paper-to-game

Design memo — Benjamin’s Veil

Walter Benjamin, Das Kunstwerk im Zeitalter seiner technischen Reproduzierbarkeit, third version (1936–39). A game in which the player keeps one face through seven ages, decides each age how near to let the people come to it, and watches the halo go.

0. Given

Mine, no ideas. No precedent overridden by instruction; the choices below are the engine’s.

1. Paper

Walter Benjamin, “Das Kunstwerk im Zeitalter seiner technischen Reproduzierbarkeit (Dritte Fassung)”, in Gesammelte Schriften I.2, ed. Rolf Tiedemann and Hermann Schweppenhäuser (Frankfurt: Suhrkamp, 1974), pp. 471–508; written 1936–39, first published in this form 1955 (Schriften). No DOI. The second version is in Gesammelte Schriften VII, pp. 350 ff.; the French version appeared in the Zeitschrift für Sozialforschung 5 (1936). An English translation of this third version is in Illuminations (1968) and, under the title “The Work of Art in the Age of Its Technological Reproducibility”, in Selected Writings vol. 4 (Harvard, 2003). Available: the 38-page PDF of the German text (typeset, with the footnotes), nothing else. The paper is an essay; there are no tables, figures or data, and no numbers at all beyond the dates of inventions and Huxley’s reading-public arithmetic in footnote 21.

2. What the paper hands over

An argument, primary and only. The essay is a verbal mechanism: technical reproduction detaches the work of art from its here-and-now, which is the seat of its authenticity and its aura; the masses’ wish to bring things close, together with the techniques that let them, changes perception itself, so that the aura decays everywhere, not only in copies; with the aura goes the work’s grounding in ritual (cult value), and what replaces it is exhibition value and, finally, a grounding in politics. The essay then reads photography and film through this mechanism (the face as the last refuge of cult value; the film actor tested by the apparatus; the star cult as a rebuilt aura; the cameraman as surgeon; reception in distraction) and closes with the political fork: fascism aestheticises politics, communism politicises art.

To play it, the argument is formalised into a toy (section 16): one face, seven ages, one control (the veil: how near the face is let), and two quantities Benjamin keeps apart, cult value and exhibition value, traded against each other at a price that is ours. The places where the toy chose something the paper did not are listed in section 15.

3. The discovery, and its kind

A counterintuitive result carried by a mechanism made intelligible. The stranger’s belief: the halo (the aura, the thing that makes the original special) belongs to the picture, and copies are the threat to it, so a keeper who refuses to copy keeps it. Benjamin: the aura is “the unique appearance of a distance, however near” (section III), a relation between the thing and the eyes looking at it; and the eyes are made by the age. Once an age has learned to want things close and to take them in copies, the aura withers even for the thing that was never copied: the veiled original is seen with the eyes of people who own the postcard. The keeper who hides the face does not keep the halo. He keeps a certificate (footnote 8: authenticity as the collector’s fetish). The mechanism underneath: cult value lives on unapproachability (footnote 7), exhibition value on reach; each age’s technique raises what reach is possible; and with photography, “exhibition value begins to drive back cult value along the whole line” (section VI).

Second discovery, at the end: when the aura is gone the longing for it is not, and where that longing is served, Benjamin says what comes of it: the market’s authentic original, the star’s “personality” (section X), the leader’s face at the rally and the war it leads to (epilogue), or the people filming themselves (section X, the Soviet film; section XIII). The game stages this as a committed fork, not a lecture: the player chooses, then sees.

4. The core narrative

There is one face, painted long ago. People love it, and part of why they love it is that they hardly ever get to see it: it lives in the dark, far away, and it has a halo. You are its keeper. Every age, a new way of making copies arrives, and the people press closer; you decide how near to let them. If you show the face, everyone can see it, but the halo gets thinner and never grows back. If you hide it, the halo stays bright, for a while. Then something strange happens: the halo fades even though you never lifted the veil, because the people’s eyes have changed. They have seen so many pictures that nothing looks far away any more. In the end there is no halo left to keep, only a face that everyone owns, and the last question is what that face is for.

5. Who lives in the paper

6. Precedent

Game Position Format Feedback device Chassis
Hubble’s Twelve Nights (hubble) experimenter (the paper’s own instruments, a budget of nights) discovery under a budget the plate (Figure 1 drawn live), with a reveal observatory at night (Instrument): sky black, starlight, arc amber, shift red, brass; Jost + IBM Plex Mono; brass-rimmed instrument plates
Cantor’s Bargain (cantor) agent (the constructor of E0, in a contest) strategic the strike (rows struck in the Lister’s book) fairground bill (Loud): big-top red, cream, ink, brass gold, teal; Rye + Alfa Slab One + Work Sans; ticket buttons
Watson’s Cardboard (dna) experimenter (builds the model) discovery, closed by a real-time copy the rails (two ribbons bow, pinch, run straight) cardboard model kit on kraft (Drawn): Permanent Marker + Courier Prime; luggage-tag buttons

What this game does differently.

7. The aim and the verdict

Aim, on the start screen, at the top of the play screen, first line of How to play:

Keep the face’s hold on the people through seven ages.

“Hold” is defined on the start screen in one line before the player presses anything: A face holds people in two ways: by its halo, which needs distance, and by being seen, which needs nearness. Each age you get one veil to set.

Verdict, the lines the end screen leads with:

What you did: You held the people for about four fifths of what was possible. You kept the veil down through the cathedral and the court, and lifted it in the age of the photograph.

What Benjamin says: The halo was never in the face. It was the distance between the face and the people, and the age sets the distance. Hide the face and the halo still goes, because the people’s eyes have changed; show it early and the age cannot yet carry it far. Once the copies are everywhere the face belongs to politics.

What that means: [the fork’s line, by choice: the vault / the studio / the rally / the newsreel]

The yardstick during play is in the aim’s units: after each age, “hold this age” as one bar, gold for the halo’s part and red for the crowd’s part, and the seven-age strip at the top fills one slot per age. The best possible is not shown until the end.

8. Three concepts

A. The Keeper (agent; challenge with the discovery inside). One face, seven ages, one control: the veil, five positions from “in the cella, priests only” to “on the screen, everyone together”, capped each age by what the age can copy. Before pressing the player weighs: the halo now, how much of it this veil spends, how far this age’s technique carries the face, and the sense that the people are pressing closer. After committing they see the crowd arrive, the halo thin, the copies appear in the margin, the age’s hold bar. Discovery moment: the age of the photograph arrives and the halo thins to almost nothing before the player has touched the veil; the counterfactual row under the picture shows the halo thin at every veil position. Fun: anticipation (the age passes after the commit), agency (when to open is a real plan with a real optimum), discovery (the halo goes anyway), mastery (a second run opens at a better age). Replay: the always-veiled run and the open-early run are different strategies with different fates, and the fork at the end has four endings.

B. The Screen Test (subject; discovery). The player is the film actor of sections VIII–X. A scene is shot in fragments out of order (the knock, the jump from the window, the shot fired behind the actor), the cutter assembles it, and the public examines. The player controls how much of themselves to give each take; the aura of the stage (the here-and-now before a live audience) is gone and the studio offers a “personality” instead. Deliberation: which take to hold back for the live stage, which to give to the apparatus. Discovery: the assembled film is better than any take and the actor is not in it. Fun: anticipation and discovery; little agency, since the cut is the cutter’s. Replay: none worth having. The walkthrough stalls at minute two: the actor has no quantity to trade, and the essay does not supply one.

C. The Examiner (subject; reckless; real time, no words on the play screen). Sections XIII–XV. The player is the public in the cinema. Images come at them as shocks, in the ballistic manner Benjamin describes; something in each must be caught (a hand reaching for a lighter, slowed; a crowd, enlarged), first in a painting that holds still and can be contemplated, then in a film that will not. The player gets better at catching in distraction, round by round: Benjamin’s claim that film is the training instrument of a new apperception. No words during play; a counter of catches. Discovery: the second mode works better than the first once learned, and the player did not concentrate to do it. Fun: mastery, agency, a little discovery. Replay: a skill run. It would fail if the training effect is not felt within five minutes, and it carries only sections XIII–XV; the aura, the cult, the fork would be left out.

9. The first five minutes

A, the pick. Minute one: a cream poster with a face in a black frame, a gold halo behind its head, the title in red condensed capitals. Two lines say the face is theirs to keep and that it holds people by its halo (distance) or by being seen (nearness). They press BEGIN. Age one, “The cave”: a line says the face is painted on the wall of a cave, that there is no way to copy a face yet, and that only the painter and the spirits see it. Two veil slabs are live (in the dark; shown to the clan), three are grey with “no way to copy a face yet”. They tap “shown to the clan”: the curtain over the frame parts a little and the face comes a little nearer. They press LET THE AGE PASS. A handful of figures walk in and stand far off; the halo loses a little of its thickness; one line: “The clan saw it. The halo is a shade thinner, and it will not grow back.” The age’s bar: nearly all gold, a sliver of red. The strip at the top gets its first slot. Minute two: the temple, the cathedral. The player is now holding the veil down (the halo is most of the hold, and the techniques carry nothing far), and grinning slightly at having worked that out. The cathedral says “new this age: the woodcut”, and the print slab comes alive. They look at it: print spends three quarters of the halo and reaches a few hundred. Not yet. Minute three: the court, then the print shop. The halo is visibly thinner at the start of the court age although they did nothing; the line says “the age’s eyes have changed: there are engravings in the houses now”. They frown. The print shop: the halo arrives at about half; the print slab says “the lithograph: in every house”. They weigh it: half a halo kept against most of the people reached. Many open here; some hold one more age. Minute four: the studio. The halo arrives at a quarter and the counterfactual row shows it at a quarter whatever they do. The discovery: the halo was going anyway. They open the veil if they have not, and the crowd floods in close. Minute five: the cinema, everyone together, then 1936: the halo is gone, the longing is not; four slabs, each a place to put the face; they choose, the face is drawn there, three lines say what Benjamin says of it; then the verdict. They can say: I kept it veiled until the photograph, then opened it; the halo went anyway.

C, the runner-up. Minute one: a black screen, a painting of a street, a question: “Where is the lighter?” They look, find it, tap. Minute two: the same street as a film, fast; the lighter is there for a third of a second; they miss it twice, then catch it. Minute three: the catches come easier; the counter climbs. Minute four: they notice they are not looking for it any more, it arrives in the corner of the eye; this is the point. Minute five: a verdict about distraction as a mode of mastery. Honest, playable, and about one fifth of the essay.

10. The pick, and why

A, The Keeper. The essay’s central mechanism (aura as distance; perception made by the age; cult value giving way to exhibition value; then politics) is the thing the player operates, and the one decision the essay allows a keeper (how near to let the people) is the player’s one control. The deliberation is real: an optimal-stopping problem whose optimum lies where Benjamin dates the turn, at lithography and photography. The discovery lands inside the play, after a commitment, at the age of the photograph. The sentence afterwards has no “guess”.

Author test. Benjamin would recognise the cella, the veiled Madonna, the woodcut arriving late in the Middle Ages, lithography putting graphic art into everyday life, photography driving cult value back along the whole line, and the face as cult value’s last refuge; he would enjoy the player hiding the face through four ages and finding a certificate where the halo had been; and he would insist on the fork, which the game gives him as its last age. The colleague who knows the answer still has to choose the age to open, and the heuristics table (section 17) shows the choice matters.

Runner-up: C. Switch if the author wants a game about the new perception (sections XIII–XV) rather than about the aura and its fate. It would be shorter, wordless, and a skill run.

What the reckless concept would have given. A play screen with no words and a real-time loop in which the player feels their own apperception change; the paper’s claim that film trains perception, staged as training. Given up: the clock, the catching, the silence. Kept from it: the counterfactual row, which puts the discovery in front of the player’s eyes rather than in a sentence.

11. The shape

Seven ages, one veil each, then one fork, then the verdict. Nine screens: start, play (the same screen, redrawn seven times), the fork, the end; plus How to play and About. No presets: the ages are the essay’s own sequence, in its own order, since the discovery needs the halo to have been held before it goes.

The ages and their sources:

Age Date on screen New this age (the technique) Veil ceiling Source
The cave long ago none: the painted wall shown to the clan section V (the elk, a magic instrument, shown to fellow men, meant for the spirits)
The temple antiquity casting and coining, not for faces in the forecourt section I (the Greeks’ two techniques); section V (the god in the cella, only for the priest)
The cathedral the Middle Ages the woodcut in the nave section I (woodcut); section V (Madonnas veiled nearly all year; sculptures invisible from the ground)
The court about 1500 the engraving in print, a few hundred section I (engraving, etching); footnote 2 (authenticity born with the woodcut); section V (panel painting more exhibitable than fresco)
The print shop about 1830 the lithograph in print, every house section I (lithography; the illustrated daily)
The studio about 1880 the photograph in print, every house, and in every hand section I; section VI (exhibition value drives back cult value; the face as last refuge)
The cinema about 1930 film on the screen, everyone together sections I, IV, VII–XV
1936 the fork the longing for the halo four places footnote 8, section X, the epilogue, section X and XIII

Score: yes. Hold per age (cult part plus crowd part) summed over seven ages, shown at the end as a fraction of the best possible in words, with the age the player opened named. The fork is not scored; it is read.

12. Controls, mechanism, feedback, moment

13. The look

Register: a 1930s photomontage poster (Worldly family: the constructivist and Heartfield line). It comes from the essay’s own time and company: Benjamin wrote it in Paris exile in 1936 for the Institut’s journal, quoting Vertov and the Dadaists, in the years of the photomontage poster, in which a halftone face, a red diagonal and a block of condensed capitals do all the work. The poster is itself a product of the techniques the essay is about. Comic is not the register; neither is the journal.

14. What is cut

The film actor and the apparatus (sections VIII–X) beyond one line at the fork; the cameraman as surgeon (XI); the optical unconscious, slow motion and the close-up (XIII) beyond the “on the screen” veil’s close-up; Dadaism and the shock effect (XIV); reception in distraction and architecture (XV); the Vienna school and the late Roman art industry (III); the Sistine Madonna’s own story (footnote 11); the theorists’ wrong question (VII); the reader becoming a writer (X) beyond one line at the newsreel; Huxley’s arithmetic (footnote 21); Marinetti’s manifesto beyond the sentence fiat ars, pereat mundus. The preface and the epilogue’s Marxian framing are kept to the About screen.

15. Divergences

16. Mechanism kind and claim boundaries

Kind: toy. The game runs a model built to illustrate the essay’s argument. No part of it is ported from the paper (there is nothing to port) or reconstructed from data (there are none).

Where the discoveries come from. The halo going under the veil follows from the exogenous eyes, which are the essay’s claim (perception changes with the age). The optimum age to open follows from the eyes and the reach table together, both ours but read from section I and VI; it is checked in section 17 against other prices and other schedules. The fork’s endings are the essay’s own sentences, not the mechanism’s.

17. Calibration plan and result

What calibration.js checks, against the essay’s claims (there are no numbers to check against):

Result: 31 of 31 checks pass (calibration-result.txt). The heuristics table, in hold units summed over seven ages (best possible 6.20):

Way of playing Veils by age Total hold Of the best
always veiled 0000000 4.63 0.75
always the widest veil 1223334 4.59 0.74
always shown on the day 1111111 2.72 0.44
open at the court 0003334 5.88 0.95
open at the print shop 0000334 6.20 1.00
open at the studio 0000034 6.12 0.99
open at the cinema 0000004 5.54 0.89
the best possible 0000334 6.20 1.00

Both ends are punished (always veiled and always the widest veil each land at about three quarters of the best); the optimum opens the print at the print shop and the two neighbouring strategies are within a tenth of it; the spread across the obvious ways of playing is more than half the best, so the decision matters. Robustness: with the crowd at half price the best first prints at the studio; at double price, at the court; with eyes declining a third more slowly, at the studio. Always veiled stays below nine tenths of the best in all three; always widest stays below the best in all three, and below three quarters except at double price, where it reaches nine tenths. The halo under the veil still falls below a quarter by the cinema with the slower eyes. The fork’s four places are checked for the orderings the essay gives them.

18. Build log

  1. Mechanism headless (mechanism.js), calibration (calibration.js) run in Node: 29/31 on the first run. The two failures were tuning: the widest reach beat the never-shown halo one age early (at the print shop; the essay dates the turn to photography) and always-widest sat exactly at three quarters of the best. The print shop’s print reach went from .60 to .55 and the temple’s and cathedral’s wall reach down by .03 each; 31/31 after. One wording fix from the loop: the “fewer came than might have” note no longer appears when the face was kept in the dark on purpose, and the arrival line compares the halo numerically rather than by its words.
  2. Loop on placeholders (loop.js, loop-log.txt): seven ages with a plausible plan, the fork, the verdict; after every action the printout says what changed (crowd, halo, copies, the bar, the counterfactual row).
  3. Skin: a 1930s photomontage poster; Oswald and Libre Franklin (latin subsets) embedded; red slabs with an ink rule; the face drawn in eight techniques by one SVG routine with patterns (hatching, stipple, halftone dots, film sprockets, poster dots); the curtains as the control’s visible state; the crowd as heads from behind; gold only in the halo. build.py inlines the mechanism verbatim and checks it byte for byte.
  4. Text: no decimals anywhere on screen; the halo in six words from bright to gone; the verdict leads in words (“nearly all”, “about three quarters”); About with the kind and the three claim lines.
  5. Screenshots at 380 px: screenshot-*.png. Fixes from them: the curtain slit widened so the halo’s gold shows through in the dark; the verdict’s “all” reserved for the best path itself.
  6. Browser session (session-test.js, session-log.txt): veil down through three ages, shown on the day at the court, print at the print shop and the studio, the screen at the cinema; help opened from the court and returned to it; the rally and its overlay; the verdict; About and back; a second run never lifting the veil, the newsreel, its verdict. No page errors (the visitor counter’s script is blocked in the sandbox, as expected); no horizontal scroll at 380 px; the commit button is disabled until a slab is tapped and no result is visible before the commit. Length about five minutes, estimated.

Self-review from the screenshots: the aim can be stated from the start screen alone (“keep the face’s hold on the people through seven ages”); the discovery can be stated from the top of the end screen alone (“the halo was never in the face; the age sets the distance”); the look (cream poster, red slabs, condensed capitals, a halftone face) resembles no game in the table.

19. Revisions after play (31 Oct 2026)

Six changes after a first play by the author; the register, the mechanism’s numbers, the fork and the verdict are as they were.

  1. The picture carries over. The frame at the start of an age was drawn in the dark until a slab was tapped, so a keeper who had shown the face saw it locked away again. Now the frame at age start shows the world as the player left it: the curtains where last age’s veil put them, the face at that nearness, the halo as it now is (already thinned by this age’s arrival), and no crowd yet. The slabs still start unselected every age (no carry-over of the control); tapping one moves the curtains from where they are. Age one starts in the dark. The session test checks the frame’s nearness at each age start against the previous veil.
  2. Copies accumulate. “Copies out” reset to “none yet” at every age start, as if the prints of earlier ages had vanished. Copies made in earlier ages now stay in the margin across ages, each in the technique of the age that made them (engraving, lithograph, photograph, film), and this age’s are added when it passes; “none yet: the one face only” only while no age has printed. The session test checks the count never falls between ages. Presentation only: the counts follow the mechanism’s reach as before.
  3. The marble. The temple’s thumbnail read as greyscale in the strip. The stone is now warm cream on the dark cella, with warm lines, so the slot reads as a statue; the technique rule (the face drawn in the technique of its age) stands.
  4. Say the cause once. The arrival line gave “The people’s eyes have changed: " from the cathedral on, explaining the discovery four times before the photograph. From the cathedral to the print shop it now states only the fact ("The halo arrives thinner than it was: about half what it was", or "a shade thinner" where the words would not change); at the studio, the full line with the cause, once; at the cinema the fact again. A calibration check (W7) holds this. The verdict keeps its explanation and Help its sentence.
  5. The cathedral’s announcement contradicted its options. It said a face could be cut in wood and printed while “in print” was greyed as too rough. The ceiling and the reason stay; the announcement now reads: “New this age: the woodcut. Saints are being cut in wood and printed in the market, roughly — too roughly for this face. Yours hangs in the choir.”
  6. Announcements name a place, never a veil state. “Veiled” is gone from the cathedral’s line; the cave’s line now says where the face is (the wall of the cave) rather than whom it is for. All seven say where the face is this age (the cave wall, the cella, the choir, the prince’s gallery) and not whether it is hidden or shown, which is the player’s and is carried by the picture. A calibration check (W8) holds this.

Calibration 33/33; the session test (two runs) passes with the two new checks; screenshots refreshed, including screenshot-age2-temple.png (the warm marble, the curtains as left) and screenshot-age6-studio-arrives-as-left.png (the studio arriving with last age’s curtains open to print and the print shop’s copies still in the margin).

  1. Continuity at the age boundary (second play). The frame at age start had jumped straight to the new age’s technique and halo. Now, when an age arrives, the frame is first identical to the last frame of the previous age (same curtains, nearness, crowd, halo, the face in the previous technique, the copies in the margin), held still for about a second; then, over about a second, the halo thins to its new value and the face re-renders in the new technique, which is what the arrival line describes (the line appears with the change). The curtains and the crowd stay where they were: the picture moves again only when the player taps a slab (the curtains move from where they are, the crowd stays) and when the age passes (the crowd is redrawn for this age). Age one has no previous frame and starts in the dark. The session test checks that the first frame of ages two to seven equals the last frame of the previous age (nearness, crowd count, halo, technique) before the arrival runs, and that curtains and crowd are unchanged after it; screenshot-age6-as-left.png and screenshot-age6-after-arrival.png show the studio’s two states.

20. Revisions after a third play (31 Oct 2026): the tally, and the picture that would not sit still

The tally. The first version scored each age as hold = halo + reach, the two added at equal weight. The author’s objection, on playing: who cares about an intact halo if nobody ever sees it? Under the sum, a keeper who never lifted the veil scored three quarters of the best, because the unseen halo counted in full every age; there was a tax on showing, not a trade-off. Benjamin half-defends the sum (cult value is the value of a thing that exists unseen, section V), but as a game it had no interior.

The revision makes an age’s worth the people moved: how many saw the face, times the halo they saw it with, plus a fifth for merely being seen (exhibition value as a value of its own, section VI); and each age has more people than the last (doubling), counted by their square root so that the late masses weigh more without drowning the keeper’s present. The three numbers (the doubling, the square root, the fifth) were settled by playing the plans against each other before anything was built, as build.md section 3 asks:

plan of the best
the best possible: dark, dark, the faithful at the cathedral, the wall at the court, print at the print shop and the studio, the screen at the cinema 100%
open at the court 97%
open at the print shop 97%
as open as the age allows, from the cave on 89%
open at the studio 77%
open at the cinema 43%
print at the print shop, then back into the dark 39%
the faithful on the day, every age 18%
always veiled 1%

So: a halo nobody sees moves nobody; showing from the first age is a little worse than opening step by step; opening late is much worse; and the best keeper opens one step wider as each technique arrives, going to print at the lithograph, where Benjamin dates the turn. Two alternatives were tried and rejected. The product with ages counted equally has no trade-off the other way: “as open as allowed” is the best path, because the small early crowds pay full halo and the age’s tax takes the rest. Linear population (no square root) hoards until 1830 then opens, with the early ages worth nothing, which is realistic and a worse game. Robustness (calibration P7): being seen priced at a tenth and at two fifths, the people counted more flatly (power 0.35) and more steeply (0.75), and eyes that change a third more slowly, all keep the best path’s first printing between the court and the studio, always-veiled below a fifth, and the best path never narrowing the veil.

On screen: the bar is now moved this age, drawn against the most this age could have moved with the halo as it arrived, gold for the halo’s part and red for being seen’s, with the fraction in words (“about half of what this age could move, mostly by the halo”); the strip’s small bars likewise; the start screen defines the quantity in two sentences and says that each age has more people than the last; the aim is “Over seven ages, let the face move as many people as it can.” The verdict leads with the fraction of the best path and now names the best path in words, age by age. The counterfactual row still shows the halo, since the halo thinning at every veil is the discovery. The fork stays unscored. The About’s assumptions paragraph says all of this in a stranger’s words.

The picture. The author saw the picture flinch at each age. The frame markup was identical across the boundary (checked in the DOM: same transforms, halo, crowd, technique), but the page was not: the red arrival line was inserted at the moment the age arrived, pushing the picture down as the face changed; and at the cathedral the title band wrapped to two lines. Now the space for the announcement (three lines) and the arrival line (two) is reserved on every age, the arrival line fades in rather than appearing, the cathedral’s date is “about 1400” so the band does not wrap, and the frame’s top edge sits at the same pixel on all seven ages, before and after the arrival (measured). The cathedral’s announcement was shortened to three lines; “On the day, the court, on the day saw it” became “The court, on the day, saw it.”

Calibration 36/36; session test clean; screenshots refreshed.

Fourth play, the same afternoon. The author still found the picture changing at the new age counterintuitive: the face was redrawn in the new technique on arrival, 0.9 s after the still frame. Now nothing in the picture changes at the arrival but the halo, which thins by what the age takes, which is what the red line beside it says. The age’s technique touches the face when the age passes: on commit the frame crossfades from the face as it was to the face as this age leaves it, with the crowd walking in. The veil slabs, while being tried, move the curtains on the old face. The session test now also checks that the technique is unchanged at the arrival.

21. Revisions after a fifth play (31 Oct 2026): the ladder is the techniques

The author, at the cathedral: the age announces the woodcut and the player cannot use it. The slabs were nearness rungs (the dark, the day, the wall, print, the screen) and the ages were techniques, and they lined up at only two points, so five ages announced something the ladder did not show. Now the rungs are the paper’s techniques, one per age, and the ladder grows: in the dark and on the day at the cave; coins at the temple (section I: “casting and stamping”, bronzes, terracottas, coins); woodcuts at the cathedral; lithographs at the print shop; photographs at the studio; film at the cinema. The old rungs stay (a face can still be shown on the day in 1930); the newest is marked NEW; what an age announces is what it offers (calibration P5c checks the announcement names the rung). The court and the engraving are gone: six ages, seven rungs. Posters, which the author proposed as a rung, went to the fork instead: the rally slab is now THE POSTER, the leader’s face on every wall and over the square, which is where Benjamin’s epilogue puts the face and where the game’s own look (a 1930s poster) was pointing all along. Film stays as the last rung because sections VIII to XV are about film, and the fork’s politics (the rally filmed, the people filming themselves) exist only because of it.

Each rung has its own halo cost and reach (CONFIG.VEILS, tagged). The coin is the designed temptation: the god on the drachma spends more of the halo than a woodcut (0.55 against 0.45) and reaches fewer eyes that matter (0.25 against 0.4), so coining at the temple then following the ladder lands at 87% of the best. Settled by simulation before building, over a grid of eyes schedules, rung costs and reaches:

plan of the best
the best possible: dark, dark, woodcuts at the cathedral, then lithographs, photographs, film 100%
wait for the lithograph, then each new rung 93%
coins at the temple, then each new rung 87%
the newest rung every age, from the day at the cave 86%
wait for the photograph 72%
print at the print shop, then back into the dark 43%
film only 36%
the faithful on the day, every age 19%
always veiled 1%

So the best keeper takes each new way of printing as it comes, from the woodcut on, and the only ways to do badly are to hide, to coin, or to wait too long. Being seen is now worth a seventh of a full halo (0.15); the eyes schedule is 1, 1, 0.9, 0.7, 0.45, 0.2. Robustness (P7, five variants) keeps the first print at the cathedral or the print shop, always-veiled below a fifth, always-newest below the best, and the best path never narrowing.

Copies go out by the rung chosen, not the age: coins are drawn as worn brass faces, woodcuts in heavy black line, lithographs in grain, photographs in halftone, film in frames; the copies of earlier ages stay in the margin. The counterfactual row shows only the rungs the age offers. The slabs are a little shorter (42 px) so seven fit under the picture. The space reserved for the arrival line is three lines, so the picture sits at one height on all six ages (measured). Calibration 40/40; session test clean.

22. The fork is cut (31 Oct 2026)

The author, on the 1936 screen: putting the face in the vault looked like a return to the status quo ante, and was that believable? It was not, and the fault was structural rather than in the drawing. The fork was a reading dressed as a choice: four places, nothing at stake, no mechanism, four pictures the game had to invent for outcomes it could not play. Three options were weighed. Keep it unscored (as it was): clean, but still a screen of pretend decision. Score it in the same unit as the ages: then the poster moves the most people of anything in the game and the engine’s best path ends in war, which is Benjamin’s warning exactly, and which a player would read as the game’s recommendation. Cut it: the game ends where its argument ends, with the halo gone under a veil the player kept down, at the cinema. The author chose the cut, as the cleaner game.

What stays of the epilogue is one paragraph on the verdict screen, “And then, 1936”, read with the player’s own six ages above it: the longing for the halo outlives it, and where that longing is served, Benjamin says what comes (the collector’s certificate, the star’s “personality”, the leader’s face over the square, the war), ending on his last line. The game’s look, a 1930s poster, still points at it.

The verdict now shows the engine’s target as a picture as well as a sentence: two strips of thumbnails with their small gold-and-red bars, YOUR SIX AGES and THE MOST A KEEPER COULD HAVE DONE, so the player sees age by age where their path and the best path parted. The last Next age button reads The verdict. Calibration 36/36 (the fork’s four checks gone); session test clean; one screen shorter.

23. Honest miniatures, and grain (31 Oct 2026)

The author: the strip of past ages must show the actual past, and the strip at the top and the strips on the verdict must be congruent. They were congruent (one renderer) and both dishonest: each slot showed the face open with its halo, whatever the veil had been, so an age kept in the dark looked like an age on show. Now every miniature is the frame as the age left it, drawn by the same renderer as the picture at thumbnail size: curtains where the veil was set, the face at that nearness, the crowd that came, the halo as it was left, in the age’s technique. In the dark: curtains almost shut and a sliver of face. On the verdict, the best path’s strip reads as curtains opening one step wider each age, which is the lesson as a picture. The counterfactual row is honest the same way: a miniature per rung of the picture the player would have had. And film now has a fine grain where the photograph has dots and the lithograph its stipple; the smooth silver face was the one technique drawn without its hand.

24. The number (31 Oct 2026)

The author: the machinery computes a target, and the verdict said it only in words. Now the verdict opens with the economist’s tally, large: the people moved as a share of the most a keeper could have moved, out of 100 (the best path 100, always veiled 1), labelled as what it is, with Benjamin’s words beneath it. The juxtaposition is the point: a number for the keeper’s craft, and a page of prose for what the number cannot hold. The words still carry the fraction (“about four fifths”), so the number and the sentence agree. A sweep of the miniatures (top strip, both verdict strips, the counterfactual row, each against the frame as the age ended) found them identical in the DOM.

25. The verdict says why (31 Oct 2026)

The author: “moved all the people it could have moved” points the wrong way when the halo was lost early; and do we need an apparatus that measures how much people are moved by the halo? We have it: every age is split into the part the halo did and the part being seen did, for the player’s path and for the best path. The verdict now uses the comparison. It finds the ages where the player fell short of the best path by more than a thirtieth of the best total and names the cause the mechanism can tell apart: the face kept farther than the age allowed (the best path’s veil was wider: fewer saw it), or the halo spent earlier (the veil as wide or wider, but the halo they saw it with thinner). One or two sentences under the score: “But the halo had been spent at the temple, so the crowds of the cathedral and the print shop saw a face with little of it left.” / “But at the cathedral and the print shop the face was kept farther than the age allowed, and fewer saw it than could have.” Nothing is said within three hundredths of the best. Calibration 38/38.

26. The halo never goes on its own (31 Oct 2026): the original, the copy, and the eyes

The author, at the design’s last fork: the halo should never fade in the dark; a keeper who hides the face through every age and shows it at last to a mass audience is one of the game’s most interesting paths, and under some weights it might even be the best. The first version thinned the stock itself as each age arrived, which Benjamin never claims. What he claims is about the eyes (section III: perception is historically conditioned; the masses’ urge to bring things closer by way of the likeness) and about the copy (section II: “even the most perfect reproduction lacks one thing, the here and now of the work of art”; sections VIII and IX: the film actor’s aura is gone because he performs for the apparatus and not for a present audience).

So the mechanism now keeps three things apart that it had merged. The stock: the halo as it is, spent only by showing (by one minus the nearness of the means), never by time; kept in the dark it stays whole, and is drawn whole, to 1930. The carry: how much of the halo a showing by a given means conveys: presence all of it, a coin 0.3, a woodcut 0.5, a lithograph 0.6, a photograph 0.7 (section VI: where the aura emanates for the last time), film 0.35 (the screen cannot carry presence). The eyes: each age’s multiplier on what a halo does to the people, cumulative, never on the halo itself. What moved an age’s people is the stock times the eyes times the carry. The tally, the weights and the exhibition term are unchanged.

Simulated before building (with the control that proves the eyes are needed):

plan of the best
the best possible: the faithful on the day at the cave and the temple, then woodcuts, lithographs, photographs, film 100%
the day until the lithograph, then the ladder 98%
woodcuts on, then the ladder · wait for the lithograph 96%
the newest rung every age, from the cave 86%
coins at the temple, then the ladder 84%
wait, wait, then splash on film 45%
the faithful on the day forever 23%
always veiled 2%
with no change of eyes at all: best path hoards until the photograph; the splash scores 71 rejected

The splash fails, and not because of our weights: five crowds saw nothing, and the sixth saw a screen, which carries little, with eyes that had seen a thousand faces. Whether the keeper weighs the early ages or the late ones makes no difference to that, which is why the author’s other idea, a keeper’s own weighting chosen upfront (Einstein’s merchants), was left out: it would change nothing, and a choice that changes nothing is the fork again.

On screen. The original is one painting and never changes; what changes is what is shown. Tapping a rung is a preview of what the people would see by that means: the painting itself with its halo (in the dark, on the day), or the copy, drawn in the copy’s own hand (a worn brass coin, a woodcut, a lithograph, a halftone, a film frame) with the trace of the halo it carries as a printed ring of gold dots. The control shows what it does, and what it does is the essay’s subject. When the age passes, the crowd walks into that same picture, and the heads the halo moved are rimmed in gold in proportion; in the cinema a thousand heads and a handful of rims. The arrival of an age does nothing to the picture at all; the one thing said, at the studio, is that the people’s eyes have changed. The counterfactual row is gone: the preview is the better version of it. Every miniature is the frame as the age left it. The hoarder’s verdict: “The halo is whole, and nobody has seen it.” The splasher’s: “You kept the halo whole through every age and showed it at last to everyone, on a screen, which cannot carry it, to eyes that had seen a thousand faces.” The shortfall diagnosis now groups small losses by cause. Calibration 40/40 (P1 and P2 rewritten for the new reading, including the no-eyes control); session test clean; the picture sits at one height on all six ages.

27. The last audit (31 Oct 2026): every transition, every strip, every line

The author played once more and found small nibbles: a frozen counter in one of the pictures, and continuity slips. A script now drives four paths (the ladder, the splash, the coin-and-back path, the hoarder) through headless Chromium and checks at every step that the frame at the start of an age is byte-for-byte the frame the last age left; that each strip slot is byte-for-byte the frame it stands for; that a preview has no crowd and does not move the frame; and that the crossfade at commit starts from exactly the preview on screen. What the audit found, and what changed:

Calibration 40/40 (W4 updated for the new halo word); session test clean; the audit reports no violation on any of the four paths, and the frame top is at the same pixel on all six ages.

28. No score (1 Nov 2026)

The author looked at the contact sheet and asked whether the best path was simply to show the face as widely as the age allowed. It was: the ladder 100, “always the newest rung” 87, opening late 94, the splash 35, the faithful forever 23, the dark 0. The eyes change whether the keeper shows or not, so a halo kept is a halo devalued, and the keeper’s problem collapsed to timing. Two ways out were on the table: make the tally concave in depth, so that few people moved whole weigh against many moved a little (the essay’s own polarity, cult value against exhibition value), and then the splash and the day-forever become rival ways of keeping; or drop the number. The author chose to drop it, and it is the cleaner Benjamin: the essay does not rank the two values, it says the aura goes and tells you what takes its place. A score needs an objective, the objective picks a side, and then the game says “show it” with a confidence the essay never has.

Gone: the score out of 100, the tally line, the fraction in the lead, the shortfall diagnosis, the strip and sentence for “the most a keeper could have done”, and the per-age “moved this age” bar. The aim is now a decision, not a maximum: “Six ages. Decide who sees the face, and how.” The start screen says, in its last line, that nobody keeps score.

Kept, because none of it is a number: the crowd walking in, the heads the halo moved rimmed in gold, the line after each age, the halo stock in the picture and in words, your six ages, the painting as it is now, Benjamin, and 1936. The arithmetic stays underneath so that the crowds, the rims and the lines agree with each other; bestSchedule and the heuristics remain in the mechanism for the calibration only.

The verdict now sorts the path into one of five families and describes it in the essay’s terms, ranking nothing:

Coins at the temple, the faithful before the copies, and print-then-put-away each get a clause within their family. Calibration 42/42: S now checks that the verdict carries no score, no fraction and no best path, S2 that the five families sort the paths, S3 that every lead is a sentence about what happened with no digit and no “better”, W10 that the painting-now line says where the halo went. Session test clean. The skill gains a paragraph under “The ending”: where the paper does not rank, the game does not rank; and the paragraph on second halves (play them only if a second mystery is at stake) that had been pending since Hubble.