paper-to-game

Noether’s Ball — design memo

Renamed (1 Nov 2026): the game was Noether’s Press through the press architecture and revisions 1 to 5c; after the deck rebuild the author renamed it Noether’s Ball, after the thing that features biggest in it (and not Noether’s Clock, which would echo Einstein’s on the Pantheon shelf). The file is now noethers-ball.html; noethers-press.html redirects to it. Older entries below keep the old name.

Revision 8b (2 Nov 2026): the words under the formula. Having played revision 8 the author found the gloss under the formula, now the card’s load-bearing line, too small to read even on a laptop (13 px, dim grey). It is now body size (16 px) and white, and every symbol in it (g·(1 + a·t), x′, r, w·x) is set in the formula’s own face, marked between backticks in the world’s words string, so the eye can jump from a symbol in the formula to the same symbol in the words. The author had floated setting the whole line in math script; declined, because a sentence in italic serif reads slower and would look like a second formula. Calibration 14/14; browser session clean.

Revision 8 (2 Nov 2026): the traps become traps. A fresh read found six of the nine cards answerable from the setup line alone: the trap cards announced themselves three times (the setup line, a caption in the picture, the band’s world name), so the plate, the paper’s own subject, was decoration until card 7. What changed: on cards 2, 4 and 6 the setup line and the picture’s caption no longer say what is different (“The same ball, somewhere else”; “The same throw, with a magnet in the picture”; “The same swing, somewhere else”), the bands read ANOTHER BALL and ANOTHER SWING, and the formula and its words carry the news (“as before, but g·(1 + a·t): the pull of gravity grows with t”). The magnet and the wind arrows stay in the picture; the reveal and the reason plate explain the trap afterwards as before. Two smaller fixes: the pink badges for a right answer no longer show a struck-out ENERGY (they are named by the trap: GRAVITY GROWS · no energy prize; THE MAGNET · no sideways-speed prize; THE WIND · no spin prize), and “empty” is retired in favour of “zero” everywhere the screen speaks of the slide’s energy. Calibration 14/14; browser session clean.

Revision 7b (1 Nov 2026, later): the fake’s job, and the last line. The author still found the last line opaque and asked for a version that names Hilbert. Working it out showed that the ending rests on the fake: a real law is one a fake trip can break. The game showed fakes on cards 7 and 8 but never said what they were for. The fakes stay where they are; the author chose this cheap version, and a fake on card 1 is held back as the option if “law” still feels thin. Card 7 now says what a fake is (a trip nature never makes) and that the fake breaks the ball’s energy, which is what makes energy a law. Card 8 says nothing breaks the slide’s energy, not even a fake, so it is no law. The help defines a fake. One verb, break, is used throughout: “catch” is retired, and the pink cards no longer say a law “breaks” (they say there is no prize). Cards 7 and 9 and the debrief now say “how fast the clock runs”, so the last line points back at what the player did. The last line, under the heading In short, is the author’s own with one change: Hilbert had claimed that Einstein’s gravity has no real law of energy. Noether showed why: its formula doesn’t care how fast the clock runs, so its energy stays the same on every trip, real or fake. A law that even a fake can’t break is no law. Card 7’s badge is A REAL LAW · a fake breaks energy. Calibration 14/14; browser session clean.

Revision 7 (1 Nov 2026, late): plain words. The author found the language opaque, starting with the end screen’s last line (“the slide’s clock is free”), and asked for every word to be understandable to someone of medium intelligence seeing the game for the first time. The diagnosis: shorthand coined in our design discussions (a fixed or free clock, real or empty energy, re-ticking, “this is the paper”) that the screen used without showing or defining it; and one claim, that the ball’s energy is “real”, which the game asserted but no longer showed once the fake curves had gone. What changed:

Revision 6 (1 Nov 2026, night): the final sweep. With the deck in place the author judged the drawing a gimmick and asked for it to go, and asked whether the first card alone earns the energy badge. It doesn’t, and the game now says so. What changed:

Revision 5c (1 Nov 2026, later still): one answer, nine badges. After playing 5b the author judged change your mind? too easy with both pictures in front of the player, and asked to go back to one answer before the reveal; the second picture is now the reveal itself. He also found the automatic fake curve unnecessary (now gone: the prize’s line is drawn along the real trip only, with one optional “draw a path: watch it wobble” on the first prize, which the slide then repeats and fails), and was puzzled that right answers on the traps earned nothing. The shelf is now nine badges, one per card, earned by a right answer: five red for the paper’s lessons (energy, sideways speed, spin, the empty slide energy, the rule) and four pink for the traps (the law that breaks: gravity grows, the magnet, the wind, the ball’s clock is real); a wrong answer leaves the badge dull. The debrief names the red badges won and missed in words and says, as the author proposed, when the pink ones were won but the core got away. No score. Calibration 11/11; full browser session clean.

Revision 5b (1 Nov 2026, later): two pictures, guess then decide. On the first card the author saw only the 3:00 throw: the 3:05 throw flew exactly on top of it, so an overlay hid the very thing the card is about. At his suggestion each card now shows two pictures, one above the other: the trip as it is (top), which the player can replay as often as they like, and, after a guess, the trip after the change (bottom), the two running in step. The camera of the second picture moves or turns with the change, so “doesn’t matter” always looks like two identical pictures (a moved magnet or tilted wind arrows show what did change). Then change your mind? (keep or switch), then the verdict: the formula’s number for both trips, right or wrong, and the prize. The end sentence also says how the player’s changes of mind went. Animation keeps one clock for both pictures, so a ball that lands early stops early. The shown re-ticking is stronger (0.22). Calibration 11/11; full browser session clean.

Revision 5 (1 Nov 2026, evening): the architecture rewritten as a deck. After four revisions on the press architecture the author judged it still too hard to follow, and asked for one final try with the whole architecture rethought. The diagnosis: the press had accumulated parts (press, stamp, auditor, needle, rack, blanks, six levers, four worlds), each added to explain another, while the paper’s gist needs only three things: a formula that gives every path a number, a change the formula doesn’t care about, and a number that stays the same along the trip. The author’s own proposal set the format: show different experiments and let the player say whether the change mattered; score: none, just the shelf and the sentence. The game is now: an opener (draw the fastest slide; the formula’s number; the cycloid revealed), then nine cards, each a picture with its formula on a plate, a change (start the clock later, move it, turn it, re-tick the clock), and the question does it matter? The answer is computed from eq. 11; the reveal shows the new trip against the dashed “where it would go if it didn’t matter”, and the formula’s number before and after; every change that doesn’t matter puts a prize on the shelf (energy, sideways speed, spin), with its line drawn along the real trip (flat) and a fake (wobbly). Four trap cards (gravity that grows, a magnet, a wind, the re-ticked ball) matter. The slide’s later start doesn’t matter, but its energy is empty: flat on every path, and the player draws paths and fails to move it. The last card, the slide re-ticked freely, doesn’t matter either and wins a rule instead of a prize, which is why the energy was empty. The end: the shelf, one sentence naming what fooled the player, and the paper in four short paragraphs. The press, the rack, the blanks, the levers, the auditor, the spring-sheet worlds and the reverse rounds are gone; the steel-and-neon look, the title, the formulas on screen and the slide stay. The previous sources are in previous/. Sections 0 to 18 below describe the press architecture and its revisions 1 to 4, and are kept as the record; where they disagree with this note, this note holds. Calibration rewritten for the deck: 11 checks, 11 passed (every card’s answer, the plate and the picture agree with it, Theorem I’s three prizes have teeth, the clock’s prize on the ball is ½ u′ ² + g·y, Theorem II’s identity on the slide, §6’s improper energy, the opener’s honesty, thirty drawn re-tickings, robustness over thirteen settings, no decimals). Full browser session clean.

0. Given

Mine: Steffen’s signature and counter. Ideas given, and binding: read the paper against its reputation. “Symmetry gives a conservation law” is Theorem I and almost the warm-up; the paper was written to answer Hilbert’s question of why energy conservation seems to fail in general relativity, and the answer is Theorem II (§6). Build what the paper actually does, and note that it runs both ways: every conservation law comes from a symmetry too (§3, §4). No biography: no Göttingen, no habilitation, nothing about who was allowed to lecture; the paper stands on its own, next to Einstein’s. The design: hard, steely, manly, maybe with some neon. These override precedent nowhere in position or format; in the look they pick the register (Techno, a machine hall) and the palette (steel and neon), which precedent also allowed.

1. Paper

Emmy Noether, Invariante Variationsprobleme. Nachrichten von der Königlichen Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-physikalische Klasse, 1918, pp. 235–257. Presented by F. Klein at the session of 26 July 1918; final manuscript submitted end of September 1918. No DOI. Available: the Wikisource transcription (German), exported as a 38-page PDF, typeset text, no scan; read in full, including the 28 notes. The paper is in German; every quoted term below is translated. Online: de.wikisource.org/wiki/Invariante_Variationsprobleme; English translation by M. A. Tavel, Transport Theory and Statistical Physics 1 (1971) 186–207, arXiv physics/0503066.

2. What the paper hands over

A model, primary: two theorems about a variational integral I = ∫ f(x, u, u’, …) dx and a continuous group of transformations of the x and u, with explicit identities (eqs. 12, 13, 16) that can be computed for any concrete f and any concrete transformation. Ported: the game computes Noether’s identities numerically on small integrals of the types the paper itself names (the simple integral of mechanics; the Weierstrass parametric integral, p. 7, her first example of Theorem II).

An argument, secondary: §6, the proof of Hilbert’s assertion. In a theory invariant under an infinite group that contains the displacement of the independent variables as a subgroup, the energy relations are improper: they are linear combinations of the identities of Theorem II plus divergences that vanish identically, so they hold whether or not the equations of motion do. Proper energy theorems exist exactly when no such infinite group is there (classical mechanics, special relativity: note 26). The game stages this on a one-coordinate miniature.

3. The discovery, and its kind

A counterintuitive result. Everyone knows that more symmetry means more conservation laws. The paper shows that a symmetry with a free hand in it (an infinite group, with arbitrary functions) gives no conservation law at all: it gives identities among the equations of motion (Theorem II: ρ of the Lagrange equations are consequences of the rest), and it empties the conservation law of any finite symmetry it contains. Energy conservation in general relativity is not violated; it is void: true for every motion, including the impossible ones, so it says nothing (§6, the “improper” energy relations). And it runs both ways: a conservation law implies a symmetry (§3), and a set of identities among the equations implies an infinite group (§4); so if your energy law is empty, your theory has a free-hand symmetry that contains the shift of time. Hilbert’s “failure of proper energy theorems is characteristic of general relativity” holds only if “general relativity” means invariance under such groups (last paragraph).

Secondary, a mechanism made intelligible: Theorem I as a press. Put a symmetry in; a conserved quantity comes out, equation 13, computed, not quoted.

4. The core narrative

There is a machine that stamps laws. You feed it a thing the world does not care about (start the clock a minute later; move the whole track a yard; turn it) and it stamps a steel tag: a quantity that never changes while the world runs. An inspector then tests every tag by feeding it fake motions, motions the world would never make. A good tag catches the fakes: its needle swings. One day you feed the machine the biggest thing of all: not shifting the clock, but re-ticking it any way you like at every moment. The machine accepts; the world does not care. But the tag that comes out is blank, and worse, the clock tag you stamped before has gone blank too: its needle no longer swings for anything, real or fake. The machine has not broken. A freedom that big does not give you a law; it takes the clock away, and with it everything a clock would have let you say. That is why energy seems to be lost in Einstein’s theory. It is not lost. There is no clock in the theory to keep it by.

5. Who lives in the paper

6. Precedent

Game Position Format Feedback device Chassis
Brown’s Restless Dust (brown) controller (sets the conditions of one drop, aims at stillness) impossible challenge the live drop and the shelf of suspects hand-coloured natural-history plate, 1828 (Austere): plate cream, ink, Prussian blue, almond gold, carmine; IM Fell English; labels with a double rule
Benjamin’s Veil (benjamin) agent (keeper of one work, a timing decision) challenge: optimal stopping the halo and the crowd 1930s photomontage poster (Worldly): cream, ink, red, halftone grey, gold leaf; Oswald + Libre Franklin; red slabs with an ink rule
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

What this game does differently.

7. The aim and the verdict

Aim (start screen, top of play, first line of How to play): Find what the formula doesn’t care about. Each one wins a number that never changes. (Revision 1’s aim was Find the changes this world does not notice. Each one stamps a law; the first build’s was Stamp this world’s laws. Only a tag with teeth counts, which led with two words the player had not yet met.)

Verdict (first lines of the end screen):

8. Three concepts

A. Noether’s Press (the transformation; strategic). Controls: five levers, one pull at a time, three steel blanks per world; the two free-hand levers open a strip where the player draws p(x) with a thumb. Deliberation before a pull: does this world care about this? (Does the spring care where the track starts? Does light in glass care how fast the clock ticks?) A wasted pull bends a blank. After: the press holds still and stamps, or jams and the world’s dial swings (not a symmetry). Then the Auditor feeds motions: the needle, and his stamp. Discovery moment: the glass world; the clock tag’s needle is dead on the fakes; VOID; then the free-hand clock lever stamps, and the tag reads not a law but an identity. Fun: anticipation (will it jam?), agency (reading the world is the skill), discovery (the void tag), mastery (world four is the same trap run backwards, and the player who understood the glass answers it in one pull). Second playthrough: the worlds are fixed, but the pulls are the player’s; a second run stamps the laws the first run never reached (the floor has four, the blanks allow three) and tries the free-hand track lever, which the first run usually did not.

B. The Auditor (experimenter; discovery). The player is Hilbert’s inspector: each world comes with laws already stamped by someone else, and the player’s instrument is the fake: they draw fake motions and watch which tags catch them. The player’s control is the perturbation δu. Discovery: a tag that catches nothing. Deliberation: which fake would break this tag? Fun: discovery, agency. Second playthrough: different fakes. Weaker: the player never touches the symmetry, so Theorem I is told, not operated, and the converse has no place.

C. The Rubber Clock (reckless: the subject, wordless, real time). The screen is one bead moving on one track with two clocks beside it. The player’s thumb on the screen is p(x): dragging warps the clock under the bead in real time. In Newton’s world the warp breaks the motion (the bead’s trace and the energy dial go wrong, the screen flashes); in the glass world the warp changes nothing but the ticking, and the energy dial is pinned at zero whatever the thumb does. No levers, no words on the play screen; the discovery is felt in the thumb: the world that does not resist the warp is the world with no energy. Fun: agency, discovery. Second playthrough: none; the lesson is one gesture. Risk: a demonstration with a thumb on it; no commitment, no verdict with stakes; the converse and Theorem II’s identities have nowhere to appear.

9. The first five minutes

A, the pick. 0:00 A dark steel panel, a neon strip, a title. Stamp this world’s laws. Only a tag with teeth counts. Three lines: you pull levers, the press stamps, an inspector tests. BEGIN. 0:20 World one: THE OPEN FLOOR. A window shows a bead sliding across an empty floor, trace behind it. A line: A bead on a flat floor. Nothing pushes it. The clock ticks evenly. Five levers under it: SHIFT THE CLOCK, SHIFT THE FLOOR ACROSS, SHIFT THE FLOOR UP, TURN THE FLOOR, WARP THE CLOCK, WARP THE FLOOR. Three blanks lit. Under the levers: Pull one. If the world does not care, the press stamps a tag. The player thinks: an empty floor; it won’t care where it starts. Pulls SHIFT THE FLOOR ACROSS. 0:40 The press comes down with a thud; a steel tag drops out: SPEED ACROSS, STEADY. The Auditor’s panel lights: TRUE MOTION, the needle sits still; FAKE, the trace kinks in magenta and the needle swings; stamp: TEETH, cyan. Grin: it worked and it was mine. 1:00 Two blanks. The player wonders about the warp levers. Pulls WARP THE CLOCK; a strip opens, draw how the clock should tick; they drag a wobble; the press comes down and jams, the world’s dial swings yellow: THE WORLD MOVED. NOT A SYMMETRY. One blank bent. Frown, then: of course, a bead at speed cares how the clock ticks. 1:30 Last blank: SHIFT THE CLOCK; tag: ENERGY, STEADY; teeth. NEXT WORLD. 1:50 World two: THE SPRING. The Auditor moves first: On every true motion of this world, one thing holds steady. Here it is. A picture: the bead swings about the centre, and a sector sweeps at a constant rate. Which lever stamps it? The player looks: a sweep about the centre, so turning. TURN THE FLOOR. Stamp: SWEEP RATE, STEADY; teeth; That is mine. A law with teeth means a symmetry; this one was yours. 2:30 World three: THE GLASS. A ray of light bending through layered glass. Light crossing layers of glass, denser lower down. It takes the quickest path. The clock’s tick is only a label along the ray. The player pulls SHIFT THE CLOCK, confident. Tag drops: it is blank. The Auditor feeds a true motion: needle dead. A fake: needle dead. Stamp, magenta: VOID. This holds for every motion, the real ones and the fakes. It says nothing. Frown: what? 3:10 Second blank: WARP THE CLOCK, draw anything. The press comes down and holds; the tag reads not a quantity but a sentence: the equations of motion are not all independent; one follows from the other. The Auditor: A free hand. That is why your clock tag was empty: the shift of the clock is inside the warp, and the warp stamps no law. Discovery landed. 3:40 Third blank: SHIFT THE GLASS ACROSS: SIDEWAYS SPEED TIMES DENSITY, STEADY; teeth (Snell’s law, unnamed). 4:00 World four: THE SHEET: a bead rolling on a curved sheet, the mini-Einstein world. The Auditor first: Here the clock tag is blank before you pull anything. Pull the lever that says why. The player who understood pulls WARP THE CLOCK in one; the press holds; the tag reads the identity; the Auditor: Then this world has no clock of its own. That is Hilbert’s answer. 4:40 Verdict: words; Hilbert’s question and its answer; both ways; the last line of the paper.

B, runner-up. 0:00 The same panel; the aim: Find the tag that catches nothing. 0:20 World one with three tags already on the rack: ENERGY, SPEED ACROSS, SWEEP. A window with the bead; a strip under it: draw a fake: drag the trace off its line. The player drags; the needles of the three tags swing. Fine. 1:30 World three: the glass; four tags; the player drags fakes; three needles swing; one, CLOCK, never does. Found it. 2:30 Then what? The game tells them why. The player never operated the symmetry and the discovery is “one of these dials is dead”, which is a search, not a thought. The walkthrough runs out of decisions at three minutes.

10. The pick, and why

A, Noether’s Press. The player is inside the paper at the one place the paper puts a free choice: the transformation, including the arbitrary function drawn by hand. Theorem I is operated, not told: pull, stamp, test. The discovery lands inside play, on a tag the player stamped and expected to have teeth. Theorem II appears as a thing the press does (a stamped identity instead of a law), §6 as the blank clock tag, and the converses as the Auditor’s two reverse rounds. The sentence afterwards has no “guess”: I warped the clock on the glass and the press held, and the clock tag was blank; a free hand stamps no law. Originality: no game in the table has let the player be the transformation, or draw a function with a thumb as a move, or fight an inspector for tags.

Author test. The author would recognise her identities (12), (13) and (16) computed on her own Weierstrass example, the displacement group and the “Energierelationen” of §6, “eigentlich” and “uneigentlich” as TEETH and VOID, the converse of Theorem I as the Auditor’s second round and the converse of §6 as his fourth, and the last paragraph’s caveat on the verdict screen. She would not find Göttingen anywhere. She would enjoy a player pulling SHIFT THE CLOCK on the glass with confidence.

Runner-up: B. Switch if the author wants the inspector’s seat (Hilbert’s) rather than the group’s: a game about testing laws rather than making them.

What the reckless concept would have given. C’s thumb-on-the-clock in real time, wordless. Kept from it: the two free-hand levers open a strip where the player draws p(x) with a thumb, so the arbitrary function is literally in the player’s hand. Given up: the silence, the real time, the bead under the thumb.

11. The shape

Four worlds, three blanks each, twelve pulls at most; five to seven minutes. Screens: start; play (one per world, the same layout); the warp strip (an overlay); end; How to play; About. No presets in the paper’s sense; the worlds are the paper’s own cases, ordered for the player:

  1. The open floor (free particle in the plane, f = ½(u₁’² + u₂’²), x = time). Theorem I with four parameters: clock, two shifts, a turn. Every finite lever stamps with teeth; both warps jam. The warm-up; Theorem I.
  2. The spring (f = ½ u’ ² − ½k u ²). The Auditor moves first with the sweep-rate law (angular momentum, drawn as a sector sweeping steadily); the player finds the lever: the converse of Theorem I (§3). Shifts jam.
  3. The glass (Fermat: f = n(u₂)·√(u₁’² + u₂’²), the Weierstrass parametric integral of p. 7, homogeneous of first degree). Clock shift stamps VOID; WARP THE CLOCK stamps the identity Σψᵢuᵢ’ = 0 (p. 7); shift across stamps with teeth (Snell); shift down, turn, warp the floor jam. The discovery.
  4. The sheet (geodesics on a curved surface, f = √(gᵢⱼ(u) uᵢ’ uⱼ’), the one-coordinate miniature of Einstein’s integral). The Auditor moves first: the clock tag is blank; find the lever that says why: the converse of §6. WARP THE CLOCK stamps the identity.

Score: none shown. The rack of tags (cyan with teeth, magenta void, yellow bent blanks) is the yardstick in the aim’s units; the verdict is in words.

12. Controls, mechanism, feedback, moment

13. The look

Register: a stamping press in a night-shift machine hall, neon-lit (Techno family: a machine panel, a cockpit). Not from the paper’s subject (it has no picture) and not from its setting (a lecture room in 1918 is barred by the brief); chosen for the play: a press stamps, an inspector stamps, levers are pulled, blanks are bent. Hard, steely, and lit by tube light: the user’s brief.

14. What is cut

The general n-fold integral and higher derivatives (κ, σ): the game runs the simple integral with first derivatives only, where (3) is the Lagrange central equation (4). The mixed groups. The §3 and §4 constructions of the finite transformations from the infinitesimal ones (eqs. 17–19): the game works with infinitesimal transformations throughout, as the proofs do. §5 (the invariance of the parts of the relations, relative invariance). The nonlinear dependency between first integrals (p. 16: u’ = const and u’² = const). Einstein’s I* and the Klein notes (notes 21, 24). The caveat of note 14 (transformations with δ̄u = 0, Div(fΔx) = 0, which are split off). Four coordinates: the sheet is a surface traversed by one parameter; the diffeomorphism group of general relativity is played as the reparametrisation group of that one parameter, which is Noether’s own first example (p. 7) and the case n = 1 of “the group of all transformations of the x”.

15. Divergences


16. Mechanism kind and claim boundaries

Ported. The game computes Noether’s identities numerically: the Lie condition (eq. 11) on random paths decides whether a lever is a symmetry; the current B = A − f·Δx (eqs. 3, 4, 12, 13) is the tag; for a free-hand lever the identity of eq. 16 (σ = 0: Σ aᵢψᵢ = 0, which for the clock warp is Σψᵢuᵢ’ = 0, p. 7) is checked on random paths. The worlds are four integrals of the two types the paper names: the ordinary integral of mechanics (§1, eq. 4; the floor and the spring) and the Weierstrass parametric integral homogeneous of first degree in u’ (p. 7; the glass, which is Fermat’s principle with n depending on depth, and the sheet, a geodesic in a conformally flat metric with a well).

Does the discovery come from the paper or from the rules? The blank clock tag on the glass and the sheet is the vanishing of Σu’∂f/∂u’ − f for an integrand homogeneous of first degree (Euler’s theorem on homogeneous functions; Noether, p. 7 and §6), not a consequence of the budget, the Auditor’s test set or the thresholds. The calibration check (ROB) varies the grid, the bump and re-tick sizes, the glass and sheet profiles and the spring constant, and every verdict stays. The three-blank budget decides only how many tags a player can collect, never which tags are void.

17. Calibration plan and result

Against the paper’s own equations and statements, in calibration.js: eq. 3 on random paths and random δu; eq. 12 holding exactly for the levers that pass eq. 11 and failing for those that jam (the converse of §3, lever by lever); Theorem I’s counts (four laws on the floor, two on the spring) and the first integrals steady along true motions (p. 16); the clock’s B equal to ½ u’ ² + V (p. 28); invariance of the homogeneous worlds under x → p(x) for arbitrary p and the identity Σψᵢuᵢ’ = 0 (p. 7), both failing on the Newton worlds; §6, the clock relation improper (B ≡ 0) on the homogeneous worlds and proper on the others, and the shift being the warp with p ≡ 1; re-ticking; the law of refraction from the across-shift on the glass; robustness; thirty hand-drawn warps; the heuristics table; no decimals in the strings the screen reads.

Result: 21 checks, 21 passed (calibration-result.txt).

Heuristics (section 3 of build.md). The game has no score; the table records what each obvious way of pulling collects and which verdict it reaches.

Way of pulling Teeth Void Identity Jams Verdict
shifts only (clock, across, up everywhere) 5 2 0 5 You saw the blank tag
clock first, then turn, then across 7 2 0 3 You saw the blank tag
warps first 2 2 2 6 You answered Hilbert
reads the world (the intended path) 7 2 2 1 You answered Hilbert
one pull per world, never the clock on the glass 4 0 0 0 The press kept its secret

The pulls matter: the same twelve blanks reach three different verdicts, and the full verdict needs both the clock on a homogeneous world and the warp on the sheet.

18. Build log

  1. Mechanism headless in mechanism.js; first smoke test found two numerical faults, both fixed in the mechanism: the re-ticked motion was resampled by linear interpolation, which put kinks into ψ and made a true motion read as a fake on the glass (now re-integrated with a variable step); and the random test paths could reach u’ = 0, where the homogeneous integrand has a cusp (amplitudes bounded). Calibration 21/21.
  2. Placeholder loop (loop.js, loop-log.txt): the intended path through four worlds; every pull changes something visible, says whether it worked, and the glass clock is the turn.
  3. Skin: template, fonts embedded (Chakra Petch 700, Saira variable, Share Tech Mono), build script with the byte-for-byte check.
  4. Browser session (session-test.js, session-log.txt): two shifts, help opened and returned mid-world with the chosen lever kept, a drawn warp through the strip, about, replay; no errors, no horizontal scroll at 380 px. Fixed from the first run: an animation frame timestamp earlier than the start time (a NaN bead position), a stray call in the Auditor-first setup, the Auditor’s first-move stamp overflowing its box, and the tag slot showing its verdict colour before the Auditor had stamped.
  5. Self-review from screenshots: the aim is stated on the start screen in one sentence; the discovery is stated at the top of the end screen (the headline and the first lines); nothing in the table looks like dark steel with neon tube light. Builder’s review only; no stranger has played it. Session length is an estimate (about six minutes) until someone measures it.
  6. Revision 1 (1 Nov 2026, in chat, after the author’s first play: “spectacular, but hard to understand”). The diagnosis: the game’s own vocabulary (press, tag, blank, teeth, void, identity, jam, warp, fake, Auditor) was defined on the How to play page and used cold on the play screen, and the thing every world line talked about, the clock, was not on screen. Changes, all on the decision and result screens, none to the mechanism’s verdicts: (a) every trail carries its clock ticks as dots, so a shifted clock, a re-ticked clock and a fake are visible as such; (b) tapping a lever previews the change in the window (the changed motion in blue, the real one faint underneath, a caption saying what is shown), so the pull is a visual bet: would this world run just the same?; (c) the world lines state the rule the player needs (what the world never does) and name the dots; (d) the Auditor’s test is watched: five of his seven motions play in turn with a caption in the window (REAL MOTION / FAKE · how) and a line under the needle saying what the needle is doing and what that means; all seven are still read; (e) after a pull the game states the result in plain words first (the press held or jammed, what the tag measures, what the needle did, what that means), and only then the Auditor comments, in a line; (f) the aim and the start screen rewritten without the chassis words; (g) on the spring, the Auditor’s law is drawn (the fan the bead sweeps out about the peg, growing evenly); (h) the sheet’s first move and both bench lines reworded to say what to do without naming the lever; (i) the How to play page rewritten to match the screens, with the question the player is answering spelt out. Display-only numbers added to CONFIG (preview size, tick spacing, which motions are shown). Calibration 21/21 unchanged; full browser session clean.
  7. Revision 2 (1 Nov 2026, in chat, after the author’s second play: “looks good but only its surface; why a bead, what am I doing, where is the math; this looks like a physics experiment but these are mathematical thought experiments”). The diagnosis, after a round of discussion: the game had hidden the one object the paper is about, the formula that scores paths, and so the player saw a machine with lights rather than a theorem; and “bead”, “world”, “notice”, “teeth” were words nobody says. The rebuild keeps the levers, the tags, the Auditor and the look, and changes everything the player reads and does: (a) the formula is on screen, on a plate under each formula’s band, in symbols (x′ for the derivative; ψ for the Lagrange expressions where the rule is stamped) with each symbol said in words; this is the one Pantheon game that shows the symbols, because here the formula is the subject; (b) the vocabulary is the eight-year-old paragraph’s: a formula gives a path a number; the best path has the smallest number; “the formula does not care” replaces “the world does not notice”, “the drawing” replaces “the world”, “the dot” replaces “the bead”, “a prize” (a number you can work out from the dot at any moment, the same at every moment) replaces “a tag with teeth”, “empty” replaces “void”, “a rule” replaces “identity”, “it cares” replaces “jam”; (c) the player draws the paths: an opening in the first formula (draw from A to B with a finger, see its number next to the best path’s; up to three tries), and after every prize (draw any path; the needle reads the prize’s number along it; try to make it move), so the Auditor’s “fakes” are now paths the player drew, and the void prize is felt in the hand: nothing you draw moves the needle; (d) tapping a change shows the numbers before and after on the plate, for the best path and one other (the player’s own in the first formula, a wobbly path from the same A to B after), so the pull is a judgement about numbers seen; the equals/changed mark on the plate is the press’s own test, never the two displayed numbers; (e) the press no longer thuds and jams as a machine: it comes down, and the stamp says IT CARES or stamps the prize; (f) start screen, end screen and How to play rewritten in the new vocabulary, the start screen carrying the first formula. Mechanism changes, all display-side: score, strokeToPath, changedPath (a finite version of each change, the re-ticked clock kept monotone), numbers, plateScale; and one world constant: the spring now swings four tenths of a period in the window instead of three quarters, so that its true path is the path of smallest S and the plate’s “smallest number” stays true in every formula (the quarter-swing version made the true path a saddle). Calibration 21/21; full browser session clean, including the drawn paths.
  8. Revision 3 (1 Nov 2026, same day, after the author’s note that drawing is in vain where the best path is a straight line). Every formula now opens with the guess, not only the first: draw the path you think has the smallest number from A to B; the window shows only A and B until the guess is drawn, then the guess runs with its number, then the best path is revealed with its number and a one-line reason (the straight line at a steady pace; a swing round the peg; a bent ray through the thin glass; a route that keeps off the thick middle). The free dot is the warm-up where the guess is easy; the glass is where the reveal bites. The Auditor’s opening move, where he has one, now comes after the guess. The “wobbly path” on the plate is gone: the second path on the plate is the player’s own guess in every formula. One run is about a minute longer.
  9. Revision 4 (1 Nov 2026, after the author’s objection that the draw is pointless where the best path is a straight line: the game needs best paths that go down and come back up, or it is back to square one). Worlds 3 and 4 replaced, within the paper’s own integral type (Weierstrass, homogeneous of first degree in u′, p. 7), so that every verdict stays as calibrated and the best path is visibly non-monotone. World 3, the slide: f = √(x′² + y′²)/√(h − y), the brachistochrone’s time integral; from A near the top the best path dives below B and climbs back (a cycloid), and the straight line and every detour score higher. World 4, the sheet: f = (c − y)·√(x′² + y′²), the hanging chain’s integrand mirrored (thick below, thin above); the best path is an arch, and again beats the straight line and the detours. The sheet’s rotational symmetry is gone with the well, so its prize now comes from MOVE RIGHT (lean × thickness) rather than TURN; its role (empty clock prize, the free hand as the Auditor’s opening question) is unchanged. The release height h, the thickness constant c and each world’s A and launch direction are config numbers chosen so that the best path is the path of smallest S from its A to its B (checked against straight lines and six detours over many seeds in the build) and stays clear of the singular line y = h. A thin-centred and a mountain-centred sheet were tried first and rejected: with a mountain in the middle the shot geodesic curves into the mountain and is a saddle, not the minimum, so the straight line beat it; with a thin centre the deflection was monotone. Sections 11 and 13 above describe the earlier glass and well; this entry supersedes them. Calibration 21/21 (two checks reworded for the new worlds; the slide’s across-law is still −n·sin θ); full browser session clean.

Renamed (11 November 2026)

Renamed from Noether’s Ball to Throw It Past Noether, at the author’s decision to let some titles say what the player does while keeping the name. The game file is throw-it-past-noether.html; a one-line page at noethers-ball.html sends old links to it. Nothing in the mechanism changed.