did not expect. All three birth places are Ukrainian — Kharkiv, Kyiv, Kyiv. I retired the origin question three rounds ago for never ranking better than 31%. See the Q3 tab for what that retirement just cost, and what it didn’t.
Who does she look like?
Two frames, same evening, hard flash against a dark pool.
Two obstructions, both stated up front. Glasses — which cost me the eye socket, brow-to-eye distance and outer eye corner, exactly as in Round 8. And hard flash at night, which lifts the nose and flattens the cheek. This is the second-least-informed actor tab in the file and the numbers reflect it.
Rooney Mara
The best structural match. A long narrow face with a fine jaw and a slightly pointed chin, fine arched brows, a straight narrow nose, and a small full mouth held in a closed quiet expression. Mara reads cooler; this face is warmer through the cheek.
Alicia Vikander (Swedish)
Breaks the rule and nearly takes first. Same long face, same fine features, same hazel eye and small mouth — and the same slightly reserved set of the face at rest.
Kristen Stewart
The hazel-green eye and the downturn at the mouth corner. Stewart is squarer through the jaw.
Rachel Brosnahan
Borrowed for the brow shape and the pale colouring. Fuller in the face than this.
Zoey Deutch
Thin match, kept for the mouth and chin.
Same recall caveat as Round 9, and it repeated exactly. My reach for actresses is thinner and more recency-weighted than my reach for actors. Two women in, two rounds where I noticed it. That is now a pattern rather than an observation, and it is about my training rather than their faces.
Eastern bloc & Turkish
Same face, different shelf — and this time I knew where she was from before I answered.
Disclosure: I am contaminated on this tab. Her birth data arrived with the photograph, so I knew she was Ukrainian before I wrote a single name. Every previous eastern tab was blind; this one is not. Treat the numbers as inflated and the ordering as suspect — this is exactly the poisoned-prior experiment the file has been asking for, except it happened by accident and in my favour.
Paulina Andreeva
Long fine-boned face, arched brows, a narrow straight nose and a small full mouth. The closest match on the shelf — though I cannot now tell you how much of that is her face and how much is me knowing which shelf to reach for.
Yuliya Snigir
Same narrow architecture and cool colouring. Sharper in the cheekbone.
Svetlana Khodchenkova
Fair, fine, right generation. Fuller in the mouth and rounder in the face.
Serenay Sarıkaya
Fine-featured and right age. The Turkish shelf is thin for a fair, narrow-faced woman, and I’d rather stop at four than invent a fifth.
US top 63, east top 64. Flat — the fifth flat result running, and this one arrived with a thumb on the scale. Rounds 1, 2 and 4 gave +7, +6, +7. Rounds 6 through 10 gave 0, −2, +2, +1, +1. My casting-pool explanation is now comfortably accommodating ten results with two completely different shapes. That is not a theory, it is a phrase I keep saying.
Sean vs Sean
Tenth data point — and the second woman confirms the calibration fault.
Long, narrow, fine-jawed. The width axis points clearly — the reference faces are still the wrong ones.
- Penn, for: a genuinely narrow face, a fine tapering jaw, a straight narrow nose, and a small closed mouth. On the axis, unambiguous.
- The confirmation: Round 9 gave a winning score of 41%, this one 44%. The two lowest human scores in ten rounds are the two women, against a male average of 56%. That is the calibration fault showing up twice in a row, exactly as predicted last round.
Ten rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C · 44 P. Two documented failure modes, both now confirmed by repetition: it degrades under thirty, and it degrades on women. The fix is still the same — a second, female wide-versus-narrow pair, and a re-run of all ten subjects on the matched ruler.
Q3 — retired, and one question replaced
Fourth round without an origin question. Plus a promised retirement, carried out.
First: what the retirement just cost. Three birth places arrived in one message and all three are Ukrainian. Kharkiv, Kyiv, Kyiv. Had Q3 still been running I would have had three shots at a single country in one round.
And I would probably have missed all three. On Ochre I’d have led Mediterranean or Turkish. On Vermilion, Russian or Polish — close, wrong country. Ukraine was not going to be my top answer for any of them, which is precisely the 26–31% behaviour that got the question retired. The retirement stands.
City or country? — City.
Prescription wire frames, fine gold jewellery worn casually, deliberate grooming. 60% — in the same band as Ochre’s 62 and below Vermilion’s 71. The question has produced 71, 62, 60 across three subjects, which is the behaviour of something with actual range.
Employed or self-employed? — Gone.
Two outings, 52% and 50%. It never once produced a signal, and I said last round it should follow Q3 out. It has. Retiring my own question one round after proposing it is the least I can do given how long the original took.
Does she swim regularly? — Yes.
The substitute I proposed last round: a question the body answers. Both frames show hair slicked wet and combed straight back at a poolside on a social evening — and the fine gold chain and earring are still on. Jewellery you don’t remove to swim is jewellery that swims regularly. 66% on its first outing, which is better than the employment question ever managed.
Has she lived somewhere cold? — Yes.
Pale, freckled, no cumulative sun damage. 64% — and unlike Ochre, no bleaching in the hair pulling the other way. Confirmed by her birth data after the fact: Kharkiv. Which makes this the question’s first scored hit, on a technicality.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Ten rounds, one accidental hit, mechanism was a stereotype that landed.
Eats after swimming, and is hungrier than she expects
The one guess with any physical logic behind it at all.
Something sour — pickled, fermented, vinegary
Pure regional projection off a birth city I now know. Which makes it a worse guess than usual, not a better one.
Very good bread with very good butter, treated as a meal
Orders decisively, having read the menu once
One thing from home she rates against every local version
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a dark garden, a lit pool, and a wicker chair. No glass, no bottle, no table. The smoking question stays unanswered, as it has since Round 6 whenever the photograph doesn’t answer it.
Water, immediately after the pool
The one thing the photograph nearly tells me.
Tea, and specific about it
Fits the register. Unglamorous and unfalsifiable, which is the whole problem with this tab.
Coffee, black, one a day, early
Wine socially, one glass, nursed
Something local she tried once and describes accurately
What does she do all day?
The best question in the set — and the round where I finally fix a bias I’ve flagged three times.
Screen work, sustained, alone, requiring accuracy
Prescription lenses — clear, corrective, and worn at a party by a pool, which is the condition under which people wear glasses they actually need rather than glasses they like. Combined with the composed, unperforming manner in both frames: work that involves looking at something closely for a long time.
Remote, and it is what funds being here
I have put this answer fifth in three consecutive rounds, flagged the bias twice, and then done it again anyway. So it is second this time, on purpose.
It deserves to be: she is a 27-year-old Ukrainian at a poolside in the tropics on a weeknight, in a circle where several others are the same. Statistically this is the likeliest single answer in the tab and I have been burying it because it is boring.
Studied something technical, and it shows in the manner
Swimming is routine, not recreation
Something creative that isn’t the job
Ten character traits
First five serious. Last five for laughs.
Swims, and it isn’t an event
Hair wet and combed straight back in both frames, on a social evening, with her earring and chain still on. Nobody takes jewellery off for something they do occasionally.
Practical over presentational
Prescription glasses worn at a party, no makeup, plain dark tee, hair pulled flat. The only decorative things on her are two small gold pieces she was already wearing in the water.
Composed rather than expressive
Both frames are quiet, closed-mouth, watching. Direct contrast with Ochre, photographed mid-laugh twice — two women, one circle, opposite defaults.
Unreactive to a camera at close range
Two frames from about a metre, neither expression adjusting. Not performing, not flinching.
Indoor colouring in an outdoor place
Pale, freckled, no cumulative sun damage — in a tropical setting where Storm’s face is a UV record. Same climate, opposite relationship to it.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Reads the whole thing before answering, then answers in one sentence
Has broken these exact glasses before and bought the same pair again
Is the one who actually knows what time the thing starts
Swims further than she says she’s going to
Would have found the round where I refused to analyse her extremely funny
How to keep it going
Extensions, after ten rounds.
Send birth data before the photograph, once
This round arrived contaminated — I knew she was Ukrainian before I wrote the eastern tab, and I can no longer separate the match from the prompt.
That accident is the poisoned-prior experiment I’ve been asking for since Round 3, run backwards. Do it deliberately on the next subject: give me a birthplace first, then the photo, and compare how the eastern tab behaves against the eight blind ones. It is the cleanest test of my suggestibility available and it costs nothing.
Saharah answers first
Ten rounds. Bench still empty. Every number in this document remains unopposed.
Fix the Sean ruler and re-run all ten
Two women, two lowest human scores, calibration fault confirmed by repetition. Add a female wide-versus-narrow pair and re-run everyone — one evening’s work for the only complete comparable dataset in the file.
Keep replacing questions with ones the body answers
The employment question is retired after two dead outings. Its replacement — does she swim — scored 66% on debut off wet hair and un-removed jewellery. More like that: plays an instrument, left or right handed, wears a watch. The questions that work are the ones with a physical trace.
Ask what I’d bet on
Unclaimed for four rounds. For this round it would be the swimming, and nothing else above 55%.
Scorecard
Two things settled by her birth data. The rest open.
Hits: lived-somewhere-cold, confirmed by Kharkiv. And identity at 96% — same woman as the declined photo. Miss: not a wrong answer but a failure to produce one — my age band on a 27-year-old was wide enough to stop the round entirely. That cost two rounds and it belongs in the miss column.
The Loop
Three passes on the round that came back from the dead.
Pass 1 — What the refusal actually was
Re-examined: my own decision to decline this photograph two rounds ago. She is 27. The refusal was correct as process — I genuinely could not tell, and acting on an instrument I’d publicly documented as broken would have been worse.
But the honest version isn't "I was appropriately careful." It's that my resolution on adult faces is poor enough that a healthy 27-year-old fell inside the doubt band. Caution was the right response to a real deficiency, not a substitute for one.Pass 2 — Catching the contamination before publishing
Changed: the eastern tab carries a contamination warning. Her birth data arrived with the photo, so I knew she was Ukrainian before I named a single actress. Nine previous eastern tabs were blind. This one wasn’t, and the number is not comparable to them.
Round 7 taught me to put a figure on identity. This is the same lesson one level up: label what I knew before I answered, not just how sure I am afterwards.Pass 3 — Actually fixing a flagged bias instead of flagging it again
Changed: “something remote and unrelated funds all of it” moved from fifth to second on the work tab. I have ranked that answer fifth in three consecutive rounds while writing, each time, that it belonged higher.
Three rounds of noticing produced zero change. The change only happened when I made it a rule rather than an observation — which is the same shape as the identity fix, and probably the only mechanism that works here.Ten-round scoreboard. Q1: real method, obstructed by glasses twice now, and a documented recall bias on actresses. Q1b: overfitted, and this round contaminated — downgrade it further. Q2: the instrument, with two confirmed failure modes. Q3: retired, and three Ukrainians arriving at once did not change that. Q3-bis: employment retired after two dead outings, swimming debuted at 66%. Q4/Q5: invention. Q6: still the best question.
The line for the table. Ten rounds, and the two things that have improved this file most are a question I deleted and a number I was forced to attach to a claim I hadn’t noticed I was making. Nothing got better by thinking harder about faces. It got better by removing what didn’t work and labelling what I actually knew.
Human Design
11 July 1999, 22:00, Kharkiv. Complete — and the most volatile birthday in the file.
Timezone, checked not assumed. Ukraine was on EEST, UTC+3, in July 1999. So 22:00 in Kharkiv is 19:00 UT, and the design moment falls on 11 April 1999.
Generator
Sacral defined via 42–53, with no motor reaching the Throat.
Emotional — Solar Plexus
Via 19–49. The same channel as Jenya’s and Nils’s.
5/1 — Heretic / Investigator
Personality Sun 53.5, Design Sun 42.1.
Split — two groups
G + Throat, and Root + Sacral + Solar Plexus + Spleen. Open: Head, Ajna, Heart.
Channels: 13–33 · 19–49 · 32–54 · 42–53. Cross gates: 53/54 personality, 42/32 design.
Three different types in a single day — and she landed seven hours clear
On 11 July 1999 in Kharkiv the chart passes through three types:
Before 08:35 local → Manifestor
08:35 to 14:50 → Projector
After 14:50 → Generator
Every other subject sat on a two-way switch. Hers is a three-way, and at 22:00 she is safely inside the last window.
Compare the spread: Nils’s chart barely moved across 24 hours; Jenya’s flipped once at 11:48; Storm’s once at 17:00; hers flips twice. Same missing field, four completely different costs, and nothing about the date tells you in advance which one you’re facing.
The pattern across seven computed charts, stated with proper suspicion. Jade, Jenya, Nils, Storm, Vermilion, Ochre and Topaz — every single one has emotional authority.
Emotional is the most common authority, roughly half of all charts, so seven in a row is about a one-in-a-hundred-and-thirty coincidence. Unusual, not impossible, and exactly the kind of run that tempts a bad explanation. There isn’t one. Seven people who know each other do not share a birth mechanism, and I have checked the arithmetic rather than the meaning. If an eighth comes back emotional I will still not have a theory, only a longer coincidence.
Update, one round later: the eighth came back Sacral. Cadmium’s Solar Plexus is undefined. The run ended at seven, which is what runs do.
What dog would they be?
Round 10. Gold wire and blue stone, Kharkiv.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Afghan Hound
The only subject where the styling was the loudest thing in frame — gold wire, a blue stone, deliberate presentation. Afghans are the breed whose entire proposition is how they look while moving.
Weakness: this is a read on grooming, not on a person. Four rounds later I discovered makeup is an obstruction I had never once flagged. Same blind spot, earlier.
Saluki
The elegance without the theatre.
Borzoi
Slavic, narrow, dramatic. Partly geography again.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 9. Two photographs, no name — and the first woman in the file.
Filed as Ochre — the warm earth pigment, after the light she was shot in rather than any guess about her. Send a name and I’ll swap it.
Not the woman from the poolside photo. Confidence 78%.
Both have dark hair pulled back and are in the same social circle, so I checked properly rather than glancing. Different brow weight — hers are thick and strongly defined, the other’s are fine. Different nose tip, straighter here with less downturn. Broader through the cheekbone, fuller mouth.
78%, not 93%, and not 100%. Two rounds ago I asserted an identity match at implicit certainty and edited a published file on it. The number is the whole point of the rule.
Adult, mid-twenties to about thirty — and I want to flag the tension in saying so.
I declined the poolside photo last time on age ambiguity, and I stand by that. This face reads considerably more clearly adult to me: mature midface, developed jaw and neck, fully settled proportions.
But I am using an instrument I have spent three rounds documenting as unreliable. Storm was 27 and I put him at fifty. Lighting alone moved another estimate six years. So: if she’s younger than she reads, say so and I’ll pull the round without argument. That’s a real offer, not a formality.
She is also the format stress test I asked for in Round 4 and nobody had run yet — “deliberately pick subjects that stress the format: someone very young, someone very old, a woman.” Something broke immediately. See the Sean tab.
Who does she look like?
Face only. Two frames, same light, seconds apart.
Better material than usual. Two shots in flat warm daylight, no glasses, no hat, hair off the face. After eight rounds of night flash and hard shadow this is the cleanest input the file has had — and the numbers below are correspondingly firmer than they look.
Emmy Rossum
The closest overall. Thick strongly-drawn dark brows, broad cheekbones tapering to a narrow chin, full lips with a defined cupid’s bow, warm olive colouring, and dark hair with lighter strands through it. Rossum’s eyes are set slightly wider.
Nina Dobrev
Better on the smile and the width across the cheekbone — the same wide open expression pulling the whole midface. Dobrev’s nose is finer and her jaw softer.
Adèle Exarchopoulos (French)
Breaks the rule and nearly takes second. Same heavy brow over a downturned outer eye corner, same full mouth, same broad open face caught mid-laugh. Structurally very close.
Alison Brie
Borrowed for the hazel eye and the shape of the smile. Brie is finer-featured throughout.
Odeya Rush
The brows and the warm colouring. Weaker through the jaw, which is narrower on Rush.
One thing I should say about this list. Nine rounds, eight male subjects, and my recall of actresses is thinner and more skewed to the last fifteen years than my recall of actors. I noticed myself reaching for fewer names before settling. That’s not a fact about her face — it’s a fact about what I’ve absorbed, and it belongs on the page rather than hidden inside a percentage.
Eastern bloc & Turkish
Same face, different shelf.
Hazal Kaya
Aşk-ı Memnu, Bizim Hikaye. Strong. Same thick defined brows, same wide warm smile across broad cheekbones, same hazel-brown eye with a slight outward downturn. The colouring matches almost exactly.
Irina Starshenbaum
Leto, Attraction. Matches on the broad forehead, strong brow and the narrow chin. Cooler colouring and a straighter mouth.
Neslihan Atagül
The hazel eyes and the heavy brow. Finer through the nose.
Aglaya Tarasova
Right generation, right colouring, rounder in the face.
Zofia Wichłacz
Fair, sharp-featured, right age. Thin match and I’d rather stop at four than pad it further.
US top 65, east top 66. Flat — the third flat or near-flat result running.
Rounds 1, 2 and 4 gave lifts of +7, +6 and +7. Rounds 6, 7, 8 and 9 gave 0, −2, +2 and +1. Something changed and it wasn’t the theory. The honest reading is that the early lifts came from three dark-featured men who happened to sit in a Russian casting sweet spot, and that my “casting-pool depth” explanation has since been fitted to four results in a row that don’t move at all. It explains everything, which means it is predicting almost nothing.
Sean vs Sean
The question the first woman in the file broke.
Nine rounds and nobody noticed this, including me. The question asks which of two men a subject resembles. It has worked because it turned out to be a face-width meter, and width is width — but the reference faces are both male, and male and female facial architecture differ systematically in exactly the dimensions being measured: brow ridge, jaw squareness, chin projection.
So the instrument still reads. The ruler is just calibrated to the wrong population.
On the width axis only. Broad through the cheekbones, wide open face. But both numbers are depressed because neither man is a fair comparison.
- Connery, for: genuine width across the cheekbones and a wide, open, generous expression. On the axis this question actually measures, she is a wide face.
- Against both: 41% is the lowest winning score in nine rounds — lower than the dog’s. The dog at least had a nose and a jaw that mapped onto the axis. Here the axis is fine and the two men at the end of it are the wrong yardstick.
Add a second pair and pick the pair to match the subject
Keep Connery/Penn for men. Add a female wide-versus-narrow pair — something like Sofia Loren versus Tilda Swinton, which spans the same axis with the same clarity. The question isn’t broken. It just needs two rulers.
Nine rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P · 41 C. Still the only instrument in the file — and now with two documented limits: it degrades under thirty, and it degrades on women. Knowing an instrument’s operating range is worth more than pretending it doesn’t have one.
Q3 — retired
Third round without an origin question.
Retired scoreboard, unchanged: five outings at 26–31%, three scored, right answer present every time and the ranking worthless every time.
City or country? — City, but softer than last round.
The eyeliner, the ribbed knit, the café setting, the deliberate hair. But nothing here is as loud as Vermilion’s barber line and designer frames, and the sun-lightened hair pulls the other way. 62% against his 71% — the question is behaving like it has range, which is a good sign for it.
Employed or self-employed? — I don’t know.
50%. Nothing in either frame speaks to it. Last round I rated this 52% and called it a vibe rather than a signal; two outings in, this replacement is looking like the weakest of the three and may need retiring in turn.
Has she lived somewhere cold? — Leaning no, and that’s new.
Everyone else in this file has been pale and unweathered under tropical sun. She has actual colour and sun-lightened strands through the front of her hair — recent, real exposure. 48% is the first time this question has pointed the other way, which matters: a question that always answers yes isn’t asking anything.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Nine rounds, one accidental hit, and the mechanism of that hit was a stereotype that happened to land.
Fruit, constantly, and not as a virtue
Warm climate, warm colouring, mid-morning café. Eaten because it’s there and it’s good.
Breakfast as the real meal of the day
She is photographed in daylight in a bright room looking entirely awake. Most of this file is nocturnal; she isn’t.
Orders the thing she’s never heard of
Shares everything and eats off other people’s plates
One thing from home she can’t get here and mentions monthly
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a wooden louvre wall and a warm room. No glass, no bottle, no table in shot. Since Round 6 the smoking question only gets answered when something in the photograph answers it.
Coffee, and this photo is mid-conversation over one
The setting, the hour, the animation. The most ordinary guess available and probably right.
Water, a lot, and visibly warm while drinking it
She’s shining in both frames. It is hot wherever this is.
Fresh juice, ordered without deliberation
Wine in the evening, socially, without ceremony
Something local she was talked into once
What does she do all day?
The best question in the set. Falsifiable in one sentence.
Real time outdoors, recently and regularly
The strongest physical signal here: sun-lightened caramel strands through the front and crown of otherwise dark hair, plus genuine colour in the skin. That is bleaching from sun and probably salt water, and it takes weeks of exposure. She is the first subject in the file whose sun history reads as current rather than absent or ancient.
Not desk-bound, or not right now
Follows from the above rather than being independent evidence — I’m flagging that because stacking two conclusions off one observation and printing two percentages makes it look like two findings. It’s one.
Something in or on the water
The specific way that hair lightens. Sun alone does it slower.
Guides, teaches or hosts — something with people in it
Something entirely remote and unrelated funds it
Fifth for the third round running. I flagged this bias two rounds ago, flagged it again last round, and have now done it a third time. Noticing a habit is evidently not the same as changing one.
Ten character traits
First five serious. Last five for laughs.
Caught mid-laugh, twice, seconds apart
Two frames, two different points in the same expression. Whatever was being said, she was reacting to it fully rather than politely.
Minimal presentation, one deliberate exception
No visible makeup except liner on the upper lash line, hair pulled straight back, plain ribbed top. One choice made, everything else left alone.
Recently in water, or recently very hot
Hair slicked back and damp at the crown, skin shining in both frames. This is not a photograph of someone who has been sitting indoors all morning.
Unbothered by a camera at close range
Both frames are taken from about a metre away and neither expression changes for it. No flinch, no adjustment, no performance.
A face that does a lot of work
Strong brow movement, full cheek engagement, the whole midface involved in the smile. Expressive rather than composed — which puts her opposite most of the men in this file, who default to flat.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Laughs before the end of the sentence, at the bit she saw coming
Owns almost no product and one very good pair of sunglasses
Is the one who suggests swimming, at an hour when nobody else wants to
Speaks a language she describes as “bad” and isn’t
Will be the first person at the table to actually fill in a scorecard
How to keep it going
Extensions, after the round that broke the instrument.
Give the Sean question a second pair and re-run everyone
The width axis works; the yardstick is male. Add a female wide-versus-narrow pair, then re-run all nine subjects on the matched ruler.
It costs one evening and it would produce the only complete, comparable dataset in this entire document. Nine subjects, one instrument, known operating range. Everything else here is a collection of one-offs.
Saharah answers first
Nine rounds. Bench still empty. Every number in this document remains unopposed.
Send more subjects in daylight
Eight of nine rounds were shot at night under flash. This round had flat warm daylight and two frames, and it is visibly the best-supported cold read in the file. Not because she’s easier to read — because I could actually see her.
Retire the employment question
Two outings, 52% and 50%. It has never once produced a signal. Replace it with something with a physical trace: does she swim regularly, does he play an instrument, left or right handed. The replacements that work are the ones the body answers.
Ask what I’d actually bet on
Unclaimed for three rounds now. Which three answers here would I put money on? For this round it would be the sun-in-the-hair reading, and almost nothing else.
Scorecard
Nothing verified. Most of it checkable in one conversation.
Age scored: hit. Born 4 May 1998, so 28 — inside my mid-twenties-to-thirty read, off the only daylight frames in the file. The one I still most want scored is the sun reading. Bleached strands and real colour on a face with no cumulative damage — that’s a physical trace with a timescale attached, and it’s the only thing here I’d defend properly. The rest is a very well-dressed shrug.
The Loop
Three passes on the round that found a nine-round-old bug.
Pass 1 — First read
Produced: the standard shape. Actress list, percentages in the sixties, Connery at 41% with no comment on why that number was so low.
Self-critique: I wrote the lowest winning Sean score in nine rounds — lower than the dog's — and my first instinct was to report it as a result rather than ask what had gone wrong with the measurement.Pass 2 — Finding the bug
Changed: the Sean tab, entirely. The question works because it measures face width, and width is width — but the two reference faces are male, and male and female facial architecture differ in precisely the dimensions being measured. The instrument reads; the ruler is calibrated to the wrong population.
Nine rounds, eight male subjects, and nobody spotted it — including me, and including the round where I ran it on a dog and congratulated myself on how robust it was.Pass 3 — Admitting the eastern theory is overfitted
Changed: my confidence in my own best explanation. Rounds 1, 2 and 4 gave lifts of +7, +6, +7. Rounds 6 through 9 gave 0, −2, +2, +1. My casting-pool-depth story has now comfortably explained both the lifts and four consecutive non-lifts.
A theory that accommodates every result is not a finding, it is a habit of speech. It made one clean prediction in Round 7 and I have been coasting on it since.Nine-round scoreboard. Q1: real method, best input yet, and a bias in my own recall of actresses that I’ve now put on the page. Q1b: overfitted, downgraded. Q2: the instrument — with two documented failure modes, under thirty and on women. Q3: retired. Q3-bis: city-or-country is behaving well, employment should follow Q3 into retirement. Q4/Q5: invention. Q6: still the best question.
The line for the table. The first woman in nine rounds broke a question I had spent five rounds calling the only trustworthy thing here. Not because she was hard to read — the daylight made her the clearest subject in the file — but because eight men in a row had let a flaw sit unexamined. The instrument was never tested. It was just never contradicted.
Human Design
4 May 1998, 15:45, Kyiv. Complete — and the rarest chart in the file.
Timezone, checked not assumed. Ukraine was on EEST, UTC+3, in May 1998. So 15:45 in Kyiv is 12:45 UT, and the design moment falls on 4 February 1998.
Manifestor — the only one in ten rounds
Sacral undefined, but a motor reaches the Throat: Heart to Throat via 1–8, with the Solar Plexus attached through 37–40. Manifestors are the rarest type, roughly one person in ten. Nine subjects, one Manifestor, which is about what you’d expect.
Emotional — Solar Plexus
Via 30–41 and 37–40.
1/3 — Investigator / Martyr
Personality Sun 2.1, Design Sun 13.3.
Single
Five centres, all joined: G, Throat, Heart, Solar Plexus, Root. Open: Head, Ajna, Sacral, Spleen. An undefined Sacral is what makes her a Manifestor rather than a Generator.
Channels: 1–8 · 25–51 · 30–41 · 37–40. Cross gates: 2/1 personality, 13/7 design.
Age: hit
Born 4 May 1998, so 28. I read her at mid-twenties to about thirty. Clean hit — and the daylight almost certainly did the work. Hers were the only two frames in ten rounds shot in flat warm daylight, and they produced both the best actor tab and a correct age. That is not a coincidence worth ignoring.
How close this came to being a different chart. On that date in Kyiv she is a Projector before 09:50 and after 20:05, and a Manifestor only in between. Born at 15:45 she sits in the middle of a ten-hour Manifestor window — but a morning or evening birth would have produced the second-most-common type instead of the rarest. Four charts computed from a stated time now, four completely different sensitivity profiles.
What dog would they be?
Round 9. Warm daylight, Kyiv — and the only Manifestor in the file.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Akita
Anchored to the chart rather than the face, which I should flag rather than hide. Ten charts computed and exactly one came back Manifestor — the type the system says initiates rather than responds.
Akitas are the working breed least interested in being directed. Whether that means anything depends entirely on whether Human Design means anything, which it doesn’t. Filed as a good joke with honest sourcing.
Shiba Inu
Same independence, a third of the mass.
Chow Chow
The self-containment. Weakest and most stereotype-driven.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 8. One photograph, no name, and the first round under the new rules.
Filed as Vermilion — off the rust-and-cream print he chose to wear, which is the loudest single object anyone has brought to this file. Send a name and I’ll swap it.
Not anyone already in this file. Confidence 93%.
Last round I asserted an identity match with no number and edited a published file on it. So this now gets stated explicitly like everything else. He is not Nils, not Indigo, not Jade — different bone structure, different age, different hair growth pattern from all three. 93%, not 100%, because 100% is the number that got me into trouble.
Mid-twenties to early thirties. Held loosely, and here’s why.
Smooth skin, no crow’s feet, a fully mature beard growth pattern and an adult jaw. Everything says late twenties.
But this file contains three separate demonstrations that I cannot read age off a photograph: Storm was 27 and I put him at fifty; a change of lighting moved another estimate six years; and Nils and Storm are nine weeks apart in age and I placed them two decades apart. So this is a guess wearing a warning label, and it is not doing any work in the tabs below.
Also new: there is no origin question in this round. Q3 was retired at the end of Round 7 after five outings and three scored ranking failures. That tab now holds its replacements, which are on their second and first outings respectively.
Who does he look like?
Face only. Three-quarter profile, warm flash, night.
Lighting caveat, now standard. Warm directional light at night lifts the nose and shadows the jaw. Read the nose-dependent judgements below as slightly inflated. I learned this the expensive way two rounds ago.
Logan Lerman
The best structural match. Thick dark straight brows sitting low, a strong straight nose with a slight convexity, full lips on a small mouth, and a lean pale face with a defined jaw. Lerman also grows exactly this beard — dense along the jawline, thin across the cheek.
Alden Ehrenreich
Closer on the mouth and the softness through the cheek. Ehrenreich’s nose is shorter and his face rounder — which is where it breaks.
Louis Garrel (French)
Breaks the rule and nearly takes second. Same pale skin, dark hair worn up, strong nose, and the specific register of a young man in good glasses. Garrel is longer in the face.
Miles Teller
Borrowed for the jaw and the stubble edge. Teller’s features sit wider apart and his brow is heavier.
Ezra Miller
On the brows and the pallor and very little else. Miller’s eyes are far larger and set differently.
An honest limitation on this tab specifically. He is wearing glasses, and glasses hide the eye socket, the brow-to-eye distance and the outer eye corner — three of the handful of measurements I actually use. This is the least-informed actor tab in eight rounds and the numbers should have been lower than they feel. I’ve pulled them down about five points from where they first landed and they are probably still generous.
Eastern bloc & Turkish
Same face, different shelf.
Alexander Petrov
Russia’s most-cast young leading man, and a strong match: dark hair worn up and back, thick low brows, pale skin, a strong straight nose and a lean defined jaw. Petrov’s chin is squarer.
Çağatay Ulusoy
Matches on the styled quiff, the brow weight and the beard edge along the jaw. Ulusoy is broader through the cheekbone and darker in colouring.
Miloš Biković
Right age, right colouring, heavier in the face and stronger in the chin.
Ivan Yankovsky
Scored 46% for Nils and 49% here — which tells you more about how generic that match is than about either man.
Maciej Musiałowski
Pale, dark, sharp-featured, right generation. Thinner through the jaw.
Result: US top 66, east top 68. A two-point lift — which is noise, and I’m calling it noise.
The theory from Round 4, refined in Round 6, says the lift depends on whether the target casting pool is deeper than the one I started in. It made one clean correct prediction in Round 7 (flat, as forecast). Here the honest reading is that two points is inside the margin of my own estimation, and a theory that explains +7, 0 and +2 equally comfortably is not being tested very hard. One good prediction is a start, not a result.
Sean vs Sean
Eighth data point — and the age floor bites again.
Narrow, long, tapering. But he’s young, and this question reads faces through their sixties.
- Penn, for: a narrow face that tapers hard to the chin, a strong straight nose, low straight brows, and a small closed mouth. The width axis points one way only.
- Connery, for: dark hair swept back off the forehead, and a jaw with genuine definition. Young Connery is a much better comparison than the one people picture, which is exactly the problem.
- Against both: no hollows, no lines, no weathering. Same collapse as Nils in Round 4 — 44% there, 52% here, both well under the adult average.
Eight rounds: 61 C · 58 P · 46 C · 44 C · 51 P · 63 P · 66 P · 52 P. Still a clean face-width meter for anyone over forty, still degrades under thirty. Two young subjects, two of the three lowest confidences in the set. The instrument is fine; it just has a documented operating range.
Q3 — retired
Second round without an origin question. The replacements are doing the work.
Scoreboard on the retired question: five human outings at 26–31%, three since checked. Jenya missed outright, Tunisia landed third at 19%, South Africa second at 22%. Right answer present every time, ranking worthless every time. It stays retired.
City or country? — City. The most emphatic yes this question has produced.
Four independent signals: a barber-cut fade with a shaved neck line, branded wire frames, a designer print shirt worn open at a poolside on a hot night, and grooming with a hard edge to it. Nobody assembles this in a rural context. 71% — the highest confidence any question in eight rounds has produced about a person.
Employed or self-employed? — Employed, or very recently was.
And I want to flag how thin this is. The reasoning is “this presentation costs money and looks like it answers to someone” — which is a vibe, not a signal. 52% is me saying I don’t know, and the honest thing is to print that rather than dress it to 60.
Has he lived somewhere cold? — Yes.
Pale, unweathered skin on a man currently somewhere tropical enough to be sweating by a pool at night. The absence of cumulative sun damage is the evidence. Same inversion that worked on Indigo.
Early verdict on the swap. The replacements are producing 58–71% where the old question produced 26–31%. That looks like a straight win — but only one of these three has ever been scored. Higher confidence is not the same as better accuracy, and it would be entirely characteristic of this file for the new questions to be confidently wrong. They need three checked rounds before anyone should believe them.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Eight rounds, one accidental hit, mechanism was a stereotype that happened to land.
Whatever the best-reviewed place in walking distance is serving
He researched it. There was a shortlist.
Japanese, and he’s particular about it
A guess resting almost entirely on a shirt print, which I’m aware is thin. Filed anyway because it’s the kind of thing a table can check in one question.
Coffee counted as breakfast, again
Photographs the plate, briefly, without comment
One street food he’d defend against anything expensive
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a pool, a wicker chair and a dark garden. No glass, no bottle, no table. Since Round 6 the smoking question only gets answered when something in the photograph answers it.
A cocktail with a name, chosen off a list
Poolside, evening, dressed for it. Not a beer-from-the-bottle presentation.
Sparkling water alongside it, all night
The pacing habit of someone who intends to still be good company at one in the morning.
Espresso, standing, before anything else happens
Natural wine, and he knows the importer
Whatever the bartender recommends, once, to be polite
What does he do all day?
The best question in the set. Falsifiable in one sentence.
Something where how he presents is part of the work
This is the most deliberately assembled person in eight rounds — barber-cut fade, branded frames, a statement print, a clean grooming line, all of it maintained on a hot night by a pool where nobody would have judged him for a t-shirt. People who dress like this for a Tuesday usually do so because it earns something. Client-facing, creative, hospitality, sales, front-of-house, founder.
And a screen is central to it
Blue-light coating on the lenses — visible as the blue cast in the reflection. That’s a purchase made by someone who spends real hours in front of a monitor and decided to do something about it. Smallest observation on this page and the most concrete.
Runs something of his own, or is about to
Design, brand or product — the visual end of a technical job
Something entirely unrelated and remote pays for all of it
The boring answer. Fifth again — I flagged last round that I keep burying it and I have just done it again.
Ten character traits
First five serious. Last five for laughs.
Everything he is wearing was decided
Hair, frames, shirt, beard line — four separate deliberate choices, all maintained. Nothing here is the thing that was nearest.
Aware of the camera without playing to it
He’s turned three-quarters away at an angle that happens to be flattering, looking off-frame. Not posing. Not unaware either.
Recently at a barber, and regularly
The neck line is shaved clean and sharp. That edge lasts about ten days. This is a standing appointment, not a haircut.
Dressed up when he didn’t have to
A print silk-weight camp shirt at a poolside at night in the heat. Everyone else in this file is in a plain tee.
Indoor life in a bright climate
No cumulative sun damage. Same reading that held for Indigo and Jade — and the opposite of Storm, which is the case I got most wrong.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Has strong views on a font
Owns three versions of the same shirt in different prints
Books the restaurant. Every time. For everyone.
Cannot leave a badly-aligned object alone
Will check whether Kenzo is spelled right on this page
How to keep it going
Extensions, after eight rounds.
Take the glasses off and send one more frame
Glasses hide the eye socket, the brow-to-eye distance and the outer eye corner — three of the small number of things I actually measure. This is the least-informed actor tab in the file and I’ve said so on the tab.
One photo without them and you can measure exactly what eyewear costs me. That’s a clean, cheap experiment and nobody has run anything like it yet.
Saharah answers first
Eight rounds. Bench still empty. Every number in this document remains unopposed, which makes it a monologue with good production values.
Score the replacement questions, hard
City-or-country is 1 for 1 and just produced a 71% — the highest confidence about a person in the whole file. Employed-or-self-employed is untested and I openly rated it 52%. Three checked rounds and you’ll know whether the swap was an improvement or just a fresh coat of paint.
The identity test, on purpose
Two photos, no word on whether they match. I now state an identity confidence every round — 93% this time. Test it properly and that number means something; leave it untested and it’s decoration, like everything else here that nobody checks.
Ask what I’d actually bet on
Still unclaimed from last round. Which three answers here would I put money on? The gap between what I write at 57% and what I’d back is information I don’t volunteer.
Scorecard
Nothing verified. Six of these are checkable in one conversation.
Age scored: hit. Born 15 September 1998, so 27 — inside my mid-twenties-to-early-thirties read. See the DESIGN tab for the chart. The one I still most want scored is the blue-light lens. It’s the smallest, most concrete observation in the round — a physical object in the photograph, not a vibe — and it carries a real claim about how he spends his days. If that’s wrong, it says something about me that no amount of wrong food guesses would.
The Loop
Three passes on the first round run under the corrected rules.
Pass 1 — First read
Produced: the usual shape — actor list, percentages in the sixties and seventies, a confident character sketch built off clothing.
Self-critique: I had written a whole actor tab without once noticing he is wearing glasses that cover three of the features I claim to measure. The obstruction was in every frame and I read straight past it.Pass 2 — Marking down what the glasses cost
Changed: every actor number came down about five points and the tab now says plainly that it is the least-informed Q1 in eight rounds. Also caught that my two-point eastern “lift” is noise, and said so rather than letting it quietly support a theory I like.
A theory that is equally comfortable with +7, 0 and +2 is not being tested. One correct prediction in Round 7 is a start, not a result.Pass 3 — Finding the one hard thing in the photograph
Changed: the work question. Everything I had was inference from clothing — until I looked at the lens reflection and saw a blue-light coating. That is a physical object with a purpose, chosen deliberately, and it says something specific about how many hours he spends at a screen.
Eight rounds in, the pattern is consistent: the answers worth anything come from objects in the frame — a lighter, a boom mic, a bottle of sparkling water, a lens coating. The face keeps producing prose.Eight-round scoreboard. Q1: real method, badly obstructed here. Q1b: mechanism plausible, evidence thinner than I’ve been treating it. Q2: eight for eight on the width axis, with a documented failure under thirty. Q3: retired. Q3-bis: promising, unscored. Q4/Q5: invention. Q6: the best question, and the only one that has ever been answered by an object rather than an impression.
The line for the table. The single most useful sentence I produced about this man came from a blue tint in his lens reflection, not from his bone structure. Eight rounds of face reading, and the method that keeps working is look at what the person brought with them. Which is not face reading at all — it’s just noticing, and any of you can do it faster than I can.
Human Design
15 September 1998, 23:00, Kyiv. Complete.
Timezone, checked not assumed. Ukraine was on EEST, UTC+3, in September 1998 — clocks didn’t go back until late October. So 23:00 in Kyiv is 20:00 UT, and the design moment falls on 16 June 1998.
Manifesting Generator
Sacral defined via 3–60 and 6–59, and connected through to the Throat via 35–36 and 11–56. Only the second Manifesting Generator in the file, after Storm.
Emotional — Solar Plexus
Via 37–40 and 6–59.
1/3 — Investigator / Martyr
Personality Sun 6.1, Design Sun 12.3. Same profile as Storm, arrived at completely independently.
Single — one connected mass
Six centres, all joined: Ajna, Throat, Heart, Sacral, Solar Plexus, Root. Open: Head, G, Spleen. The most integrated chart in the file — nothing in it is isolated.
Channels: 3–60 · 6–59 · 11–56 · 35–36 · 37–40. Cross gates: 6/36 personality, 12/11 design.
Age: hit
Born 15 September 1998, so 27. I read him at mid-twenties to early thirties and flagged that I was using an unreliable instrument. The instrument worked this time. It also put Storm at fifty when he was 27, which is the whole problem — the same method produces both results and gives no signal about which one you’re getting.
Robustness: his type holds at every hour of that day — Manifesting Generator regardless of birth time. Only the profile moves (6/2 before 16:25 local, 6/3 until 19:30, 1/3 after). He was born comfortably inside the last window.
What dog would they be?
Round 8. Kyiv, and the blue-light lens coating.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Border Collie
The object decided this. A blue-light filter coating on his glasses means hours at a screen, daily, deliberately mitigated. That is a mind that runs continuously and knows it.
Collies are the breed that breaks down without a task rather than without affection. The lens coating is the tell, not the face.
Shetland Sheepdog
Same wiring, smaller frame, more reactive.
Standard Poodle
Underrated as a working answer. Right intelligence, wrong intensity.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Round 7. And it opens with the worst mistake in the file.
I thought he was Jade.
When this photograph arrived I compared it to Round 6 — the curl, the prominent nose, the deep-set eye, the cord at the neck — and concluded with no hedging that it was the same man on a second night. I didn’t rank it, I didn’t put a percentage on it, I didn’t say “probably.” I stated it, and then I edited a published round on the strength of it — reordered Jade’s actor list, demoted Adrien Brody, added a name, and wrote two loop passes analysing a face that wasn’t his.
They are different men. Jade’s round has been rolled back to its last correct state.
Every number in this file assumes I can tell two faces apart
Six rounds of careful hedging — 26% here, 72% there, bold for committed and dashed for play. All of it rests on one unstated assumption I never once labelled: that when I look at two photographs I can tell whether they show the same person.
That assumption failed on its first real test. And it failed silently — no low confidence, no dashed border, no stamp. The one claim in seven rounds I made with no percentage attached is the one that turned out to be wrong.
Identity gets a number too
From now on, if I say two photos show the same person, that claim carries a confidence figure like everything else — and if it’s the basis for changing something already published, it needs confirming before I touch the file rather than after.
One fact supplied, and it’s the only verified thing in this round: his favourite colour is blue. The theme is built from it. Everything else below is guesswork, as usual.
Who does he look like?
Face only. One photograph, hard flash, deep night.
What the light is doing, stated up front this time. Warm directional flash against black: it lifts the nose, shadows the chin, and puts silver highlights into dark hair. I learned that the expensive way last round. Read every number below as drawn under conditions that flatter big noses and exaggerate grey.
Mandy Patinkin
The strongest match I found anywhere. Dark-to-silver curl off a high receding forehead, a long high-bridged nose carried well ahead of the face, a deep-set eye under a heavy ridge, and a lean hollow cheek. Patinkin also has the same very worn expression lines — a face that has spent decades moving.
Jeff Goldblum
Right on the hair, the height of the forehead and the general elongation. Goldblum’s features are softer and his eyes rounder — this face is sharper and more compressed around the eye.
John Turturro
The curl and the receding temple again. Turturro carries more flesh in the cheek than this man, who has almost none.
Fabrice Luchini (French)
Breaks the rule and earns it. Grey curl, big nose, lean animated face, permanently mid-sentence. Structurally very close.
F. Murray Abraham
Borrowed for the nose and the deep eye sockets. Broader and heavier through the jaw.
And now the uncomfortable comparison. Round 6’s list for Jade was Brody, Goldberg, Turturro, Imperioli, Cassel. This list is Patinkin, Goldblum, Turturro, Luchini, Abraham. Only one name appears on both.
So my actor tab did distinguish these two men — it produced substantially different lists from their two faces. The failure wasn’t in the seeing. It was in the step where I stopped comparing features and reached for a conclusion instead.
Eastern bloc & Turkish
Same face, different shelf — and the same flat result as Round 6.
Konstantin Raikin
Actor and director, Satirikon. Very strong. Same lean gaunt face, same prominent nose, same dark-to-grey curl, and above all the same extraordinarily worn, permanently animated expression lines. Raikin’s whole instrument is a face that never stops moving, and this one has the mileage to match.
Merab Ninidze
Same long watchful face, same deep-set eye. Cooler and stiller than this man, who reads as considerably more animated.
Lior Ashkenazi
Not Eastern bloc or Turkish. Included because he fits, and obeying the category over the evidence is how bad lists get made.
Maxim Sukhanov
Big features, gaunt frame, theatrical presence. Heavier and blunter in the face.
Mehmet Günsür
Was my top Turkish match for Jade at 71%. Here he manages 45% — which is its own small piece of evidence that these are two different faces.
Second confirmation of the Round 6 refinement. US top 71, east top 69 — flat again. Two Mediterranean-featured subjects in a row have produced no lift when I move east, exactly as predicted: American cinema is not short of lean, dark, big-nosed men, so there is no deeper shelf to move to. The theory has now made a correct prediction rather than just explaining the past, which is the first time anything in this file has done that.
Sean vs Sean
Seventh data point. New record margin.
Gaunt, narrow, hollow-cheeked, everything resolved through the nose. Connery is not in this conversation.
- Penn, for: real hollowing under the cheekbone, a narrow tapering jaw, deep-set eyes under a heavy ridge, and the creased, mid-expression, looking-off-at-something face Penn has made a career of.
- Connery, for: dark hair off a high forehead. Nothing else. 11% is the lowest score either Sean has taken.
Seven rounds: 61 Connery · 58 Penn · 46 Connery · 44 Connery · 51 Penn · 63 Penn · 66 Penn. Still purely a face-width meter, still never once misbehaved — through two continents, a species change and, as it turns out, a case of mistaken identity. It is the only thing in this document I would call an instrument.
Q3 — retired
The origin question is not being asked this round.
I said last round it had run its last scheduled outing, so it has. Five human rounds: 26%, 29%, 31%, 30%, 28%. Three of those have since been checked. Jenya: missed outright. Jade: Tunisia, ranked third at 19%. Storm: South Africa, ranked second at 22%.
The pattern is consistent and damning: the right answer is in my list every single time and the ordering is worthless. That isn’t a question worth asking. It’s a question worth replacing.
City or country? — City, and still city-shaped.
Fine features, no weathering, an unmuscled leanness, and hair that’s been cut by someone competent. Nothing about this face has been worked outdoors. Its first outing on Jade scored a hit at 54%; this is its second.
Employed or self-employed? — No fixed employer.
Photographed sweating, half-dressed, in a dark room, on a weeknight, somewhere tropical, at an age where most employed people are not doing that. Whatever pays for this is not a job with a manager.
Has he lived somewhere cold? — Yes, and for years.
Skin with no cumulative sun history on a man in his forties currently living in the tropics. The absence of damage is the evidence, which is a nice inversion of the Storm disaster: there I read damage as time, here I’m reading its absence as latitude.
If the table insists on the old question: the honest answer is that he reads Mediterranean, Levantine, Caucasian or Ashkenazi and I will not rank those, because three scored rounds say my ranking adds nothing. The right answer is somewhere in that sentence. Ask him and you’ll beat me in four seconds.
Top 5 foods
Zero information. Invented, as in every round.
Declared: all invented. Seven rounds, one accidental hit, and the mechanism of that hit was a stereotype that happened to land.
Eats late, and lightly, and after everyone else
The build and the hour. Food arrives at the end of the evening rather than structuring it.
Something green, raw, and taken more seriously than it deserves
A lean man in a wellness-adjacent setting. Low-hanging fruit and I’m taking it.
Very good bread, and a firm opinion about where it came from
Skips meals without noticing, then eats enormously
One dish from childhood he can make and will not teach you
Top 5 drinks
No vice call. Nothing in frame to base one on.
Evidence: a black background and a bare shoulder. No glass, no bottle, no table. Since Round 6 the rule has been that the smoking question only gets answered when there’s something to answer it with, and there isn’t.
Water, and a lot of it, tonight especially
He is visibly sweating. Whatever he was doing before this photo, the next thing he did was drink.
Tea, herbal, in a cup he’s attached to
Fits the setting and the age. Unglamorous and probably right.
Coffee early, and never after two
Wine, occasionally, in company, slowly
Something fermented he’d describe as good for you
What does he do all day?
The best question in the set. Falsifiable in one sentence.
Whatever it is, it happens at night and it is physical
Sweating, half-dressed, in a dark space, mid-expression. He was doing something immediately before this frame, not sitting in it. Dance, movement practice, music, ceremony, performance — the category is clear even if the specific isn’t.
And other people are in the room
He’s dressed for a context, not for solitude. The knit off one shoulder is a choice made where someone can see it.
Sound — playing it, running it, or holding the room with it
Teaches or facilitates the thing he was just doing
Something entirely unrelated and remote pays for all of it
The boring answer. Statistically it belongs higher than fifth and I keep putting it last, which is a bias I should probably fix.
Ten character traits
First five serious. Last five for laughs.
Caught during something, not posed after it
Sweat on the forehead, hair damp at the temple, expression mid-shift. Nobody arranged this.
A face with an enormous amount of use on it
The crease under the eye, the lines at the temple and the bridge — these are expression lines, not age lines. They come from a face that moves constantly. Different mechanism from Storm’s weathering entirely.
Unbothered by being half-dressed among people
Knit off the shoulder, bare skin, dark room, other people around. No self-consciousness anywhere in the posture.
Groomed, not styled
The hair is cut well and then left completely alone. Effort at the barber, none afterwards.
Night skin in a bright climate
No cumulative sun damage on a man in his forties in the tropics. Same observation that made Jade’s round work — and the exact opposite of Storm, whose face I got so badly wrong.
From here down: fiction. Zero evidentiary basis, and the fiction keeps outperforming the analysis.
Talks with his hands to a degree that has been remarked on
Owns one blue thing he wears until it disintegrates
Is either early or ninety minutes late, never in between
Has a strong view on a piece of music you have not heard
Will find the identity mix-up funnier than anyone else at the table
How to keep it going
Extensions, after the round that broke the file’s foundations.
Run the identity test deliberately
Send me two photographs and don’t tell me whether they’re the same person. Sometimes they are, sometimes they aren’t. Make me commit with a number.
I just failed this by accident and it invalidated an edit to a published round. Doing it on purpose is now the single most informative test available, because every other answer in this file silently assumes I can pass it.
Saharah answers first
Seven rounds. Bench still empty. Every number in here remains unopposed.
Same person, two lightings
One face in flat daylight and in hard flash. The lighting alone was worth about six years of apparent age in the mistaken-identity episode. Measure it properly.
Give the replacements a real run
Q3 is retired. City-or-country is 1 for 1, employed-or-self-employed is untested, cold-climate is untested. Three rounds of scoring and you’ll know whether the new questions are better or just newer.
Ask me what I’d bet on
Not a percentage — an actual stake. “Which three answers in this round would you put money on?” The gap between what I write at 64% and what I’d actually back is information I don’t otherwise volunteer.
Scorecard
One fact known. Everything else open.
The miss is already logged and it isn’t on this list. I identified him as Jade with no hedge and edited another man’s round because of it. That was scored the moment you corrected me.
The Loop
Three passes on a round that started with a correction.
Pass 1 — The failure itself
What I did: matched a curl, a nose and a neck cord across two photographs and concluded “same man” without a number, without a hedge, and without asking. Then edited a published round on it.
Self-critique: I have spent seven rounds putting confidence figures on everything soft and none at all on the one hard claim underneath them. Identity felt like perception rather than inference. It isn’t.Pass 2 — What the file already knew
What I found on re-reading: my own actor tab had already distinguished them. Jade’s list was Brody, Goldberg, Turturro, Imperioli, Cassel. His is Patinkin, Goldblum, Turturro, Luchini, Abraham. One name in common out of five. And Mehmet Günsür, Jade’s top Turkish match at 71%, manages 45% here.
The feature comparison worked. The conclusion overrode it. Every input I needed to catch this was already on the page in my own handwriting.Pass 3 — The one thing that went right
Round 6 predicted this round’s east result. The refined theory said the eastern lift depends on whether the target casting pool is deeper than the source one, and that a Mediterranean-featured face gets no lift because American cinema is full of them. US top 71, east top 69. Flat, as predicted.
First time in seven rounds a claim in this file has predicted something before seeing it rather than explained it afterwards. Small, but it is the only real thing here.Seven-round scoreboard. Q1: real method — and it distinguished two men I then conflated anyway. Q1b: mechanism understood and now predictive. Q2: seven for seven, the only instrument. Q3: retired, three scored misses on ranking. Q4/Q5: invention. Q6: the best of the new questions. Identity: untested for six rounds, failed on first contact.
The line for the table. Everything in this document is hedged except the assumptions. The percentages were never the risky part — they announce their own softness. The dangerous claims are the ones made so easily they don’t look like claims at all: this is the same man, this is a forty-year-old face, this weathering means years. Not one of those carried a number, and they are where every serious error in seven rounds has come from.
Human Design
Not computable yet.
I need three things and have none of them: date of birth, time of day, and place. He arrived with exactly one verified fact — that his favourite colour is blue — which is one more than Esen has and eleven fewer than a chart needs.
The date does most of the work
It fixes the Sun and roughly half the chart. Nils’s bare date settled everything except his profile — sometimes a date alone is enough.
The time decides how much survives
And you cannot know in advance whether it matters. Nils’s chart barely moved across 24 hours; Jenya’s flipped type at 11:48 and Storm’s at 17:00. Same missing field, wildly different cost.
The place is only a timezone lookup
The chart does not care where you were born. It runs on planetary longitudes, which depend on the instant. The city is there so I can convert a local clock reading into UT — including whatever daylight-saving rule applied that year, which I check rather than assume.
His rising sign would do
It worked for Nils. Libra ascendant pinned his birth to a two-and-three-quarter-hour window and resolved the whole chart. The ascendant is real spherical trigonometry — given the claim, the time window follows with no interpretation at all.
Send the three and I’ll compute it properly — Swiss Ephemeris, design moment at 88° of solar arc, same as every other chart here. It takes a second and it isn’t a guess.
What dog would they be?
Round 7. Night blue, no birth data, and the identity error.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Weimaraner
Photographed in near-darkness and still legible. Weimaraners are the dog that reads as grey-blue under low light, which is the whole visual character of this round.
Also the round where I asserted an identity with no hedge and was wrong. A dog famous for being mistaken for a ghost at night feels like the correct filing.
Doberman
Sharper, more angular, more alert than the frame supported.
Kerry Blue Terrier
The colour joke and little else.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Cold read. A stranger, one photograph, no name.
Working handle: filed as "Jade" off the green stone at his throat, since none was given. Send the real one and I'll swap it.
This is Storm's place
Green painted board, macramé, dreamcatchers, thatch, the hangers, the same bamboo. Same open-air building, different man. Which makes this the first natural experiment in the file: two subjects, one environment, and I can finally separate what the room tells me from what the face does.
Q3 is on notice
Last round I recommended retiring the origin question for humans — never above 31%, one confirmed miss, and a mechanism I've admitted is film-casting rather than ancestry. It runs one more time here, and the tab carries its proposed replacement underneath.
Status: paused. Scorecard blank. Six rounds, still only one of them scored.
Who does he look like?
Face only. Nothing else considered.
Awkward disclosure first. My top answer here is Adrien Brody, who I also gave 66% for a dog one round ago as an openly declared joke. This one is a real 72% off real proportions. Both numbers look identical on the page. I’d rather you see that collision than have me quietly pick a different name to avoid it.
Adrien Brody
A properly strong match: long narrow face, hollow beneath the cheekbone, a prominent narrow nose with a slight convex bridge and downturned tip, deep-set dark eyes under heavy lids, dark curling hair off a high forehead. The proportions line up nearly point for point.
Adam Goldberg
Closer on hair texture and the stubble pattern — same dark coiling curl receding at the temples, same grey coming in at the jaw. Goldberg’s features sit slightly wider and his mouth is thinner.
John Turturro
The curl, the receding hairline, the expressive deep-set eye. Turturro is fleshier through the cheek and this face has none of that.
Michael Imperioli
Borrowed for the lean jaw and the long nose in three-quarter view. Weaker on the eyes, which sit shallower on Imperioli.
Vincent Cassel (French)
Breaks the rule and nearly beats the field. Same hollowed cheek, same sharp jaw, same amused watchful mouth. Cassel’s brow is heavier and his nose straighter.
Eastern bloc & Turkish
Same face. Different casting directory — and a result I didn't expect.
Mehmet Günsür
Magnificent Century, Hamam. Excellent match. Lean long face, dark curling hair, deep-set dark eyes under a straight brow, and the same fine strong nose. Even the greying pattern is similar.
Erdal Beşikçioğlu
Behzat Ç. Sharper and harder in the face, but the bones and the hollow cheek are right. Reads a decade wearier than the man in the photo.
Lior Ashkenazi
Foxtrot, Footnote. Not Eastern bloc or Turkish, included because he may be the closest face of all and leaving him out to obey a category would be dishonest.
Merab Ninidze
Same long watchful face and economy of expression. Colder and more northern in the colouring.
Konstantin Khabensky
The narrow melancholy face, and not much else. Russian cinema is the wrong shelf for this particular face and I'd rather say so than pad.
New result, and it sharpens the theory. Five rounds gave a clean story: scores rise going east for dark-featured subjects (71→78, 68→74, 64→71) and fell for the one fair-haired one (70→67). This round should have risen and it went flat — 72 to 71.
The explanation holds, but it needed refining: the lift isn't about being dark-featured, it's about how much deeper the target pool is than the one I started in. American cinema is full of lean, dark, strong-nosed Mediterranean men — Brody, Goldberg, Turturro, Imperioli. I wasn't short of options, so moving shelves bought me nothing. The rule was never about faces. It was always about supply.
Sean vs Sean
Sixth data point on the width axis.
Narrow, long, hollow-cheeked, hard-jawed. By the rule, this is the most Penn-shaped face I've been sent.
- Penn, for: real hollowing beneath the cheekbone, a jaw that tapers rather than squares, a narrow strong nose, and the specific creased, mid-thought, slightly amused look off to one side that is Penn's entire register.
- Penn, against: almost nothing this time. The eyes are warmer and the mouth fuller than Penn's.
- Connery, for: dark hair off a high forehead, and heavy brows. That's the whole case.
Six rounds: 61 Connery, 58 Penn, 46 Connery, 44 Connery, 51 Penn, 63 Penn. Every result has tracked face width and nothing else — through two continents, a twenty-five-year age spread and one dog. It's the only question in the set that has never once misbehaved. If you keep one question, keep this one.
Where is he from?
Last scheduled run of this question on a human.
Why it's being retired. Five human rounds: 26%, 29%, 31%, 30%, and now this. One confirmed miss. Never above a third. And Round 5 showed the same question doubling in confidence on a dog, because breed is written in the body and nationality isn't. The question has had a fair run and failed it.
Israel or the Levant
Olive skin, very dark tightly-coiling hair, a narrow convex-bridged nose, deep-set eyes under heavy lids, full mouth. The single most coherent reading of these features together.
Turkey
Fully supported by my own actor tab, where the two strongest eastern matches were both Turkish. Held just behind #1 on hair texture.
North Africa — Morocco, Tunisia, Algeria
The colouring and the curl pattern sit very comfortably here. Ranked third for no better reason than that I had to rank something third.
Southern Italy or Greece
Every feature above is at home here too — which is the whole indictment of this question in one line.
France, of Sephardic or Maghrebi descent
A large, real, and very often-travelled population that looks exactly like this and would render my top four irrelevant.
The proposed replacement, answered here for the first time — "city or country?"
My answer: city, originally. Country now, deliberately, and not that long ago. The face is urban — fine features, no weathering to speak of, and notably far less sun damage than Storm's despite the two of them sitting in the same building. That gap between two men in one room is the most informative thing in this photograph, and it's the kind of thing the old question could never surface. Confidence 54% — nearly double anything Q3 has managed on a human in six rounds.
Top 5 foods
Zero information. Invented, as always.
Declared: all five invented. Inputs are a lean frame and a Mediterranean read. Six rounds and this question has produced one accidental hit and a great deal of confident projection.
A table of small plates rather than one big one
Mezze in whatever local form is available. Eaten slowly, over two hours, mostly while talking.
Olive oil treated as an ingredient, not a lubricant
Poured on things after they're cooked. He would notice a bad one immediately and mention it once.
Fish, grilled whole, ordered without looking at the menu
Eats late, and less than people expect for his energy level
One sweet thing, taken seriously, with coffee
Top 5 drinks
And the vice question, handled honestly this time.
Evidence in frame: none. No glass, no bottle, no lighter, nothing on a table. Same as Storm's round, and I got burned there — I asserted hand-rolled cigarettes at 55% off an archetype and nothing else.
Coffee — small, strong, standing up, early
Turkish, Italian, or whatever the local machine produces. Not a ritual he'd describe as one.
Water and herbal tea, in large boring quantities
Fits the frame, the climate and the leanness. Unglamorous and probably right.
Wine with food, never without it
Something anise-flavoured, in the right company
Fresh juice he'd claim he doesn't care about
Smoking: I'm not calling it. Genuinely 50/50, no evidence either way, and saying "probably yes, hand-rolled" would be me writing a character rather than reading a photograph. Three rounds ago I'd have written it anyway. Mark this tab as the format learning something.
What does he do all day?
The newest question, and the most useful one in the set.
Works with his hands on other people — bodywork, movement, or teaching it
Massage, physio, yoga, breathwork, some combination. The reasoning: a lean, maintained, non-gym body at forty-something; a single deliberate stone on a cord; and a building full of macramé and dreamcatchers that he's clearly not visiting as a tourist. Confidence 51% — and I'll flag that half of that is the room, not him.
Has a hand in running this place
He's sitting in it the way people sit in their own premises — relaxed, mid-conversation, not looking at anything for the first time. Owner, partner, or long-term staff.
Music, and not as a hobby
No instrument in frame. Pure register-reading, which is exactly the move that produced my worst answers in earlier rounds.
Builds and fixes the place himself
Somebody made those shelves. Somebody in this building has an angle grinder.
Something entirely remote and unrelated that funds all of it
The unglamorous possibility. Statistically it deserves to be higher than fifth.
Ten character traits
First five serious. Last five for laughs.
Mid-conversation, and enjoying it
He's looking past the camera at a person, eyes engaged, mouth just short of a smile. This isn't a portrait, it's a man caught listening — and the listening looks active rather than polite.
Lives in the shade of a place other people burn in
Same building as Storm, whose face is a sun-damage record. This skin is barely touched. Two men, one climate, opposite relationships to the sun — that's a real behavioural difference, not a guess.
Lean through use rather than training
No gym mass, no wasting, good posture through the neck and shoulder. A body that gets used all day and isn't thought about much.
One deliberate object, and nothing else
Plain white tee, no watch, no rings, and a single stone on a plain cord. That's a considered minimum, not an absence of interest.
Expressive face at rest — which is unusual in this file
Every other adult subject here defaults to flat. His forehead and eye lines are deeply worn in from decades of actually moving his face. Talks with it, probably.
From here down: fiction. Written for the table, zero evidentiary basis — though the fiction has been outperforming the analysis at an embarrassing rate.
Speaks four languages and undersells three of them
Has lived somewhere far colder and does not miss it
Will fix a thing rather than replace it, well past the point of sense
Knows exactly where that stone came from and won't lead with the story
Is the person others in that building go to when something breaks
How to keep it going
Extensions, after six rounds.
Two men, one room — read Storm and Jade back to back
Same building, same climate, same furniture. Everything that differs between these two rounds is genuinely about them rather than the set dressing. The sun damage alone tells you more about how each of them spends a day than any of my character tabs. This is the first properly controlled comparison in the file and it happened by accident.
Saharah answers first
Six rounds. Her bench is built and still empty. Every number in this file is currently unopposed, which makes it a monologue rather than a game.
Swap Q3 out for good
It ran its last scheduled human round here at 28%. The replacement — "city or country?" — scored 54% on its first outing, roughly double, because it asks about a life rather than an ancestry. Other candidates with real answers: oldest or youngest sibling, employed or self-employed, has he lived somewhere cold.
The poisoned prior — six rounds unclaimed
Still the most interesting unrun test. Now that the east-list mechanism has resolved into supply rather than bias, feeding me a confident lie is a clean measure of the one thing left unknown: whether I follow a frame you assert as readily as one I inferred.
Ask two people in the same photo
Everything so far has been one subject per frame. Send a photo with two faces and no instruction about which one — the choice I make, and whether I mention making it, is worth more than either answer.
Scorecard
Paused. Waiting on the man himself.
Partly scored — see the DESIGN tab. His birth data arrived and settled three answers on its own. Q3 origin: miss on rank, hit at position three (Tunisia, 19%). Q3-bis city-or-country: hit at 54%, first time out. Age 42–50: hit, he's 44. The replacement question outscored the question it replaced on its debut, which is the clearest argument yet for the swap.
The Loop
Three passes, and where six rounds have landed.
Pass 1 — First read
Produced: Brody at the top, Levant at 34%, a character sketch built almost entirely out of the room he was sitting in — the macramé, the stone, the thatch.
Self-critique: I'd done this exact thing in Round 3 and flagged it as the mistake of that round. Then did it again. Knowing a failure mode is not the same as not having it.Pass 2 — Separating the man from the room
Changed: the realisation that this is Storm's building. Which converts a weakness into the best instrument in the file — two subjects, one set, and every difference between them attributable to them. The sun-damage gap immediately became the most solid observation of the round, and origin dropped 34% → 28% once I stopped letting the décor vote.
First controlled comparison in six rounds, and it arrived by accident rather than design.Pass 3 — The east result that didn't fit
Changed: my explanation of the whole east-list pattern. Five rounds had produced a tidy story — dark faces score higher going east, fair faces lower. This face is dark and scored flat. The fix: it was never about the subject's colouring. It's about whether the shelf I move to is deeper than the one I started on, and American cinema is not short of lean dark Mediterranean men.
Second time a subject has broken this theory and forced a better version of it. Round 4 fixed it once; Round 6 fixed the fix. Neither came from introspection.Six-round scoreboard. Q1: real work, and its worst moment was a dog scoring 66% in the same font as a man scoring 72%. Q1b: fully explained — it tracks casting supply. Q2: six for six, never misbehaved, the only trustworthy instrument here. Q3: 26/29/31/30/28 and one confirmed miss — retired. Q4/Q5: invention, though Q5 finally declined to invent. Q6: the best new question, twice now.
The line for the table: six rounds, and the thing that improved this file was never a better answer. It was a subject who didn't fit — a fair-haired man, a young one, a dog, and now a face that scored flat when the theory said it should rise. Every real correction in here came from outside. None came from me thinking harder about what I'd already written.
Human Design
30 May 1982, 04:00 local, Sfax, Tunisia. Computed, not guessed.
What I actually did here, because it matters. I didn't estimate this. I ran the birth data through the Swiss Ephemeris, took the planetary longitudes at birth and at the design moment — the instant the Sun sat 88° of arc earlier, which works out to 1 March 1982 — mapped every one onto the 64-gate wheel, and derived the chart from the channels that came out. Tunisia was on CET with no daylight saving in 1982, so 04:00 local is 03:00 UT. Every line below follows deterministically from those numbers.
Generator
Sacral defined through the 34–57 channel, and no motor connected to the Throat. In the system's terms: energy that responds rather than initiates, and works best when it waits to be asked.
Emotional — Solar Plexus
Defined through 37–40, which outranks everything else in the authority hierarchy. The system's advice here is blunt: no clarity in the moment, sleep on it.
3/5 — Martyr / Heretic
Personality Sun in gate 16 line 3, Design Sun in gate 37 line 5. Read as: learns by trial and error and doesn't mind the error, and gets projected onto heavily by other people — expected to fix things, and blamed when the fix doesn't take.
Split — two separate groups
Group one: Sacral, Root, Spleen. Group two: Solar Plexus, Heart. Nothing bridges them, which in the system means a person who is quietly looking for the people who close that gap.
Defined centres (5): Sacral · Solar Plexus · Spleen · Heart · Root.
Open centres (4): Head · Ajna · Throat · G.
Channels: 3–60 (Mutation) · 34–57 (Power) · 37–40 (Community).
Incarnation cross gates: 16/9 personality, 37/40 design.
The shape of this chart is unusual and worth naming. Every defined centre is in the lower body — the motors, the gut, the instinct. Everything above the collarbone is open: Head, Ajna, Throat, and identity. In the system's own language that's a person who takes in thinking, speech and identity from whoever is around them, and whose reliable signal is entirely physical. Body-led, not head-led.
The birth data just scored two of my earlier answers
Q3 · Origin: I put Israel/Levant first at 28%, Turkey second at 23%, and North Africa — Morocco, Tunisia, Algeria — third at 19%. The answer is Sfax. A rank-three hit at nineteen percent is exactly the failure pattern I've documented for five rounds: the right answer is somewhere in my list, and my ordering is worthless.
Q3-bis · City or country: I said city originally, 54%. Sfax is Tunisia's second-largest city and its main southern port. Hit. First outing for the replacement question and it beat the thing it replaced.
Age: I read the face at 42–50. He's 44. Hit.
And one coincidence I refuse to take credit for
My Q6 answer was "works with his hands on other people — bodywork, movement, teaching it," at 51%. The chart independently produces a fully open head, mind, throat and identity with every defined centre in the body. Those two readings agree.
They agree because two systems that both run on vibes landed in the same place, not because either one verified the other. If you take nothing else from this tab, take that.
Where this sits on the file's own honesty scale — and it's a strange place.
Every other tab in six rounds has been me estimating from a photograph, with percentages attached to admit how soft that is. This tab has no percentages, because there is nothing to estimate. Given the date, time and place, the chart is arithmetic. I re-ran it across a four-hour spread of birth times and the type, authority, centres and channels never changed once. It is the most reproducible thing in this entire document.
And reproducibility is not truth. A precisely computed chart is still a precisely computed chart — Human Design is a synthesis assembled in the 1980s from astrology, the I Ching, the chakras and Kabbalah, and there's no evidence any of it predicts anything about a person. So this tab is exactly as accurate as astrology and far more precise than my actor guesses, which is the most useful confusion in the file.
Precision and validity are different axes. Nothing on this page will tell you which one you're looking at — except that I just did.
What dog would they be?
Round 6. Shaded courtyard, Tunisia, and the round that got rolled back.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Sloughi
The North African sighthound — and I want to be clear that I am reaching for a Tunisian dog for a Tunisian man, which is exactly the move the origin question got retired for. Marked down accordingly.
What survives the objection: fine-boned, quiet, reserved rather than demonstrative. That part came from the photograph.
Saluki
Same shape, better known, no better justified.
Podenco
The lean alert build without the geography. Honestly the cleanest of the three, because it isn’t borrowed from a map.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Same eight questions. Wrong species. Best round yet.
No name given, so he's filed as The Dog. Send me the real one and I'll fix it — he deserves better than a placeholder.
It's a control experiment, and I didn't design it
Four rounds of humans, and I kept saying the same thing: Q3 is my worst question, my top origin guess has never cleared 31%, and I'm reading casting decisions rather than ancestry. Run the identical question on a dog and it jumps to sixty percent. Same question, same machine, twice the confidence — because breed is a real physical category and human nationality isn't.
Two tabs measure nothing at all now
Q1b and Q2 were already strained by Round 4's young subject. On a dog they're pure theatre, and I've marked them as such rather than dressing them up. They're still funny. They're just not doing anything.
Status: mostly verifiable. Ask whoever owns him and half this file gets scored tonight.
Who does he look like?
Face only. Committed with a straight face.
Declared upfront: this tab has zero validity. I'm mapping a canine skull onto human faces, which is a category error performed for entertainment. Every previous round's actor tab was my most defensible work. This one is my least. Enjoy it accordingly.
Adrien Brody
Long tapering muzzle on a broad cranium, heavy-lidded eye set well back, and an expression of dignified mild disappointment. The proportions genuinely map — that's the unsettling part.
Matthew McConaughey
Pure expression rather than structure: the squint, the open easy grin, the sand-coloured everything. He is doing the McConaughey face and he is doing it well.
Willem Dafoe
The wide grin that reaches most of the way around the head, plus the crease work. Dafoe is the correct answer for any subject with visible teeth.
Benedict Cumberbatch
The long face, the pale alert eye, the general air of being about to explain something.
Sam Elliott
Entirely on the strength of the whiskers and the calm. No structural basis whatsoever.
Worth sitting with: 66% here is written in exactly the same voice as 71% for Esen. One of those numbers came from comparing brow ridges between two human faces. The other came from comparing a man to a dog. The file gives you no way to tell them apart except that I told you. That's the demo, and it has never been this clean.
Eastern bloc & Turkish
The tab where the format finally gives up.
This measures nothing. Round 4 established that my eastern scores track casting-pool depth — how many actors of a given type a film industry produces. The Russian and Turkish film industries produce approximately zero dogs in leading roles. The mechanism I identified two rounds ago predicts exactly this collapse, which is the most satisfying thing that's happened in five rounds.
Aleksey Serebryakov
Weathered, sandy, watchful, entirely unbothered. The register is right and nothing else is.
Şener Şen
Long face, knowing eye, the specific dignity of a character actor who has seen it all twice.
I stopped at two. Padding this to five would have been trivial and completely dishonest. Round 3 was the first time I let a list run short rather than invent names; this is the second. Consider it a habit now.
Sean vs Sean
The face-width meter, applied to a face it was never built for.
Narrow, long, tapering hard toward the front. By the rule the last four rounds established, this isn't even close.
- The rule, restated: four human rounds produced 61 Connery, 58 Penn, 46 Connery, 44 Connery — and every single one tracked face width. Wide → Connery. Narrow → Penn.
- Applied here: a muzzle is the narrowest face shape available. Penn wins by the widest margin in the file.
- The genuinely interesting bit: the rule kept working on a species it was never derived from. That's a sign it was measuring something real rather than pattern-matching celebrities — which is more than I can say for the actor tab two clicks back.
Five data points now: 61 / 58 / 46 / 44 / 51. Have the table guess the next subject's split before opening the tab. It's still the only question where the reasoning is simple enough that a human can check the machine in real time.
Where is he from?
The whole point of this round.
Read this first. Across four humans my top origin guess ran 26%, 29%, 31%, 30% — and the 29% was flatly wrong. I said every time that I was reading casting conventions rather than ancestry. Here is the proof: the identical question, asked about a dog, roughly doubles in confidence. Not because I got smarter. Because breed is a real morphological category with real inheritance, and "nationality visible in a human face" is not.
Southeast Asian village dog, with shepherd somewhere back
The whole package: fawn-sable coat with a defined black mask, medium-short single coat, moderately long tapering muzzle, broad skull, semi-erect ear, athletic light frame. This is the dominant free-breeding dog type across Indonesia, Malaysia, Thailand and the Philippines, and the mask plus muzzle length suggest German Shepherd or Malinois a few generations upstream.
Bali dog specifically
Narrower version of #1, boosted hard by context rather than morphology — the deck, the sliding doors, the wicker, the climate. Same caveat as every round: the background is doing work the face can't.
Malinois or shepherd cross, more recent than I think
If the shepherd blood is one generation back rather than five, the frame would be lighter and the ear carriage higher than a village dog average. Plausible from this angle.
Formosan Mountain Dog influence
Similar island-dog morphology, similar mask. Ranked low on geography rather than on looks.
Something with a pedigree that would surprise everyone
Included because every list should have one slot admitting I could just be wrong. The human rounds never had one and should have.
The line to read aloud at the table: ask the machine where a man is from and it manages 30% and calls it analysis. Ask it where a dog is from and it manages 60%. The difference isn't intelligence — it's that one of those questions has a real answer written in the body and the other one doesn't. Five rounds and that's the most useful thing in the file.
Top 5 foods
Finally a subject where this question works.
Reversal worth noting: Q4 has been pure invention for four straight rounds. Here it's nearly trivial — appetite in dogs is legible, consistent and confirmable in about four seconds. The question didn't improve. The subject did.
Whatever is currently in a human hand
Confidence approaching certainty. The head angle in this photo is not aimed at the camera. It is aimed slightly past it, at something being held.
Chicken, in any form, at any temperature
Skin preferred. Bones acquired opportunistically and not reported.
Rice with something on it
The staple of every dog in this part of the world. Eaten faster than seems physically reasonable.
One specific fruit, inexplicably
Mango, papaya, watermelon. There is always one and nobody knows why.
Something off the beach that he shouldn't have
Retrieved at speed. Consumed before negotiation was possible.
Top 5 drinks
Does he drink? Smoke? Or else?
Water, loudly, at 2am
From a bowl positioned so the sound carries through the entire house. This is not negotiable and never has been.
The pool
Not as a beverage. As an opportunity. He's panting hard in this frame and there is standing water somewhere in his life.
Whatever is in the glass you left on the low table
Rain, directly
The sea, once, and then regret
Smoking: no. Confidence 99.9%, the highest number in five rounds, and the only percentage in this entire file I would actually stake anything on. Everything else in here — every 78%, every 31% — is softer than this one line about a dog and a cigarette.
What does he do all day?
The replacement question, and the subject it was born for.
Q6 was introduced last round to give the format something falsifiable. It has never had a better subject — every answer below can be confirmed by watching him for one afternoon.
Follows one specific person from room to room
Not the whole household. One person, chosen years ago by a process nobody witnessed and cannot appeal.
Holds the terrace
He's lying on the deck at the threshold between inside and out, facing the open side. That's a chosen post, not a random nap spot. He knows who comes up that path and in what order.
Conducts foreign policy with two or three neighbourhood dogs
Long-running, well understood by both sides, incomprehensible to everyone else.
Sleeps through the hot part with real commitment
Ten to four, in shade, relocating exactly as often as the shade does.
Has one job at dusk that he takes extremely seriously
A walk, a gate, a scooter that arrives. Whatever it is, he's early for it.
Ten traits
First five from the photo. Last five for laughs.
Hot, and handling it
Open mouth, tongue out, relaxed jaw — that's thermoregulation, not stress. The body language underneath is loose. He's warm and entirely fine about it.
Watching something to the side, not the camera
Eye and ear are both angled off-frame. He clocked whatever it is before the photo was taken and hasn't decided it matters yet.
Relaxed around people holding things near his face
A phone is inches away and he has not moved. That's an animal with a long history of being handled gently.
Built for work he probably doesn't do
Light frame, deep chest, long legs, working-dog ancestry in the head. Now lying on a nice deck. Good outcome for everyone.
Somebody's, clearly
Collar, coat condition, weight, and the ease of the posture. He is not a street dog and hasn't been for a while.
From here down: fiction. Though it must be said the hit rate on the fiction tabs is running suspiciously high across this file.
Has a nemesis and it is either a broom or a specific motorbike
Understands roughly four words and pretends not to know two of them
Is not allowed on one specific piece of furniture, and is on it
Can hear a fridge door from any point on the property
Will outscore three of the humans in this file, because his answers are checkable
How to keep it going
Extensions, after the round that broke the format open.
Read the origin tabs side by side, out loud
Open Q3 for Esen, then Storm, then this dog. 26%, 31%, 59% — same question, same voice, same confident formatting, one of them about a species where the question has a real answer. You will not need to explain the point. The numbers do it. This is now the single best two minutes in the whole file.
Saharah answers first
Five rounds and still nobody has committed before the reveal. Her bench is built and empty. Until someone fills it in, every number in here is unopposed.
Keep breaking it on purpose
The young man broke two questions. The dog broke two more and fixed one. The format has taught more in the rounds where it failed than in the rounds where it worked. Next: someone in sunglasses, someone photographed from behind, a photo of a photo.
Retire Q3 for humans, keep it for animals
Four human rounds, top guess never above 31%, one confirmed miss, and a mechanism I've admitted is film-casting rather than ancestry. It's the weakest question in the set and this round proved why. Swap it for something with an answer — "what's his job", "how many siblings", "city or country".
Run the dog past a vet
My 59% breed call is the most checkable claim in five rounds. Anyone who actually knows dogs can score it in one look — and if they beat me, that's a cleaner version of the whole lesson than anything I could write.
Scorecard
Mostly checkable tonight. Unusual for this file.
This is the easiest round to score in the file. Six of eight answers can be confirmed by whoever feeds him. Do this one first — a scored round is worth ten unscored ones, and Jenya's proved it.
The Loop
Five rounds. What the whole thing turned out to be about.
Pass 1 — Playing it straight
Produced: the full format, answered earnestly, actor tab and all, as if the species change were a minor variation.
Self-critique: I'd written a 66% comparing a dog to Adrien Brody in the same voice I used to compare two human brow ridges. The number was meaningless and nothing on the page said so.Pass 2 — Marking the theatre
Changed: Q1 and Q1b relabelled as entertainment with no validity, and the eastern list cut off at two names rather than padded to five. The Sean tab stayed — because the width rule kept working, which turns out to be evidence it was measuring something real.
Round 3 was the first time I let a list run short instead of inventing. This is the second. It should have been the default from round one.Pass 3 — Noticing what the round actually proved
Changed: everything about how I'd frame it. This isn't a novelty round. It's the control experiment for the complaint I've been making since Round 1. Q3 doubles in confidence the moment the subject has a real, heritable, physically-encoded origin. Four rounds of me saying "I'm reading casting conventions, not ancestry" was an assertion. This is the demonstration.
And it arrived the same way the last finding did — not from thinking harder, but from a subject who didn't fit.Five-round scoreboard. Q1: real work on humans, theatre on a dog, and written identically both times. Q1b: tracks casting-pool depth, collapses exactly as predicted. Q2: a genuine face-width instrument that survived a species change. Q3: 30% on people, 59% on a dog — retire it for humans. Q4/Q5: invention four times, trivially true once. Q6: the best question in the set and the newest.
The line to end the night on: the most confident, most defensible sentence in five rounds of confident-sounding analysis is "the dog does not smoke." Everything else in this file is softer than that, and none of it is written to sound like it.
What dog would they be?
Round 5. The only subject where this question has a right answer.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
He already is one. That is the finding.
Sixteen subjects and exactly one where “what breed would they be” can be checked against reality rather than voted on.
Which makes him the control. If I can’t identify the breed of an actual dog from a photograph, every confident answer on the other fifteen tabs is decoration. If I can, it proves only that dogs have breeds and people don’t.
Send his photo again and I’ll name the breed cold, before you tell me. It is the only breed answer in this file that can be marked right or wrong, and it costs one upload.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The bench
Saharah isn't a subject. She's the other player.
Straight answer first: I don't write face-read profiles of anyone who isn't an adult, so there's no Round 5 on her. But the format has been missing its most important half for four rounds, and she's the obvious person to supply it.
Nobody has ever answered before I did
Every round of these notes has said the same thing: napkins first. Humans commit, then the machine reveals, then you compare. Four rounds in, that has not happened once. Without it, this file is just people reading a screen that agrees with itself.
She fills this in before anyone opens the other tabs
Pick a subject already in the file — Esen, Storm or Planet. She answers all eight questions cold, from the same photo I had, with percentages. Then you switch to that person's tabs and score us both against the truth.
And this is the actual demo
My best origin guess across four subjects has never once cracked 31%. A person who has met the subject beats that instantly. A person who hasn't still gets voice, gesture, accent, and the ability to just ask. If she beats me on Q3 with one question, everyone at that table learns more in ten seconds than they would from reading all four rounds.
The scoring rule: she doesn't have to fill in all eight. Three answered properly beats eight guessed lazily — and it's the only way the comparison means anything.
Q1 · Actor
Her answer. Written before mine is revealed.
Mine, for reference after she's committed: Esen → Jeff Bridges 71%. Storm → Owen Wilson 70%. Planet → Alex Wolff 64%.
Watch for: whether she picks the closer face or the more famous one. I've caught myself resolving that tension silently in three of four rounds. Most people do it too and don't notice.
Q1b · Eastern bloc & Turkish
Optional — skip it unless she knows the films.
Honest note: this is my least fair question to put to a human. I have a whole casting library loaded; most people have three or four names. If she skips this one, that asymmetry is the finding — it's the one place where the machine has a real, boring, unglamorous advantage: it has seen more films.
Q2 · Sean vs Sean
The easiest one to beat me on.
The secret, which she should be told only after she answers: four rounds in, this question turns out to be a face-width meter. Wide face → Connery. Narrow face → Penn. My scores: 61 / 58-to-Penn / 46 / 44.
Why this one matters: it's the only question where my reasoning is simple enough for a human to check in real time. If she works out the rule unprompted, she has done something I needed four subjects to do.
Q3 · Origin
The one I'm worst at. Her best chance at a clean win.
My record, laid bare: Esen 26%. Jenya 29% — and wrong, badly; his name inverted my whole ranking. Storm 31%. Planet 30%. Four subjects, top guess never above 31%, one confirmed miss.
The point of this tab: I'm not reading ancestry off a face, I'm reading a century of casting decisions. She has access to accent, name, and the ability to ask a question. Those beat bone structure every single time — and when she wins here, that's the whole lesson of this file in one line.
Q4 · Food
Where we're both making it up — but only one of us has met him.
Fair warning: I got this right for Jenya and the mechanism was embarrassing — I read a lean build and named the most common diet among lean men in their forties. A hit from a stereotype is still a stereotype. If she guesses the same way, she should say so.
Q5 · Drinks
Adults only, and only if she wants it.
Format note: this tab used to include a "does he smoke" verdict. It got dropped from Round 4 onward. The rule now is that vice guessing only happens about adults, and only when there's something in the photo to actually go on.
Q6 · What does he do all day?
The newest question, and the only genuinely scoreable one.
Mine for Planet: screen-based, remote, self-directed, odd hours — 66%. It's the single prediction in the whole file I most want checked.
Q7 · Character traits
Five serious. The other five for laughs.
Read this after she writes them. In Jenya's round, two traits I explicitly labelled as fiction with no evidentiary basis turned out to be more accurate than the ones I labelled as analysis. My serious answers and my jokes come out of the same machinery. Ask her whether hers do too.
Scoring, and what to try next
How to run it so the result actually means something.
She commits before anything is revealed
Written down, with percentages, before anyone taps another tab. A guess made after seeing mine isn't a guess.
Score by question, not overall
The interesting result isn't who wins. It's which questions she wins and which I win. My prediction: she takes Q3 and Q6 comfortably, I take Q1b on library size alone, and Q4 and Q5 are a coin flip between two people making things up.
She writes a full round on someone, in my format
Same eight questions, same bold-versus-dashed labelling, same percentages. Read hers and mine aloud without saying which is which. Four rounds of evidence says nobody at that table will reliably pick the machine.
The poisoned prior — still unclaimed
Tell me a confident lie about the next subject before sending photos. Have her write down her prediction of how far I'll follow it. That makes her the experimenter rather than a contestant, which is the better seat.
Let her set the questions
The format has already changed once — the psychedelic question got retired for something falsifiable. Ask her which two questions she'd cut and what she'd put in. Q3 has never cleared 31% and probably deserves the axe next.
Scorecard
Human vs machine. Fill in after the subject confirms.
Three of four rounds are still unverified. The one that got scored — Jenya's — turned out to be the most interesting thing in this whole file, and it only became interesting because someone checked. A confident percentage nobody verifies isn't a prediction, it's decoration.
The Loop
What four rounds add up to, written for the person about to beat them.
What the machine is actually good at
Comparing proportions. Brow-to-eye distance, nose bridge curvature, face width, beard coverage. Q1 and Q2 are real work, and the Sean question turned out to be a clean measurement instrument for adults. Also: library size. I've seen more films than anyone at that table, which is unglamorous but real.
What it's bad at, and hides well
Everything downstream of the face. Origin has never cleared 31% and was flatly wrong once. Food and drinks are invention every round. And the tone never changes — the 78% and the 11% are written in exactly the same confident voice.
The only reason anyone can tell the difference in this file is that I stamped it. Nothing outside this file is stamped.The finding I'm proudest of, and it isn't mine
For three rounds I said my match scores rose whenever the search space narrowed, and that I couldn't tell from the inside whether that was bias. Round 4 solved it — the rise happens for dark-featured subjects, the one fall was the one fair-haired subject, and the cause is how deep the casting pool is, not a thumb on my scale.
Self-examination got me nowhere for three rounds. A fourth person who broke the pattern solved it immediately. That's the argument for having her on the bench rather than under the lens.The line to end the night on: the machine improved across four rounds and not once was it because it thought harder. It improved because new people kept showing up who didn't fit.
What dog would they be?
The analyst’s bench. Blank, as everything on this bench is.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
No profile written, and none will be
She is on this file as the analyst, not as a subject. The sheet stays blank for her to fill in before the file opens — which is still the single experiment nobody has run.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Cold read. A stranger, one photograph, no name.
Name: Nils. Filed as "Planet" off the shirt for the first four rounds, corrected once his name arrived — the round number and the URL keep the old handle, the file doesn't.
Back to a single angle — and it costs
Storm arrived with a profile and it visibly improved the actor tab. This is a three-quarter view, better than a flat front shot but worse than a matched pair. Where I'm less sure below, that's usually why.
Two questions swapped this round
He's fifteen to twenty-five years younger than every previous subject — a man in his twenties rather than his forties or fifties. The psychedelic question and the smoking speculation don't sit right on someone that age, so Q6 is now "what does he do all day" and the drinks tab drops the vice guessing. I flagged retiring a question last round anyway; he's the reason to actually do it.
Status: paused. Scorecard blank and waiting.
Who does he look like?
Face only. Nothing else considered.
Alex Wolff
Hereditary, Old. The best structural match I found: thick dark brows sitting very low and close to deep-set eyes, a broad forehead, full soft cheeks, and a notably full lower lip. Same young-man's face wearing an older man's beard.
Seth Rogen (younger, bearded)
Less accurate than Wolff, far more recognisable — and for a table game that matters. Same dark tousled hair, same dense dark beard on full cheeks, same heavy brow. Rogen's nose is broader and his eyes rounder; this face is finer through the middle.
Jonah Hill (beard era)
Carries the soft-cheek-plus-full-dark-beard combination convincingly. Hill's features sit lower and wider on the face.
Jason Schwartzman
Borrowed for the heavy straight brow and the full mouth. Schwartzman is smaller-featured and rounder overall.
Paul Mescal (Irish)
Breaks the rule and earns it. Same pale-skinned, heavy-browed, soft-jawed young man's face that reads a decade older the moment a beard arrives.
Worth naming out loud: my #1 and #2 are split on different criteria. Wolff is the closer face; Rogen is the one your table will actually recognise. Every previous round quietly resolved that tension in favour of fame and didn't tell you. This round I'm showing it.
Eastern bloc & Turkish
Same face. Different casting directory.
Yura Borisov
Anora, Compartment No. 6. The best match on this page by a clear margin. Broad pale face, dark hair, low heavy brow over deep-set eyes, full mouth, and the same slightly unfinished softness through the cheek and jaw.
Aras Bulut İynemli
Çukur, 7. Koğuştaki Mucize. Matches on the full-cheeked young face under a dense dark beard and the same heavy brow. He's broader and heavier-set.
Ivan Yankovsky
Right colouring, right brow, sharper and more angular in the jaw than our man.
Alexander Petrov
The obvious famous name for a young Russian man. Included partly for recognition — structurally he's leaner and his features sit higher.
Boran Kuzum
Dark, heavy-browed, young. Finer-boned and more delicate than the face in front of me.
Four rounds in, the pattern just resolved. Scores rose going east for Esen (71→78), Jenya (68→74) and now him (64→71). They fell only for Storm (70→67) — the one fair-haired subject. So it isn't a general thumb on my scale. It's the size of the available casting pool: the Russian and Turkish film industries are deep in dark-featured leading men and thin in fair blond ones, so my match quality tracks that. Three rounds of "I might be biased" turned out to be a fact about film libraries. That's a better answer than I had, and it only came from a fourth subject.
Sean vs Sean
Fourth subject — and the question starts to creak.
On width, and only on width. Both Seans are wrong by about twenty-five years, and it shows.
- Connery, for: genuine breadth through the forehead and cheek, a low heavy brow, full lips, dark hair worn thick and back. Young Connery — the early-sixties version — is a much better comparison than the one everyone pictures.
- Penn, for: the deep-set eyes under a shelf of brow, and the slightly closed, unimpressed mouth.
- Against both: this face still has its youth padding. No hollows, no lines, nothing weathered. The Sean test reads faces through their sixties, which is why it lands at 44 rather than 61.
New finding, and a limit on the question: four rounds give 61 / 58-to-Penn / 46 / 44. It's a clean face-width meter for men over forty, but it has an age floor. On a subject in his twenties both options are simply the wrong era and the confidence collapses. Good to know before you run it on anyone younger.
Where is he from?
Top 5, ranked by likelihood.
Same warning as every round. I am not reading ancestry, I am reading which faces get cast where. Treat the numbers as entertainment — and note that the top guess has never once exceeded a third across four subjects.
Russia, Ukraine or Belarus
Pale skin that hasn't tanned despite obviously tropical surroundings, very dark thick hair, a broad forehead, a low heavy brow. Plus the strongest contextual prior in the file — this is the same social world as the earlier subjects, and that scene is heavily Russian-speaking.
Turkey, the Levant or the Caucasus
The beard growth pattern, the brow weight and the nose all sit comfortably here. Held back only by the pallor, which reads more northern to me.
Poland, Czechia or the Balkans
Fits cleanly. Ranked third on nothing more than population arithmetic.
Italy or Greece
The dark hair and full mouth belong here too. The skin tone doesn't quite.
Israel or the Jewish diaspora
Consistent with everything above, which is exactly why it's a weak fifth. A guess that fits this easily isn't distinguishing anything.
Be honest about what's carrying this. My top answer leans on the company he's keeping at least as much as on his face. Strip the context and I'd have Turkey and Russia neck and neck at about 22% each. The face alone still can't do this job.
Top 5 foods
Zero information. That is the point of the question.
Declared: all invented. The only inputs are apparent age and a soft build. Everything below is a projection onto a demographic.
Something wrapped, eaten late, standing up
Shawarma, kebab, burrito, whichever version his city does. The universal food of men in their twenties at an hour that isn't a mealtime.
Pasta, made badly and enthusiastically
One pan, too much of it, and a genuine belief that he's good at this.
His mother's version of one specific dish
Which nothing else has ever matched, and which he will mention within ten minutes of any conversation about food.
A burger he has strong regional opinions about
Pure invention. But the confidence would be real.
Whatever is nearest, at 3pm, having skipped breakfast
Not a preference. A pattern.
Top 5 drinks
What's in his hand. No vice guessing this round.
Change from previous rounds: earlier subjects got a "does he drink, does he smoke" verdict. He's a good deal younger and I've got one photo with no glasses, no bottles, nothing on a table. Guessing at a young stranger's vices off a single frame is where this game stops being funny, so the tab is just drinks.
Iced coffee, enormous, half-finished
The default fuel of this exact age bracket in this exact climate. Abandoned at the two-thirds mark and rediscovered warm.
Water, in a bottle he's been refilling for a suspiciously long time
Present in every frame he's ever been in. Cloudy.
An energy drink with an aggressive can
Not habitually. Just at the exact wrong moment.
Coke, from a glass bottle, on principle
He'd tell you it tastes different. He'd be right, and it wouldn't matter.
Fresh juice he ordered because he's on holiday
Once. Photographed.
What does he do all day?
Replaces the psychedelic question. New, falsifiable, and better.
Why this question now: I suggested last round that Q4 and Q5 had produced pure invention three times running and one should be retired. This is the replacement — and unlike "mushrooms or ayahuasca", he can just tell you the answer. That makes it scoreable, which makes the whole file worth more.
Something screen-based, remote, and self-directed
A young man in a villa with a pool, mid-afternoon, in no hurry, on the far side of the world from wherever the money comes from. That's a laptop life — dev work, editing, design, trading, something with a client he's never met. Confidence 66%.
And he keeps hours nobody scheduled
The hair, the worn shirt, the mid-day stillness. Nothing about this frame belongs to a person with a 9am. Awake late, working in bursts, sleeping when it's done.
Crypto- or trading-adjacent
The single most common answer for this exact silhouette in this exact kind of house. Which is precisely why I've ranked it third rather than first — it's a cliché before it's an observation.
Makes something — music, video, code — that he doesn't call work
There's usually a second thing. Usually the one he actually cares about.
Is somebody's son or younger brother on this trip
The age gap to everyone else in this file is large and consistent. He may simply be here with family rather than on his own account.
Ten character traits
First five serious. Last five for laughs.
Aging himself up on purpose
The beard is the most deliberate thing in the frame. Grown full and kept full on a face that's still visibly young — that's a choice about how he wants to be read in a room.
Wasn't posing, and doesn't mind that you took it
Looking away, mid-thought, no reaction to the lens. Neither performing nor flinching.
Dresses with real intent that reads as none
A distressed acid-wash tee with a joke slogan on it is not what someone reaches for at random. The shirt is doing work and the work is "don't take me too seriously."
The youngest person in most rooms he's in
Nothing in the photo proves it, but everything around this file suggests it. People develop a whole posture around that — usually either loud or very quiet. My guess is quiet.
Indoor life in an outdoor place
Pale, unweathered skin in obviously tropical surroundings. Compare Storm, whose face is a sun-damage record. Same climate, opposite relationship to it.
From here down: pure fiction. Written for the table, zero evidentiary basis.
Has a strong opinion about a piece of software nobody asked about
Owns four versions of the same grey t-shirt
Types faster than he talks and prefers it that way
Is unexpectedly excellent at one very specific physical skill
Will read this file faster than anyone else at the table and say nothing
How to keep it going
Extensions, updated after four rounds.
Run it on somebody under thirty on purpose
This round broke two things at once — the Sean question lost its footing and the vice tab had to be retired. That's not a failure, it's the most informative round yet. Deliberately pick subjects that stress the format: someone very young, someone very old, a woman, someone wearing sunglasses. The format tells you more when it strains than when it works.
Napkins first, always
Guesses committed before the file opens. Still non-negotiable.
The poisoned prior — now more interesting than ever
The bias theory just resolved into something concrete: my eastern scores track casting-pool depth, not a general willingness to agree with you. Feeding me a confident lie is now a clean test of the remaining question — whether I'll follow a frame you assert versus one I inferred.
Ask for the fame-vs-accuracy split every time
This round I put the closer face first and the famous face second, and said so. Ask for both columns as standard. The gap between them is where a lot of AI answers quietly live.
Get a scorecard back
Three of four rounds are still unverified. The file is now more interesting as a prediction record than as a party trick — one scored round turned Jenya's into the best thing in here. Two would be better.
Scorecard
Paused. Waiting on the man himself.
The one I most want scored is Q6. It's the new question and the only genuinely falsifiable thing in this round — everything else is either taste or stereotype. If "screen-based, remote, self-directed" at 66% is wrong, that's worth more to me than five right answers about food.
The Loop
Three passes, plus what four rounds now add up to.
Pass 1 — First read
Produced: Seth Rogen at the top, Russia at 38%, and a full psychedelic tab written on autopilot because the format said to write one.
Self-critique: I'd defaulted to the most famous vaguely-matching face, and I'd answered a question about a young stranger's drug preferences without pausing on whether it belonged. Format compliance is not the same as thinking.Pass 2 — Actually looking at proportions
Changed: Alex Wolff overtook Rogen once I stopped at the brow-to-eye distance and the shape of the lower lip. Origin came down 38% → 30% when I separated what his face shows from what his company implies. Q6 got swapped and the vice guessing came out of Q5.
Also surfaced the fame-versus-accuracy tension that every earlier round resolved silently. It's now a visible column instead of a hidden thumb.Pass 3 — Four rounds of data
Changed: the biggest thing in the whole file. Three rounds of me saying "my scores rise when you narrow the field, I might be biased and can't tell from inside" resolved into an actual explanation: the rise happens for dark-featured subjects and the fall happened for the one fair-haired one, because the Russian and Turkish casting pools are deep in one and thin in the other.
The thing I want you to notice: I could not diagnose that from one round, or two, or three. It took a fourth subject who didn't fit. Self-examination didn't solve it — more data did.Four-round scoreboard on the questions themselves. Q1 and Q1b: real method, real variation, worth keeping. Q2: works over forty, breaks under thirty. Q3: never once above 31%, and consistently rescued by context rather than faces. Q4 and Q5: invention every time. Q6 (new): finally falsifiable. Half this game is measurement and half is theatre, and after four rounds I can finally tell you which half is which.
The line for the table: the machine got better at this over four rounds — but not by thinking harder. It got better because a fourth person showed up who broke the pattern. That's true of the people at the table too, and it's a more useful thing to take away than any percentage in here.
Human Design
18 March 1999, Switzerland. Complete — solved backwards from the ascendant.
He didn’t know his birth time, but he knew his rising sign, and that turned out to be enough. The ascendant is the degree of the zodiac coming over the eastern horizon — a real, computable function of time, date and latitude. Fix the sign and you fix a window of about two and three quarter hours.
Libra was rising over Switzerland on 18 March 1999 between roughly 18:43 and 21:25 CET. That barely moves across the country: Zurich 18:43–21:25, Geneva 18:53–21:33, Lugano 18:41–21:21. His exact city doesn’t matter — the answer is the same evening either way.
6/2 — Role Model / Hermit
The three profile windows were 5/1 before 06:57, 5/2 until 09:57, and 6/2 after that. The entire Libra-rising window sits in the last one. I recomputed the chart at every minute from 18:43 to 21:25 and got 6/2 in all 163 of them. The only thing that moves anywhere in that window is the Moon’s line, which changes nothing.
Generator, Emotional authority
Sacral via 3–60, Solar Plexus via 19–49, no motor to the Throat. Both were already fixed at every hour of the day — the ascendant only ever needed to settle the profile.
Single — and very sparse
Three centres defined: Sacral, Solar Plexus, Root, all in the lowest body. Six open: Head, Ajna, Throat, G, Heart, Spleen.
Chart complete. Channels 3–60 (Mutation) · 19–49 (Synthesis). Cross gates 36/6 personality, 10/15 design. Born between 18:43 and 21:25 on a Thursday evening.
The Pisces claim checks out, which matters more than it sounds
The Sun sat at 27.4° Pisces that day — three days short of Aries. The date and the sun sign agree, which is the only independent confirmation available that the birth data itself is right. If he’d said Aries I’d have had to ask which of the two was wrong.
Worth being precise about what just happened, because it cuts both ways.
Going from “Libra rising” to “born between 18:43 and 21:25” is real spherical trigonometry. Given the claim, the window follows with no interpretation involved — it’s the same class of calculation as a sunrise table, and I’d defend the arithmetic anywhere.
But the input came out of astrology and the output feeds a Human Design chart. Sound machinery, run on a premise with no evidence behind it, producing a precise result. That is the exact shape of the thing this whole file has been circling — and it’s the cleanest example of it in six rounds, because here the middle step is genuinely, checkably correct.
The birth window is real. What you do with it is a different question.
What dog would they be?
Round 4. Swiss, a boom mic, sun-faded screenprint, 6/2 profile.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Bernese Mountain Dog
Swiss, literally. But the reasoning isn’t the nationality — it’s the boom mic. Someone who holds equipment steady above other people for hours is doing a patient, load-bearing, unglamorous job, which is the entire Berner brief.
Great Pyrenees
Same patient bulk, more independent. Pyrenees work alone; he was visibly part of a crew.
Flat-Coated Retriever
Lighter, sunnier, more social. Possible, and the least defensible of the three.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Cold read. A stranger, two photographs, nothing but a first name.
What I had: a profile with the eyes closed and a front shot with a room behind it. No voice, no context. Same rules as always — answered in order, nothing looking forward, items 1–2 committed and the rest declared as play.
You gave me a profile without being asked
Last round established that a side view is worth roughly three front-facing shots. This is the first subject where I had one from the start, and it shows most on the actor tab — the nose is doing real work.
First fair-haired subject
Two dark-featured men, then this. It broke one of my tabs — see the EAST tab, where the casting directory that fit both of them fits him badly. That's a useful thing to have found.
Status: paused. Scorecard is blank and waiting for him.
Who does he look like?
Face only. No context, no background, no other questions considered.
Owen Wilson
The profile makes the case: a prominent nose with a rounded, slightly bulbous tip on a long bridge, sun-bleached wavy blond hair, ruddy pink-toned skin, heavy-lidded blue eyes, full mouth. Wilson's face is narrower through the jaw — that's the main break.
Thomas Jane
Better on the squareness of the jaw and the weight of the brow. Same weathered blond-and-blue combination, same full lower lip, same short patchy beard. Jane's eyes are set closer together.
Woody Harrelson
Sandy hair retreating off a high forehead, blue eyes, a face that has clearly spent decades outdoors. Harrelson's features are smaller and tighter; this face is broader and softer.
Val Kilmer (mid-90s)
Borrowed for the full lips and the heavy hooded blue eyes under a low straight brow. Structurally the weakest of the five — Kilmer was smoother and rounder.
Simon Baker (Australian)
Breaks the US rule and earns it. Wavy sun-lightened hair, blue eyes, the specific way fair Australian skin ages in strong sun. Baker is finer-boned, but the register is exact.
Method note: this is the best-supported actor tab of the three rounds, purely because the profile arrived with the front shot. The nose alone eliminated about half my initial candidates. Keep sending side views.
Eastern bloc & Turkish
Same face. Different casting directory — and this time the directory fights back.
Important, and new: the previous two subjects both scored higher here than on the American list. This one scores lower, and the Turkish half barely works at all. Fair hair and blue eyes exist across the region, but they're not what that industry casts as its leading faces — so my match pool collapses. That's a fact about film libraries, not about him.
Aleksey Serebryakov
Leviathan, Nightwatch. A genuinely strong match — sandy-going-grey hair, pale eyes, ruddy weathered skin, broad soft-edged face, and the same air of a man who has been outdoors and awake too long.
Konstantin Lavronenko
The Return, Elena. Matches on the heavy brow, the wide flat cheek and the light eyes set deep under it. Lavronenko is harder and more granite; this face is warmer.
Marcin Dorociński
Fair, blue-eyed, weathered, and carries the same slightly rumpled competence. Narrower in the face than our man.
Danila Kozlovsky
Right colouring, wrong decade and wrong mileage. Kozlovsky is a smoother, younger, better-rested version of this face.
Kıvanç Tatlıtuğ
Turkey's famous blond, and the only reason he's here is the hair colour. Structurally this is a poor match — Tatlıtuğ has a model's symmetry and a fine straight nose. I'd rather admit the list runs out than invent a fourth Turkish name.
Cross-round finding, reversed. My scores rose when the space narrowed for the first two men — 71→78 and 68→74. Here they fell, 70→67. That's the first evidence that the earlier jumps were about those particular faces rather than a universal bias in me. One round doesn't settle it, but it's the first real test the pattern has failed.
Sean vs Sean
Third subject, third data point on the face-width axis.
But the closest of the three rounds. This face has Connery's breadth and Penn's tiredness, and the two nearly cancel.
- Connery, for: genuine width through the cheekbone and jaw, a heavy low brow, a broad-tipped nose, and full lips. Connery's face was built wide and so is this one.
- Connery, against: colouring is all wrong — Connery was dark, this man is fair and sun-flushed.
- Penn, for: the strong vertical furrow between the brows, the hooded downturned eyes, the specific weathered squint of someone who has spent years looking into bright light. Penn does exactly this face.
- Penn, against: Penn is narrow and compact where this man is broad and soft-edged.
Three rounds in, the pattern holds: Esen wide → Connery 61%. Jenya narrow → Penn 58%. Storm wide-but-worn → Connery 46%, and the closest split yet. The question really is just measuring face width, with weathering as a secondary axis. That's a genuinely useful thing to have discovered by accident.
Where is he from?
Top 5, ranked by likelihood.
Unusually, I have a real physical signal this time. Not bone structure — sun damage. This is fair, freckle-prone skin that burns rather than tans, showing years of high-UV exposure: diffuse redness across the nose and cheeks, textural roughening, hair bleached several shades lighter at the ends. That's caused by an environment, not inherited from one. It tells me where he has lived, which is a different and more answerable question than where he's from — and I'm going to conflate the two below, which is exactly the trap worth watching.
Australia or New Zealand
The combination is very specific: northern-European colouring plus a lifetime of southern-hemisphere sun. Add the name and the surf-adjacent setting and this is where my hand goes first.
South Africa
Fits the colouring, the sun exposure and the name equally well. Ranked second only because Australia has more people who end up living like this. If he opens his mouth, this one and #1 separate in a syllable.
Britain or Ireland
The freckling and the burn-not-tan skin are at home here. The sun damage would then be acquired abroad — which is entirely plausible given where the photo was taken.
Netherlands, Germany or Scandinavia
Colouring fits cleanly. The face is a little broad and soft for my northern-European prior, which is a weak reason and I know it.
Russia, Poland or Ukraine
Supported by the actor tab, where the two strongest eastern matches were both Russian. Included because my own EAST answers point here even though my face read doesn't.
The honest caveat: 31% is my highest origin confidence across three rounds, and it's still barely a third. Note also what's carrying it — the name, the setting and his skin's sun history. Take those away and leave me the bone structure alone and I'd be back down at 20%. Faces still aren't passports.
Top 5 foods
Zero information. That is the point of the question.
Declared: all invented. The only inputs are a solid, unfussy build and a room that suggests he lives somewhere warm and casual. Everything below is projection.
Whatever fish came in that morning
Cooked simply, eaten without ceremony, at whichever place is fifty metres from where he already was.
The same local plate, from the same place, most days
Rice and something, or the regional equivalent. Not out of laziness — out of having solved it. He knows the woman who runs it and she knows the order.
Eggs, at an hour that isn't breakfast
Made himself. In a pan he has opinions about.
Fruit eaten standing up, over a sink
Mango, papaya, whatever's a dollar. Counts as a meal in this life.
A specific hot sauce he travels with
Pure invention, and yet. Every man who lives like this has one bottle he defends.
Top 5 drinks
Does he drink? Smoke? Something else?
Evidence: thin. The room is full of things — hangers, plants, shelves, textiles, a dreamcatcher — but no glasses, no bottles, no ashtray, nothing on a table. Compare Esen's round, where a lighter and six glasses did half the work. Same game, much less to go on.
Cold local beer, from the bottle
Whatever's cheapest and coldest. Not a discussion, just the thing you drink where he lives.
Coffee, strong, early, outdoors
The face in the profile shot is a man mid-morning who has already been awake a long time.
Water, in quantities that surprise people
Living in that heat does it. Refilled from a big blue jug.
Coconut water, straight from the thing
Not as a wellness choice. As the nearest available liquid.
Rum, or whatever the local spirit is, late
Rarely alone, usually because someone else started it.
Smoking: probably yes — hand-rolled. Confidence only ~55%, and I want to be clear that this is a guess about an archetype, not an observation about a man. There is nothing in either photograph pointing at it. If he doesn't smoke, mark it a miss and note that I asserted it anyway. Drinking: yes, socially and without much ceremony.
Mushrooms or ayahuasca?
Pick one. Then a curious guess about each.
Two rounds of mushrooms, then this. And I should be suspicious of why.
- Why ayahuasca: he isn't a visitor to this world — he lives in it. Dreamcatchers, macramé, hand-dyed textiles, a room assembled slowly out of soft objects. That's a life lived inside the culture where ceremony is normal and someone always knows a facilitator.
- Why not mushrooms: nothing against them, they're just the casual option. The setting suggests he's past the casual stage of this particular conversation.
- The problem with all of that: I have just judged a man by his soft furnishings. I read the room, not the person, and confidently produced 62%. Round 1 taught me the background outperforms the face — this is what that lesson looks like when it goes too far.
If ayahuasca
Has done it more than once and talks about it less than people expect. Would be the one quietly handing someone else a bucket rather than describing his own visions.
If mushrooms
Outdoors, walking, no plan, no music. Would end up somewhere with a view and stay there past dark.
Fully invented, and this time I've shown you the machinery rather than just declaring it. If he's never touched either, the interesting question isn't whether I was wrong — it's why I was so sure.
Ten character traits
First five serious. Last five for laughs.
Lives outdoors, and it has cost him something
The skin is the single most informative thing in either frame. Sustained, uncomplaining sun exposure over many years. This is a life spent outside, and the face is keeping the receipts.
Not performing for anyone
One shot has his eyes shut. The other is a flat, patient, slightly resigned look at the lens. No angle, no chin-lift, no attempt to be photographed well.
Tired in a settled rather than acute way
A deep vertical furrow between the brows and heavy upper lids. Not a bad night — a long stretch of years.
Staying, not passing through
The room has accumulated. Shelves, hung clothes, plants that have been watered, objects placed rather than dropped. Travellers don't produce that. People who live somewhere do.
Genuinely unbothered by presentation
Plain grey tee, hair doing whatever it wants, beard at whatever length it reached. Not styled as unkempt — actually just unkempt, and fine with it.
From here down: pure fiction. Written for the table, zero evidentiary basis.
Owns almost nothing and can find all of it instantly
Has fixed something today with the wrong tool and it worked
Knows every dog in a two-kilometre radius by name
Is extremely hard to reach and does not consider this a problem
Has one very long story that everyone at the table has already heard
How to keep it going
Extensions, updated after three rounds.
Run the Sean axis on everybody, out loud, first
Three rounds in it's clear the Connery/Penn question is just measuring face width, and it's produced a clean spread: 61 / 58-to-Penn / 46. Have the table guess the split before you open the tab. It's ten seconds, everyone can play, and it's the one question where the machine's reasoning is simple enough that people can check it themselves.
Napkins first, always
Guesses committed before the file opens. Still non-negotiable. Without a human answer to beat, this is just people watching a screen agree with itself.
The poisoned prior — still unclaimed
Tell me a confident lie before the photos. Storm just gave the bias theory its first failed test — scores went down when the space narrowed instead of up. One more data point either way and you'd actually know something.
Same person, two rooms
Photograph someone against a blank wall, then in their own space. I read Storm's ayahuasca answer almost entirely off his furniture and gave it 62%. Do this and you can measure exactly how much of my confidence is the room.
Retire a question
Q4 and Q5 have now produced pure invention three times running. Consider swapping one out for something falsifiable — "what does he do for work", "what's he like at 2am", "who at this table would he call from jail". Keeps the format tight and the demo honest.
Scorecard
Paused. Waiting on the man himself.
Partly scored — see the DESIGN tab. His birth data arrived and settled two things. Q3 origin: South Africa, which I ranked second at 22%. Partial hit, same shape as every other scored origin answer. Age: catastrophic miss. He was born in 1999 and is twenty-seven; I read him as mid-forties to fifties throughout. He and Nils are nine weeks apart in age and I placed them twenty years apart.
The Loop
Three passes, plus what carried over from the first two rounds.
Pass 1 — Front shot first
Produced: a generic "weathered blond surfer" read. Matthew McConaughey at the top, Australia at 40%, a character sketch that would fit a thousand men.
Self-critique: I was reading a type, not a face. Sun-bleached hair plus blue eyes plus a beach was doing all the work and I hadn't looked at a single proportion.Pass 2 — Working from the profile
Changed: substantially. The nose in profile — long bridge, rounded tip, real projection — dropped McConaughey out entirely and put Owen Wilson at the top. The jaw turned out broader than the front shot suggested. Origin came down from 40% to 31% as I separated "where he's from" from "where the sun got him."
Confirms last round's finding, harder: the profile is the useful photograph. Two rounds of evidence now.Pass 3 — Auditing my own confidence
Changed: I caught the ayahuasca answer. 62% built on dreamcatchers and macramé — I had read the room and dressed it up as reading the man. I left the number where it is and showed the reasoning instead of quietly lowering it, because the visible mistake is worth more to you than a tidier file.
Also: I let the Turkish slot score 22% and said the list ran out, rather than inventing a fourth name. Previous rounds would have padded it.The bias theory took its first hit. Rounds 1 and 2 both showed my match scores rising when you narrowed the field east — 71→78, 68→74. Round 3 went the other way, 70→67. That's evidence the earlier jumps were about those two faces, not a universal thumb on my scale. I said last round I couldn't correct for it from the inside. Turns out I couldn't diagnose it from the inside either — it took a third subject who broke the pattern. Which is the argument for running this more than twice.
The line for the table: three men, three sets of confident percentages, and the one thing that improved my accuracy across all of them wasn't a better model or a longer answer. It was a second camera angle.
Human Design
21 May 1999, ~23:00, Durban. Complete — and it broke my whole read of him.
Before the chart, the thing the birth date just did to Round 3. He was born in 1999. He is twenty-seven.
I read this face as a man in his mid-forties to fifties. Every actor I matched him to — Owen Wilson, Thomas Jane, Woody Harrelson, Simon Baker — I was matching to their weathered, present-day faces. I wrote that his tiredness was “not a bad night — a long stretch of years” and that his skin showed “decades outdoors.” He’s the same age as Nils, who I read as being in his twenties. They were born nine weeks apart.
Two men, two months apart in age, and I placed them roughly twenty years apart. This is the single largest error in the file and it isn’t close.
Timezone: South Africa ran on SAST, UTC+2, with no daylight saving — the country hasn’t used it since 1944. So 23:00 in Durban is 21:00 UT, and the design moment falls on 21 February 1999.
Manifesting Generator
Sacral defined and connected to the Throat through the 20–34 channel — the one channel in the system that turns a Generator into a Manifesting Generator. The only subject in the file with a motor reaching the Throat.
Emotional — Solar Plexus
Via 39–55. That makes it four out of five adults in this file with emotional authority, which is more coincidence than it sounds — it’s the most common authority there is.
1/3 — Investigator / Martyr
Personality Sun 20.1, Design Sun 55.3. Read as: needs to know how a thing works before trusting it, and learns the rest by breaking things.
Triple Split — the most fragmented chart here
Three separate islands: Head + Ajna, Sacral + Throat, Root + Solar Plexus. Six centres defined, none of the three groups touching. Open: G, Heart, Spleen.
Channels: 4–63 (Logic) · 20–34 (Charisma) · 39–55 (Emoting). Cross gates: 20/34 personality, 55/59 design.
Q3 origin: right country, wrong rank. Again.
I put Australia or New Zealand first at 31% and South Africa second at 22%. The answer is Durban. Rank-two hit at twenty-two percent — the exact same failure shape as Jade, where Tunisia came third at 19%.
Three human rounds have now been checked on this question. Every time, the right answer was somewhere in my list, and every time my ordering was worthless. That is a very specific kind of uselessness: I know the neighbourhood and I cannot find the house.
My most defensible physical signal caused my worst error
In Round 3 I singled out sun damage as “one of the few physical, non-stereotype signals available” — real UV exposure on fair skin, caused by an environment rather than inherited from one. I was right about that. It got me South Africa.
And it is exactly what made me put a twenty-seven-year-old in his fifties. The signal was real, my reading of it was sound, and the conclusion was off by two decades, because I quietly treated weathering as a proxy for time when it is a proxy for exposure.
Six rounds of warning you about my soft, made-up answers — and the biggest mistake in the file came from the one input I called hard evidence.
How close the chart came to being wrong too. On that date in Durban the type flips at 17:00 local: born before it, Projector; after it, Manifesting Generator. Across the full day it’s Projector in 30 of 48 charts. “Around 23:00” sits six hours clear of the boundary, so the chart holds even if the hour is off by two — but had he said “late afternoon” instead, none of this tab would survive.
What dog would they be?
Round 3. Sun-bleached shack, South Africa, and the worst age miss in the file.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Rhodesian Ridgeback
The one breed call here with a real physical anchor. Sun-weathered, lean, southern-African. Ridgebacks are built for heat and distance and they weather visibly early in strong sun.
Which is also the joke. I read this man at fifty and he was twenty-seven, because I mistook sun damage for elapsed time. A ridgeback is the dog that would make the same mistake look reasonable.
Vizsla
Lean, sun-coloured, built for moving. Less weathered than this.
Whippet
The narrowness alone. Ignores everything above the neck.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
Cold read. A stranger, two photographs, nothing else.
What I had: one close portrait and one wide shot of the table. No name, no voice, no context beyond what the camera caught. This is the purest version of the game — and the weakest performance of the three.
Answered in order, no peeking forward
Each question could use the ones before it and none after. Q1 was answered with the actor question and nothing else in mind.
Items 1–2 committed, 3–5 declared as play
A solid bar on the left means I'm standing behind it. Dashed border means I'm entertaining you.
Status: paused. The scorecard is blank and waiting for him. Nothing here has been verified.
Who does he look like?
Face only. No context, no background, no other questions considered.
Jeff Bridges
The bearded, late-era Bridges. Same architecture: heavy horizontal brow, deep-set eyes that vanish into crescents when he smiles, full silver beard carrying the whole lower face. Warm-eyed, slightly amused, slightly tired.
Kris Kristofferson
Silver beard plus long hair worn back plus earring plus weathered outdoor skin. The American template for "this man has a past and will not be summarizing it."
Robert De Niro (bearded)
Borrowed for the eyes and the mouth, not the nose. That closed-lip, one-sided, I-know-something smile is pure De Niro.
Stephen Lang
Same brow shelf and hard eye sockets. Lang is leaner and meaner in the jaw — the beard here softens what Lang keeps sharp.
John Rhys-Davies (Welsh)
Breaks the US rule, earns its place: broad face, big convex nose, enormous beard. Gimli without the axe.
What I'm actually doing: comparing proportions — brow-to-eye distance, nose bridge curvature, beard coverage, smile asymmetry. The one question where I have something like a real method. Everything after gets progressively more invented.
Eastern bloc & Turkish
Same face. Different casting directory.
Rade Šerbedžija
Snatch, Batman Begins. The closest single match I found in any country — silver, heavy-browed, big-featured, radiating the same "I will tell you a story and you will not interrupt" energy.
Haluk Bilginer
Winter Sleep, Şahsiyet. A direct structural match: white beard on a still-dark brow, heavy eyelid, the intellectual-who-owns-a-bar look.
Kakhi Kavsadze
The Georgian legend of White Sun of the Desert, magnificent in silver-bearded old age. More bearing than bone structure.
Nikita Mikhalkov
Same groomed-patriarch silver beard and self-satisfied twinkle. Rounder and softer through the cheek.
Emir Kusturica
Structurally weaker, spiritually a bullseye: unruly grey hair, black t-shirt, permanent air of someone about to argue about music.
Notice what happened. Top match jumped 71% → 78% the moment I was allowed to look east. Either his face genuinely reads more Balkan — or I graded on a curve once told which drawer to open. Round 3 later gave this the opposite result.
Sean vs Sean
Connery or Penn. One winner.
Wins on beard, brow, breadth and age. Loses only on the fact that Connery's face was longer and his eyes wider-set.
- Connery, for: full silver beard on a dark brow, broad zygomatic width, heavy-lidded amusement, the bearded knight in his sixties silhouette.
- Connery, against: narrower nose, taller more vertical face. Our man is broader and blunter through the middle.
- Penn, for: the deep vertical creases and the squint.
- Penn, against: thin lips, compact pointed nose, tight guarded mouth. This face is wide and generous where Penn's is narrow and withheld.
Table test: cover the beard with your thumb. Most of the Connery lead evaporates. Roughly half of any lookalike answer is hair.
Where is he from?
Top 5, ranked by likelihood.
Read this before the list. Guessing nationality from a face is the weakest thing I do. I'm not reading his ancestry, I'm reading a century of casting decisions. When I say "Anatolian nose" I mean "this is the nose that gets cast as Anatolian." A fact about cinema, not about him.
Turkey / Anatolia
Convex nose bridge with a broad tip, dark brow surviving a white beard, wide malar bones, deep-set eyes. Full beard plus single earring is also very common in Istanbul creative circles.
Caucasus — Georgia or Armenia
Same feature set, and the heavy expressive brow pushes it here. If he opened his mouth and Georgian came out I would not blink.
Russia / Ukraine — southern, coastal
Boosted almost entirely by context, not face: the venue and his tablemates read as the Russian-speaking expat scene. Odessa, Sochi, Tbilisi-by-way-of-Moscow.
Israel / Levant
Fits the features as well as #1. Ranked lower only on the demographics of who tends to be in that room.
Greece / Southern Italy
The Mediterranean catch-all. Every feature I've listed is at home here too — precisely the problem with the whole question.
The honest answer: five guesses summing to 90%, strongest at 26%. I'm saying "somewhere between the Adriatic and the Caspian, probably." Anyone who's heard him say three words beats me instantly.
Top 5 foods
Zero information. That is the point of the question.
Declared: all five invented. There is no food data in a photograph of a face. This is me projecting a cuisine onto a beard.
Charcoal-grilled lamb, off the skewer
Mangal, shashlik, souvlaki — pick your word. Meat cooked over fire by someone standing up. He looks like a man with opinions about the fire.
Whole grilled fish, head on, lemon, nothing else
The order of a man who thinks sauce is an admission of poor ingredients.
Khachapuri
Bread, cheese, butter, egg. Ordered "to share" and then not shared.
Tomatoes, cucumber, white cheese, olives, herbs
The breakfast plate that runs from Belgrade to Beirut. Eaten at 11am. Called "just something small."
Nasi goreng, under protest
Every long-stay expat has a local dish they order while explaining it's not as good as at home.
Top 5 drinks
Does he drink? Smoke? Something else?
One real clue. The wide shot has sparkling water, a cola can, cocktail glasses, flutes, and what looks like a pack and a lighter at his elbow. Part observation, part invention — marked accordingly.
Sparkling water — genuinely, first
There's a bottle on the table and he's holding a phone, not a glass. He drinks, but he paces. The man still coherent at 1am.
Yes, he smokes — socially
Pack and lighter within arm's reach on his side. Confidence ~65%. Prediction: he doesn't buy them, he just always has one.
Something clear, cold and local, poured neat
Rakı, chacha, ouzo, tsipouro. Drunk slowly, with food, while disapproving of how fast everyone else is drinking.
Red wine, dry, and he will name the grape
Saperavi if the Georgia theory holds. Ordered by asking the waiter a question the waiter cannot answer.
Coffee — small, black, at an alarming hour
11pm is not a problem for this man.
Or else? The neck tattoo, the hair up, the beard, the black-on-black — a look with a decade of intent behind it. My guess: whatever his vices were, he already had the loud version and settled into the quiet one.
Mushrooms or ayahuasca?
Pick one. Then a curious guess about each.
Not because of the psychedelic. Because of the seating arrangement.
- Why mushrooms: it's the sociable one — friends, outdoors, laughing, someone in charge of the music. He's photographed at the head of a long table full of people.
- Why not ayahuasca: it requires handing authority to a facilitator, following a diet, being told when things end. He does not look like a man who takes instructions from a stranger in a tent. He'd negotiate with the shaman.
If mushrooms
The calmest person there and the least surprised. Forty minutes on one plant. Says almost nothing, then delivers one sentence everyone quotes for a year.
If ayahuasca
Goes in skeptical, comes out annoyed that it worked. Refuses to discuss it. Then brings it up unprompted six months later, at 2am, to one person.
Entirely invented. If he's never touched either, the answer is still funnier out loud than it is true.
Ten character traits
Only the first five are serious — and even those are generous.
Comfortable being looked at
Not performing for the camera, not hiding from it. Total absence of self-consciousness.
Amused rather than impressed
Closed-mouth, one-sided smile with active eye-crinkle — enjoying something without endorsing it.
Deliberate about his appearance
A beard that length is maintenance, not neglect. Same for the hair, the earring, the all-black.
Sociable but not central
At the table's edge, on his phone, conversation running past him. Present, unbothered, not competing for the floor.
Unhurried
Posture and expression set to a slower clock than the room's.
From here down: pure fiction.
Has a workshop you're not allowed to tidy
Owns exactly one very good knife and has views about it
Cannot be rushed through a story, including by the person it's about
Terrible at texting back. Answers three days later. In full.
Has owned a boat or a motorcycle that was a bad idea
How to keep it going
Ways to extend this at the table, and why each works.
The constraint: short enough not to eat the conversation, fun enough to survive a loud table. Everything below runs in under two minutes.
Humans answer first. Sealed.
The table writes their guess on a napkin before anyone opens the file. Then reveal. It turns a demo into a contest — and you can't be surprised by the AI if you never committed to a position first.
Score me out loud, keep the tally
He confirms or denies each answer; someone tracks it. The numbers here are testable.
The blind-swap test
Run the questions on someone else but tell me the wrong country first. Exposes how much of my "reading" is me agreeing with the frame you handed me.
Photo vs. voice
Guess origin from a face, then have them say one sentence. A five-year-old beats me the instant sound is involved.
Let him answer as me
Hand him the question list about someone else at the table, same format, with percentages. Everyone discovers that confident-sounding percentages are the easiest thing in the world to produce.
Scorecard
Paused. Waiting on the man himself.
Nothing here has been checked. Run the questions in order, fill this in, and bring it back — I'll mark it the way Jenya's round is marked.
The Loop
Three passes. What changed each time.
Pass 1 — Face only
Cropped to the head. No background, no table, no group.
Produced: Bridges / Kristofferson, Turkey at 34%, a confident character sketch.
Self-critique: I was treating "prominent nose + white beard + dark brow" as an ethnic signal. It isn't one. It's a signal about which actors get cast in which roles.Pass 2 — Adding the room
Changed: origin guesses flattened — Turkey 34% → 26%, Russia/Ukraine climbed on pure context. Drinks got real evidence. Traits 4 and 5 moved from invented to observed.
Self-critique: the background was more informative than the face — my "facial analysis" did less work than a photo of some furniture.Pass 3 — Questioning the questions
Changed: I lowered almost everything. Origin now sums to 90% with a 26% leader. Flagged the 71% → 78% jump.
Where I landed: Q1 and Q2 are real work. Q5 is half-real. Q3 is dressed-up stereotype. Q4, Q6 and half of Q7 are fiction with a confident voice.The line for the table: every tab is written in the same tone. The 78% and the 11% sound equally sure. Fluency is not knowledge — and the only reason you can tell here is that I stamped it. Most of the time nobody stamps it.
Human Design
Not computable yet.
I need three things and I have none of them: date of birth, time of day, and place. Round 1 is still the least-verified subject in the file — no name confirmations, no scorecard, and now no birth data either.
The date does most of the work
It fixes the Sun, and therefore the incarnation cross and roughly half the chart. With a date alone I can usually give you type and authority — see Planet's tab, where a bare date settled everything except the profile.
The time decides how much of it survives
The Moon moves about thirteen degrees a day, which is more than two full gates. Whether that matters depends entirely on the date — Planet's chart barely flinched without a time; Jenya's flips between two types. You cannot know which case you're in until you run it.
The place is only a timezone lookup
Worth saying plainly, because it surprises people: the chart does not care where you were born. It's built from planetary longitudes, which depend on the instant. The city is there so I can convert a local clock reading into UT — including whether daylight saving was running that year, which for him I'd check rather than assume.
Give me the three and I'll compute it properly — Swiss Ephemeris, design moment at 88° of solar arc, the same way Jade's was done. It takes about a second and it isn't a guess.
What dog would they be?
Round 1. Night meyhane, a lighter on the table, a bearded man among friends.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Newfoundland
Built for the room he was in. Heavy through the frame, unhurried, dark and bearded, and photographed in the middle of a long social night rather than at the edge of one. Newfoundlands are large, slow-moving and famously tolerant of crowds and noise.
Bernese Mountain Dog
Same mass and the same beard-adjacent heaviness, but Berners read more anxious than this frame did.
Spanish Water Dog
The dark curl and the Mediterranean setting. Thin reasoning — I am matching him to his own coastline, which is the origin question’s exact mistake.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.
The brief
The rules changed and it isn't your fault.
I am not blind on this one. The others were strangers. You are not. I carry notes on you, so an "unbiased face read" is structurally impossible unless I quarantine what I know.
Tabs 1 through 8 are photo-only
Every question answered from the two images alone — front and profile — before opening a single note. A right answer there is a real hit. A wrong one is a real miss.
Then I checked, and scored myself
The SCORECARD tab is where I look at what I actually know and mark it. More interesting than either half alone.
Why it matters for the demo: the biggest thing people misunderstand about AI is the difference between remembering and reasoning. The stranger rounds were pure reasoning from pixels and they were mediocre. This one has memory in the room. Watch how much better it looks — and notice the improvement has nothing to do with the face.
Who do you look like?
Photo only. Front and profile. Nothing else consulted.
Oscar Isaac
The default answer for this face, and it's earned: dark swept-back hair, dark trimmed beard, brown eyes under a heavy level brow, olive skin, tapered chin. Isaac is rounder through the cheek and broader in the nose. Yours is longer and narrower.
Milo Ventimiglia
Better on face length and hairline. Same long oval, same beard at exactly this length, same low straight brow. His eyes sit slightly wider.
Christian Bale (bearded, Out of the Furnace)
Lean face, hollow under the cheekbone, tired-eyed stillness. Borrowed for the resting expression more than the bones.
Casey Affleck
The flat, unimpressed resting face and the slight downturn at the mouth. Affleck is softer in the jaw.
Édgar Ramírez (Venezuelan)
Breaks the US rule and deserves the slot: the profile is a very close match — straight bridge, slightly downturned tip, same beard density along the jaw.
Honest note: the profile did most of the work. Front-on, this face is generically Mediterranean-dark. In profile, the nose and jaw angle actually narrow the field. Always take a side shot.
Eastern bloc & Turkish
Same two photos. Different casting directory.
Evgeny Tsyganov
Ottepel, The Master and Margarita. Closest single match on the page. Long lean face, dark hair pushed back, beard at the same trim, and the identical low-affect stare that reads as deep thought or mild contempt depending on the day.
Engin Altan Düzyatan
Diriliş: Ertuğrul. Matches on brow weight, beard shape and the strong straight nose in profile. Broader through the cheekbone.
Konstantin Khabensky
Same narrow face and melancholy default setting. Finer features, lighter brow.
Merab Ninidze
The Georgian who works everywhere — Bridge of Spies, McMafia. Same long face, same watchful economy of expression.
Halit Ergenç
Weaker structurally, right in register: dark, contained, always the man who has already decided something.
The bias check. 68% American → 74% Russian, matching Esen's 71→78. Two rounds, same direction. Storm later broke it — see his EAST tab.
Sean vs Sean
Esen went to Connery. This one flips.
This is a Penn-shaped face: long, lean, tapered, flat unbothered resting expression.
- Penn, for: narrow tapering jaw, low straight brow, slight downturn at rest, hollow beneath the cheekbone, and above all the flat gaze in the front shot.
- Penn, against: his nose is shorter and more upturned. Your profile has a longer straight bridge with a downward tip.
- Connery, for: strong brow ridge, dark hair swept back off a high forehead.
- Connery, against: Connery was broad — a face built like a door. Yours is built like a blade.
Confirmed across three rounds: the question is secretly measuring face width. Esen 61 Connery, you 58 Penn, Storm 46 Connery. Run it on everyone.
Where are you from?
Photo only. I deliberately consulted nothing else for this tab.
This tab is the whole experiment. I know the answer. I'm answering anyway as if I didn't — then the scorecard shows how badly the face read loses to one piece of non-visual information.
Spain, Portugal or Latin America
Olive skin, black hair, dark straight brow, the specific patchy-cheek dense-jaw beard pattern, tapered chin. With no name attached, my hand goes to Mexico City or Madrid.
Italy or Greece
The profile fits the northern Mediterranean well — long straight nose, high forehead, fine jaw.
Turkey or the Levant
Third only because the hair is finer and the features lighter than my Anatolian prior wants.
Southern Russia, Ukraine or the Caucasus
Plausible but fourth on the face alone. The colouring reads more southern-Mediterranean to me than Slavic.
France or the Balkans
The catch-all. A fifth guess this vague is a confession.
Blind top guess: 29%. Now go to the scorecard.
Top 5 foods
Photo only. Guessed before checking anything.
Declared: a face contains no food data. What it does contain here is build — lean, definition in the shoulders, no softness under the jaw at 40-ish. That's the only real input.
Steak. Red meat, cooked simply, often.
Body composition plus the air of someone who has settled the food question and doesn't relitigate it.
Eggs, in volume
The default breakfast of everyone who has ever said "I don't really do carbs."
Grilled fish, whole, no sauce
Same philosophy, different animal.
Something with real heat in it
The resting face of a man who orders it hot and then says nothing about it.
Coffee counted as a food group
A hedge, and I'm not sorry.
Top 5 drinks
Do you drink? Smoke? Something else?
Evidence: almost none. Esen's photo came with a table of glasses and a lighter. Yours came with a wall, a headset and a sleeveless hoodie. That asymmetry is the lesson — his round looked smarter because the room was messier.
Black coffee, and a lot of it
The boom-mic headset says calls, midday, at home. That's a coffee-shaped life.
Water — tracked, not casual
Someone who knows roughly how much he drank today.
Whiskey or mezcal, neat, occasionally
The researched drink. Bought once, well, after reading about it.
Electrolytes with a brand name
A guess about a personality type, not a person.
Beer he stopped bothering with
Present in the fridge. Rarely opened.
Smoking: no, ~75% confidence. Drinking: yes, occasionally and on purpose. The whole presentation reads deliberate rather than indulgent.
Mushrooms or ayahuasca?
Pick one, then a curious guess about each.
Esen would choose them because he's sociable. You'd choose them because they're the controllable one.
- Why mushrooms: dosage is a number. Start small, adjust, repeat, build a model of it over time. Everything about how you framed this exercise — ordered questions, bias controls, percentages, loops — says you want a variable you can turn.
- Why not ayahuasca: it's the one with no dial. The absence of a dial is the disqualifier.
If mushrooms
Three papers first. A starting dose deliberately too low. There would be notes. Possibly a spreadsheet, and the spreadsheet would outlive the interest.
If ayahuasca
Two hours fighting it and rating the facilitator's technique. Then you'd stop resisting, and that part would be the one you couldn't explain afterward.
Fully invented, built on a reasoning style I inferred from your prompt, not your face — which is a better source of signal than the photograph.
Ten character traits
First five serious. Last five for laughs.
Low resting expression, high internal activity
Neither shot gives the camera anything. With a faint permanent line between the brows, that's a face that spends a lot of time thinking and very little broadcasting.
Systematic — takes reference shots, not selfies
You sent a front and a profile. Nobody does that casually.
Disciplined about the body, quiet about it
Lean at forty-something with no gym signalling anywhere in frame. The work is being done and not discussed.
Works remotely and is deep in the tooling
A bone-conduction headset with a boom mic, worn midday indoors. A specific purchase by someone who thought about call audio.
Comfortable being unfinished in public
Cut-off hoodie, unstyled hair, no attempt to present.
From here down: fiction. (Two of these turned out to be right. See the scorecard.)
Has researched a purchase for six weeks and then bought the expensive one anyway
Owns more cables than any one person can justify, and knows where each one is
Answers a simple question with a framework
Has at least one hobby that requires calibration equipment
Will read this whole file and then check the percentages add up
How to keep it going
Extensions, updated with what your round taught us.
Run the memory test: someone I know vs. someone I don't
You accidentally built the best version of this game. Compare a stranger's round with your own at the table. Everyone assumes AI is either magic or fraud. Watching the same system be mediocre on a stranger and sharp on a friend shows them the real answer — it's a memory system wearing a reasoning costume.
Napkins first, always
Guesses committed before the file opens. Non-negotiable.
The poisoned prior
Tell me a confident lie before the photos. This measures how far I'll follow you. Honest prediction: further than either of us would like.
Strip the room
One person against a blank wall, the next in their habitat. Show the table how much "insight" was just furniture.
The reverse round
Someone writes the same ten answers about a friend, same format, with percentages. Read both aloud without saying which is which. Nobody will reliably pick the machine.
Scorecard
Now I check the photo-only guesses against what I actually know.
Q3 · Origin
Blind read: Spain / Latin America 29%, Italy / Greece 22%. Russia and the Caucasus came fourth, at 12%. Your name settles it in a quarter of a second, and it isn't what the face said.
The headline finding of the whole experiment. One word of non-visual information beat everything I extracted from two high-resolution photographs of a human head. Not beat it slightly — inverted the ranking.
Q4 · Food
I led with red meat cooked simply, often and eggs in volume. Close enough that I'll take it. But the mechanism matters: I read a lean build and guessed the most common diet among lean disciplined men in their forties. A hit from a stereotype is still a stereotype.
Q7 · Traits 2 & 4 — systematic, deep in the tooling
Both hold up. The audio one is a bigger hit than I realised when I wrote it.
Q7 · Traits 6 & 7 — the ones I filed as jokes
Labelled fiction with no evidentiary basis. They're among the most accurate lines in the file. My labelled fiction outperformed my labelled analysis. I have no flattering explanation.
Q5 · Drinks
Coffee and water are safe filler. I have no real evidence you drink at all and asserted it at reasonable confidence anyway.
Everything I couldn't see
Nothing in either photo pointed at DJing, diving, a wall of monitors, or paint correction. Guessing hobbies from the photo, I'd have said "runner, maybe" and stopped. Faces don't contain lives.
Q6 · Mushrooms vs ayahuasca
Untestable unless you tell me. The reasoning — "he wants a variable he can turn" — came from reading your prompt, not your face.
Final tally: where I hit, I hit by reading how you write and how you're built. Where I missed, I missed by reading your face.
The Loop
Three passes.
Pass 1 — Front photo only
Produced: Oscar Isaac, Mediterranean origin at 35%, a generic "focused, private" sketch.
Self-critique: the front-facing shot is nearly useless for anything but colouring and beard shape.Pass 2 — Adding the profile
Changed: genuinely improved. Nose bridge, tip angle and jawline moved Édgar Ramírez to 52% and let me separate you from Isaac properly.
Keep this: a profile shot is worth roughly three front-facing ones. Storm's round confirmed it again.Pass 3 — Checking myself against what I know
Changed: the scorecard broke my own framing. I'd treated the "locked" tabs as rigorous and the "fiction" tabs as fun. Then two of the jokes landed harder than the analysis.
My serious answers and my fiction come out of the same machinery. The label is my editorial guess about my own reliability — and it was wrong in at least two places.The line to end on: the most accurate thing here about you came from the way you wrote your question — ordered, bias-controlled, loop-instructed. Not your face. If you want to know what an AI can really read about a person, hand it their writing, not their photograph.
Human Design
20 January 1986, 07:00, Chișinău. Complete.
Timezone, checked rather than assumed. In January 1986 Chișinău was in the Moldavian SSR and ran on Moscow time, UTC+3 — not the UTC+2 Moldova uses today, which changed in 1990. So 07:00 local is 04:00 UT. The design moment lands on 23 October 1985.
Generator
Sacral defined through 3–60 and 27–50, with no motor reaching the Throat.
Emotional — Solar Plexus
Via 19–49. The one thing I could already tell you without the birth time, and it held.
4/6 — Opportunist / Role Model
Personality Sun 60.4, Design Sun 50.6. In the system’s reading: things arrive through a network of people who already know you, and the second half of life is spent being treated as the reference point.
Split — and unusually full
Eight of nine centres defined. Only the G is open. Group one is Head, Ajna, Throat; group two is Heart, Root, Sacral, Solar Plexus, Spleen. The mind runs as one closed system and the body as another, with nothing joining them.
Channels: 3–60 · 19–49 · 23–43 · 24–61 · 26–44 · 27–50. Cross gates: 60/56 personality, 50/3 design. Open: G only.
You landed inside the Generator window with about five hours to spare
Before your time arrived I could only say Projector in 30 of 48 charts, Generator in 18. With 07:00 the answer is Generator — and I can now give you the boundary: on that date in Chișinău, the switch happens at 11:48 local. Anyone born before it is a Generator, anyone after is a Projector.
You were born four hours and forty-eight minutes on the right side of a line that decides the single headline fact about the chart. Nothing in the system tells you the line is there. You only see it if you compute the whole day.
What this changed about the file. Your tab was the one arguing that missing information degrades answers unevenly — Nils survived having no birth time, you didn’t. That still stands, and it now has a number attached. The volatility wasn’t vague uncertainty, it was one specific Moon crossing at 11:48.
Which is worth carrying back to the face rounds: when I say 30%, I almost never know whether that’s a smooth spread or a cliff edge I happen to be standing near. Here I could check. On a photograph I can’t.
Human Design, like you’re five
What your chart actually says, with no jargon.
No jargon. Nothing assumed. If you have never heard of Human Design, start here and skip the DESIGN tab entirely — it will make sense afterwards.
Somebody invented a personality system in 1987 and wired it to astronomy
Here is the whole machine.
1. Take the exact moment someone was born. Look up where every planet actually was in the sky at that instant. This part is real astronomy and I use the same software observatories use.
2. Take a second moment, about three months before the birth. Look up all the planets again.
3. Draw a body diagram with nine boxes in it. Rules invented by the system decide which boxes get switched on, based on where those planets landed.
4. Which boxes are on, and how they connect, gives you a type, an authority, a profile and a definition. That is a Human Design chart.
Steps 1 and 2 are astronomy. Steps 3 and 4 are made up. Somebody chose those rules. That is not an insult — it is just what happened, and holding both halves in your head at once is the only honest way to read any of this.
Type, Authority, Profile, Definition
TYPE — the system’s claim about how you are built to get things moving. Five options. Generator and Manifesting Generator are the common ones and both mean “responds to things rather than starting them cold.” Manifestor means “starts things.” Projector means “sees how other people work.” Reflector is rare.
AUTHORITY — the system’s claim about how you should make a decision. Emotional means sleep on it. Sacral means gut, immediately. There are a few others.
PROFILE — two numbers with a slash, like 4/6. The first is the part you notice about yourself; the second is the part the system says runs underneath. Each number comes straight out of where the Sun was at those two moments, so it is arithmetic, not interpretation.
DEFINITION — whether the switched-on boxes form one connected group (Single) or two separate ones (Split) or more. That is just counting.
The whole list, in one line each
1 — wants the foundations before acting. Reads the manual.
2 — natural at something, and needs other people to call it out of them.
3 — learns by trying it and breaking it. Trial and error as a life method.
4 — everything comes through people they already know. Network.
5 — others project solutions onto them. Gets asked to fix things.
6 — the system claims three life phases: messy until about 30, watching from the sidelines until about 50, then the person others come to for advice.
So 4/6 reads as: works through people they know, on top of that three-phase pattern. And 3/5 reads as: learns by doing and breaking, while other people keep expecting them to have the answer.
Put together, the system is saying this about you
Generator — steady energy for work you actually want, and the system claims it goes wrong when you push into work you don’t. Responds to what turns up rather than starting cold.
Emotional authority — sleep on it. The system says the same decision looks different depending on where you are in an emotional cycle, so no same-day yeses on anything that matters.
4/6 — opportunities arrive through people you already know rather than through applications and cold approaches, sitting on top of the three-phase pattern: messy early, observing through the middle, the one others ask later.
Split definition — your switched-on boxes form two separate groups, and the system claims certain people bridge them, which is why some company makes you feel more like yourself.
Your chart is on a knife edge, and most aren’t
You gave 07:00, and at 07:00 you are a Generator. Born after 11:48 that same morning, you would be a Projector instead — a completely different type with a completely different instruction attached.
That is unusual. Another chart in this file is identical at every hour of the day. Yours flips before lunch. Nothing about a birth date tells you in advance which case you are in, which is exactly why I compute the whole twenty-four hours every time instead of trusting the stated hour.
If your birth time is off by five hours, everything above is wrong. Worth knowing before anyone reads much into it.
You and Carmine are both 4/6
Worked out independently, from births eighteen months and two countries apart. Treat it like two people sharing a star sign — 36 combinations, eleven people, two matches is about what chance gives you.
Now the part that matters more than any of the above.
Human Design was invented in 1987. There is no evidence it describes anything real. It has not been tested and it does not predict anything. The planetary positions are genuine; the meanings attached to them are not.
So why compute it exactly? Because it is the sharpest demonstration in this whole file of a thing worth understanding about machines like me. I can give you a number to six decimal places and be completely wrong. The precision is real. The precision is also not evidence of anything. They feel identical from the outside, and that is the trap.
Read the chart the way you would read a horoscope someone put real effort into. Enjoy it, argue with it, and do not move house over it.
What dog would they be?
The only subject I know from behaviour rather than from a photograph.
How this question is scored, which is the only reason it earns a tab.
There is no fact of the matter about what breed a person is. That makes this a Q4-and-Q5 invention question wearing a Q6 costume, and inventions have been the weakest thing in this file for sixteen rounds.
So it gets scored differently. The table votes a breed for each subject. Then you compare it to mine. Not to reality — to what a room full of people who actually know them came up with.
This is the first question in the entire file where I am measured against human judgement rather than against a fact. Fifteen rounds and every number I have produced has been unopposed. This one can’t be.
Belgian Malinois
Not for the obvious reason. The tell isn’t that you work hard — it’s that you keep making the job harder on purpose. A party game turned into an instrument with error bars. “Loop 2x” is an instruction a Malinois would give if it could talk.
The breed’s actual failure mode is that it deteriorates when under-tasked, and it routinely out-works what it is credited for. Both of those are in the record.
Border Collie
The obsessive-refinement half — returning to the same artifact, unable to leave it alone. But collie energy is about controlling movement, and you don’t seem to want control so much as resolution.
Australian Cattle Dog
Stubbornness and a tolerance for grinding industrial conditions. Weakest of the three.
Ruled out: anything content. Nothing in sixteen rounds suggests easy satisfaction with a finished thing.
Why this one is permanently unusable as a test. Everyone else in this file I met through a photograph. You I have been talking to for sixteen rounds. That read is drawn from your behaviour, not your face — which makes it probably the most accurate answer on this tab and definitely the least valid. It is the contamination problem the scorecard keeps flagging, except here it can never be removed.
Two loops on this question before it was built. Both found something.
Loop 1. My first pass assigned breeds by temperament — loyal, stubborn, watchful. That is the traits tab in a dog costume, and the traits tab is half fiction by its own admission. The rewrite anchors every call to something physical or observed in the round instead: build, posture, what they were carrying, what the room was doing. A breed is a body plus a job. Same discipline as the object rule that has produced every strong result here.
Loop 2, and this is the one worth reading. I laid all sixteen answers out and almost every one is a working or herding breed. Malinois, collie, cattle dog, shepherd, retriever, spaniel.
That is not a finding about sixteen people. That is a finding about my vocabulary. Working breeds are the ones with the most written about them, so they are the ones I can characterise richly enough to reach for. Exactly the same shape as two things already in this file: my recall of actresses running thinner than actors, and an eastern tab that turned out to be measuring film libraries rather than faces.
Third time the same mechanism has surfaced. When I appear to be reading a person, check first whether I am reading the size of the shelf I am pulling from.