PulseCore

Chapter 9 · Section 6

Bringing the Game to the Theory

The work that brings the game closer to the theory goes in this order: the differences that change what a universe does first, then the fixed numbers that shape what is seen. Each step lands in this book before it lands in the code; its rows leave the registers of Parts 9.8 and 9.9, and its item leaves Part 9.5, when the code no longer holds it.

  1. The law and the measuring. One function for the Saturation Law in log form (Part 9.2), and the measured V for each heavy system (Part 9.3).
  2. The universe. The model level and the grid (Chapter 1); the body budget and the gravity mesh (Chapter 4); the ring list, with its dropping of the oldest well fixed first (Chapter 3); feedback and star lots (Chapter 5).
  3. The drawing. Proto Wells by the model field's own room, rows·cols/CL², not the screen's V (see "Built in the 8.4 trials" below); galaxy points by their share of the light; the cluster glows; the point budget that only falls; crystals and strings (Chapters 3, 6 and 8).
  4. Levels past 202. The level is typed as well as slid, and the large counts are carried as their logarithms so nothing overflows near level 510 (Chapter 1, Part 1.15).
  5. Records and panels (Part 9.9).

The rulings of Part 9.5 go to the author as they come up; none of them blocks the rest.

What the live 8.4 is (rebuilt 31 October 2026)

The 8.4 that is live at /game/ was rebuilt from 8.3's code on 31 October 2026, keeping five changes and nothing else: "drawn Proto Wells by the Saturation Law (8.3's room), rings keep every well, countdown loop, level box, dimmer waves; everything else 8.3". So the registers below still describe the live game, apart from these:

  • The drawn Proto Wells follow the Saturation Law in 8.3's room, a quarter of the field's cells (saturate(), protoDrawn() app.js:1610–1613; Part 3.3).
  • The ring list keeps every well; 8.3 dropped the oldest past max(4,000, cells/4) (rings.js:37).
  • The countdown loop: a returned universe continues into its full decay after 15 s, and a dissipated one plays its Ascendant Alpha after 15 s more (Chapter 10, Parts 10.12 and 10.14).
  • The level box takes any level; its slider covers 0 to 202 (Part 1.11). A level typed past about 500 is not yet guarded (Part 1.15).
  • The expansion only grows, and the waves and radiation are dimmer (drawing only).

capacity.js, the measuring of Part 9.3, is in the live folder but nothing loads it. Its device-measured V and everything below were built and run in the 8.4 trial builds, 8.4.1 to 8.4.9, kept at /game/versions/, and are not in the live 8.4. They are the ground the next version starts from.

What 8.5 / Claude / 1 is (built 4 November 2026)

/game/v8.5/claude/1/ is the live 8.4 brought to the rulings of 4 November 2026 and to the parts of this plan that change no look and no pacing. Its code is in public/game-v8.5-claude-1/; line numbers below are that folder's. It was built from the code outward, run once in a browser, and corrected once (the author: "you only have permission to watch it 1 time after you have built it and then you can make one last run of updates").

The rulings, in the code:

  • A Proto Well's size decides its fate (protoFailShare(), failProtoWells() app.js:1373–1400). No hash: each Proto Well's basin, every data cell nearest it, is counted, and the smallest basins fail first until the data they held reaches P⦰ of the scaffold. The Formation Condition cannot yet be applied as written, because the game's light runs half the field a flip and its G is set for the whole universe's collapse; the largest basin's 2GM/𝒞→² falls short of its size by many orders, and each run records how far as proto_fate.compactness_max. Until the game has a collapse test in its own units, P⦰ is the share the sky asks for, 0.069 (BPT Part 3.13), in a universe of four actual dimensions, and 2^(−D) in one of fewer. That share is a calibration, METHOD, named in the code. darkRatio() (app.js:1336) takes the same P⦰, (1 − P⦰) · 144/25 = 5.36.
  • Proto Wells merge along the scaffold before the big bang (mergeProtoWells() app.js:1699, called as they ride the data flow, flowSlice() app.js:4529): two whose data areas touch, nearer than the sum of their radii √(mass/π), become one. The merged map at the Physical Nova is kept (protoMap, filled in promoteDataWells() app.js:1401) and scored after the run (protoScore() app.js:1721): how many filled their sites, the rank correlation of a Proto Well's mass with its well's (Spearman), and the filled share of the largest and smallest quarters.
  • The grid's memory and the data's pull, measured (measureAfterBang() app.js:1736, called each physical step after the wells merge, app.js:678): at each hundredth of the life, the correlation of the matter's web with the scaffold's weight, per block of a 32 × 32 grid, and the pull of the sites on the matter, its mean strength and its coherence. They move nothing.
  • α follows the level (alphaLg(), matterLaw() app.js:4387–4392; ruling: charge is a bit). α(⚚) = α♁ × (2f)^(2·(⚚♁ − ⚚)), so stars stand where m_P/m_p > α^(−½), wherever matter is stable from about level 12 up to about 288.6 (starTop() app.js:950, from the same law), and atoms only above about 198.5, which the readout states and the game, drawing no atoms, does not simulate. 8.4's STAR_LINE = √137.036, which put the star line at level 22.2, is gone; the census's dead ends follow the band (no stars below level 12 or above 288.6).
  • Levels past 202. Every power of the level is carried as its logarithm (lg in describeUniverse(), universe.js:324; sci(x, digits, lg) universe.js:336), the work of the 8.4 trials brought over unchanged, so a typed level of any size reads as its value; the slider still covers 0 to 202, our level.
  • The notation. The readouts say ⚚ for the level and ♁ for ours, and the Pulse Tempo is shown as derived, ℨ × 2^⚚ × f^⚚ (renderUniverseReadout() app.js:2153); a new row gives α and which side of the band the universe stands. The real universe's panel shows ⚚♁ = 202 with the Nexus's reading, 200.7 to 201.9: the simulation keeps the whole level 202 for ours, since its harmonics count whole levels, and the Nexus's fractional level is shown, not run.
  • The record (runRecord()) gains alpha, proto_fate, proto_merges, proto_map, grid_memory and data_pull, and its matter state a fourth value, stars-no-chemistry.

The words and guards of items 3 and 4: the dive prompt says "one generation on"; "black holes" are Null Wells in the readout and the page's description; "Local Pixel Count" is the Local Cell Count; the Kerr–Kerr channel's range starts at the Kerr floor, 0.932, and the extreme channel's is 0.932 to 0.94 (universe.js:204–205); a child's outcome is drawn on its level, not on the heritage's step count (childOutcome(Math.round(levelOf(uni)), …)); wellCap(), which nothing called, is removed. The "Neutron star – black hole" channel keeps its name until the author names it.

The caps, brought in (10 November 2026, the author: "we were supposed to resolve everything with no caps and just caps that are created by math"). 8.5 / Claude / 1 now runs on the code of the 8.4.5 trial: the measured V of capacity.js and the model level by measurement, bodies and the gravity mesh by the Saturation Law, each simulated star one star (stars in lots by mass; STAR_DRAW's 40,000 is gone), points by the screen's room, the visual counts, swirls and reaches by law, arms with no top and the formula's pitch, the Nyquist turn, supernovae summed by cell, the IMF's lot bins, and saves by the server's sizes. Two trial changes stay out, the two that broke the look on 30 and 31 October: the Data Matter's own falling bodies (from 8.4.6) and the dilation over the whole life (the dilation is 8.3's, as in the live 8.4). The rulings of 4 November are applied on top as before. The author reviews it.

The bodies' room, the whole flip (11 November 2026, the author's lineage 842–847: "giant pink pixels instead of stars ... only 1 null well after the big bang"). The room for the bodies of matter was what a flip had left after the field, the "rest" of the physical universe and the bodies' fixed cost. On the author's desktop those three filled the flip: at a 256 field, 65 ms and 355 ms of a 390 ms flip; at 128, a rest of up to 600 ms and a fixed 276 ms against 781 ms. The room came to nothing, and every quantum of matter went out at the Physical Nova as one body: one well, no stars, and the torus's data left in view as large squares. The rest is kept on the device, so one heavy universe starved the next, and the lineage alternated good and bad. The field and the rest are already their own Saturation Law shares, and a fixed cost is paid once however many bodies there are, so neither is subtracted from the bodies' room: V = the flip's time / the measured cost of a body, and the bodies sent out are the Saturation Law of the real quanta in that V. In the same change, Start on a universe that has dissipated plays the child its end card offers (or a new universe where the line has ended) instead of restarting an empty clock.

8.5 / Claude / 1 is the Alpha, at /game (11 November 2026, the author: "set claude v8.5 as alpha at /game"). The game the site runs at /game is now 8.5 / Claude / 1, ranked Alpha on the versions page, the main line's newest step; 8.4 stays at /game/v8.4/, and the Gemini and Codex 8.5 builds and the 8.4 trial snapshots stay Zetas. What it carries beyond the live 8.4 is set out above: the 8.4.5 trial's measured capacity and Saturation Law in place of every cap, the rulings of 4 November, and the changes of 10 and 11 November that follow. A universe starts at level 202 by default, our own, and the level slider runs to 288 (the author: "set the default level to 202 and raise the ceiling to 288"): the Goldilocks level is the book's Band of Chemistry and Stars, 198.5 to 288.6, where atoms (α below 1) and stars (gravity weaker than the electric force between protons) both hold. The default had been 28, chosen on 28 October so that a lineage's children keep their stars; at 28 the fine-structure constant is about 10¹⁰⁵ and the game records its universes as stars without chemistry. The level box still takes any level.

Sprites kept to the room of their own size (10 November 2026, the author's Alpha child at generation 2: "large squares, not even pixelated"). Flashes, crystals, strings and the heavy bodies' disks were held to the room of a point, the screen's area over 36 square pixels, so a busy universe could put tens of thousands of 40-pixel sprites on screen (run 840: 40.6 ms a frame, 78 % busy). Each is now held to the room of its own drawn size, the screen's area over 40², about 1,300 on a 1080p screen, by the Saturation Law: the same law, the right unit.

Only the Alpha Well answers a tap (11 November 2026, the author: "I shouldn't see the magnifying glass over the visual ... allow the user to only click the alpha well, and if they click it, they get a 10 second popout top right"). The cursor over the field is the open hand that turns the view, the pointing finger only over Well #1. In a living universe, and through its decay, a tap on the Alpha Well shows its card for ten seconds at the top right: its generation and level, its mass and its share of the universe, whether it is Ascendant or how far it is from the Ascendant share (1/8.1 of the universe, the game's ASCENDANT_RATIO; the Ascendant Mass of Part 3.14), its child's level once the census has read it, and how and when it was born. After the return and before the decay a tap on any well still dives into its child, the way a child is chosen. A press paints the field only before a universe starts; in a living universe it turns the view.

The page while a universe runs (11 November 2026, the author). Start hides the site bar and the field takes its place; Pause brings it back, and the auto-rotate (a diagonal turn of the view, on by default since 10 November, 0.04 radians a second on its dial) holds still while paused. Full screen hides the bar too, and is left by a small "Exit full screen" pill. Every popout, the banners and prompts (the full decay's included), the rupture notice, the toasts and the Alpha Well's card, appears at the very top of the page over the site bar, in the toasts' small type and without forced capitals; the run's end card stays centred and fits the screen without a scroll bar. None of this changes what is computed; it is how the universe is shown.

Smooth at level 202 (11 November 2026, the author: "extremely choppy at level 202"). One run at the new default level, in a headless browser with its saves stopped, timed every part of a flip. The page stalled for half a second to a second and a half every few seconds through the life, and for four to six seconds at the big bang. Six causes, each fixed without changing what is computed (a check of the stars' step against the old one, 400 steps with mergers on a test universe, gave the same stars formed, living and dead, kind by kind):

  • The bodies' room. The morning's change counted only a body's own step, but the rest of the physical flip grows with the bodies too (the author's readings: 378 ms at 47,584 bodies, 1,120 ms at 153,219, the same 0.007 ms a body). The rest is now also read against the bodies, as a fixed part and a part per body, and V = (the flip − the fixed parts) / (a body's step + its share of the rest), the fixed parts taking at most half the flip, so V can never fall to the one body of the pink pixels.
  • Each body's stars. A body's stars are kept in lots by mass bin and birth flip, so a large body held two thousand lots by mid-life, and every step walked all of them for the few that swell or die. Each body now keeps its living lots in the order of their next moment and takes only those due; dead white dwarfs and neutron stars of one kind and bin share a lot (their number kept); the tallies are carried forward instead of recounted.
  • The colours of the stars. The scene sorted every body's lots by brightness, two powers per comparison, whenever its stars changed. It now keeps one of each look (kind, giant, living, mass bin), ordered by mass (brightness rises with it), and recolours a body at most every 0.4 s.
  • Mergers of wells. Each merger rebuilt the scene's whole well list and summed every body for its Ascendant test; at a level 202 big bang thousands of Proto Wells meet. The scene is now told once per pass.
  • Shaders compiled again and again. Each gravitational-wave ripple had its own material, disposed as it faded, so whenever none was showing the shader was freed and the next merger compiled it again; the star layers did the same when they grew. One ripple material is kept, and a layer keeps its material when it grows.
  • The Alpha hoop. Whether the torus hides the Alpha Well was a raycast through the whole torus every frame; it is now asked four times a second.

On the test machine (no graphics card, eight cores) the physical life ran 2.2 times as fast, the longest stalls through it fell from 0.5–1.4 s to 0.3–0.4 s, and the big bang's from 4–6 s to about 2 s, most of it compiling the shaders once; a desktop with a graphics card runs it well within that.

What the one watch showed (level 28, a desktop browser, four minutes, its saves stopped so nothing reached the database): no error from load to the expanding universe. The readout gave α ≈ 2.08 × 10¹⁰⁵ at level 28, below the band, and the record's matter state stars-no-chemistry; the Data Matter ratio 5.363. Before the big bang 980 Proto Wells became 979: the touch rule almost never fires, because the Proto Wells ride one data flow and seldom close on one another, so the merging the ruling expects needs a reason to bring them together that the game does not yet have. At the Physical Nova the size rule failed the 226 smallest, a fifth by count holding P⦰ of the data, and 753 sites waited for their wells; by a quarter of the life 583 stars shone and 8 wells had formed. The one correction after the watch: the View row printed (4·2^⚚)² as "10⁹²", its square run into the exponent.

Built in the 8.4 trials (8.4.1 to 8.4.9), not in the live 8.4

Fixed:

  • The ring list keeps every well; it dropped the oldest past max(4,000, cells/4).
  • Dilation runs over the life the run plays, 1 at the big bang to 2f at rupture; it never moved above the model level.
  • The Motion view divides by the speed of light in cells; it divided by a value nothing set.
  • Every pair of wells that meets merges in its flip, not one pair a flip.
  • Stars formed counts the stars made, and a dying lot's stellar wells count all n of them in the census.
  • The Pulsecore gas strikes the skin π/chord times a flip; the 8-bounce limit cut 3.9 % of paths.

Replaced by the Saturation Law with a measured V (capacity.js):

  • The model level and the grid. The model level is the whole level nearest to filling, the Pulsecore's flips to its Data Nova, in the game's own paced time, about 20 s: level n fills in about 128·8^n times the field's measured cost a cell, so n = log₈(20 s / (128 · cost)), rounded. Since the Pulse Mirror (2026-11-14) the fill is 2^(n+1) flips, a quarter of that; the 128 is kept as the field's room, so a device's level does not change with it (capacity.js modelLevel). The 6 and 4 by device kind and the 1,024 side are gone. A first version paired a body with each cell and ran a level coarser; a second set the level by the flip alone, which let a deeper level's longer life run eight times as long. Both were corrected the same day.
  • The bodies. Bodies = S(real quanta, the room the field leaves), where the 8,000 and 16,000 were.
  • The gravity mesh. It holds about one body a cell, where 64, 128 or 256 a side were set by device kind.
  • The drawn Proto Wells. S(law, the model field's own room, one well for each star-forming cloud it resolves, rows·cols/CL², 1,024 on any device). Level 28 draws about 46 % of its 846, level 40 about 5.5 %, level 202 under 0.01 %. A first version used the screen's room, about 29,000 on a desktop, against which every usual level drew nearly all.
  • The stars the matter forms. They are the law's whole count, each simulated star one star (starMult 1); stars are kept in lots by mass, so their cost does not grow with their number. They were at most 40,000.
  • The points drawn. The screen's room, a point for every 4 pixels, by the Saturation Law of the real stars; the ×0.8 cut is gone. A page cannot time its graphics card from the CPU's side, and a first version that tried drew a near-empty sky.
  • The galaxies' and streams' points. Each draws its share of the living stars and of the infalling matter, where 60 % and 120 an arm were, and 8 % and 42 a galaxy.
  • The visual counts. Flashes, accretion disks, ripples, crystals, strings, the eruption, the Big Bang's points, the fountain, the residue, and the harmonic rings, which now stand at (1 − 1/n) of the tube.
  • The Pulsecore gas. S(data within, cells/π), the points that strike each skin cell once a flip, since each strikes π times; its founders are the painted cells, where 16 were.

Replaced by a law:

  • Every clump that can turn its data once in the life swirls and glows. That is M ≥ (2π/T)²/G; before, the 24 heaviest.
  • Each swirl reaches to where its neighbor's pull takes over, d·√M/(√M + √M′), where 24 cells were.
  • Galaxy membership is searched over twice the wells' mean spacing, where a sixteenth of the field was.
  • A galaxy is drawn out to its own share of the space between it and its neighbors, where 0.45 was.
  • A galaxy's pattern turns at most π/m a frame (Nyquist), where 0.15 rad was.
  • Arms have no top, where 2 to 8 were; pitch is the formula's, where 5° to 45° were.
  • Supernovae in one cell are one push of their summed momentum, where 96 or 24 a step were.
  • Masses drawn a step are at most the IMF's lot bins, where 48 were.
  • Saves fill the server's own sizes, a summary of 200,000 bytes and a well_frames column of 65,535 bytes; browser lists fill the browser's storage. Panel lists show everything.

The Data Matter falls (Chapter 4, Part 4.10): its own cold, collisionless bodies from the Physical Nova carry its mass and motion; before, its mass was the data's lay, flattened and smoothed, and it never fell. What is seen stays 8.3's: a first version that lit the skin by its density looked wrong and was taken back.

Levels past 202:

  • The level is typed in a box with no top; the slider covers 0 to 202.
  • Every power of the level is also carried as its logarithm, so readouts show real values; checked to 10⁻⁶ against the computed values, and in short runs at levels 600, 1,500 and 5,000.

Reclassified, not changed:

  • The language reader's window, run length and sample sizes. The speaking margin, 0.05, was calibrated on that window, so these are METHOD with their calibration.
  • The pattern detector's windows. METHOD.
  • The render resolution of at most 2 device pixels a CSS pixel. DISPLAY: past it the fill cost grows by the square while the eye at a phone's distance gains little.
  • The cloud size's and the heat's reference level 6. These are calibration units, not caps.

Open, for a ruling:

  • The frame-drag profile. The inner ring is held to 1.5 times the outer, and the 1/ρ³ to 64; the physical bound would be the frame-dragging rate at a horizon.
  • The scaffold's 24 lines and 12 rings. They decide where the first data and matter land.
  • The device targets of 33 ms a frame and 1 s a flip (2 s on a phone).

Measured on this server (node, 30 October 2026):

  • A field cell costs about 0.0008 ms a flip.
  • A body costs about 0.026 ms a flip.
  • The rest of a physical flip takes about 40 ms.
  • At level 28, 8.4 against 8.3 on the same run:
8.3 8.4
Model level 6 6
Sites 811 → 726 803 → 738
Wells up to 70 up to 82
Bodies 8,010 12,321
Living stars 26,864 28,663
Each simulated star stands for 1.04 stars 1 star

Next: version 8.5 (planned 4 November 2026)

Version 8.5 is driven by the theory and by what writing this book from the code found. In order:

1. The rulings of 4 November 2026.

  • A Proto Well's size decides its fate. failProtoWells() drops its random hash: a Proto Well survives if its basin's data would pass the collapse test, 2GM/𝒞→² at least its size, by the end of the life, and fails otherwise. The run records the failed share of the scaffold's data beside the failed count; the sky's ratio asks for about 0.069 (BPT Part 3.13).
  • Proto Wells merge along the scaffold before the big bang, and the map of the merged Proto Wells is kept and scored after the run against where the physical wells formed: does Proto Well rank predict final-well rank and place?
  • The grid's memory. At each step after the big bang, the correlation of the matter's web with the original scaffold, giving the curve of how fast the gridlines fade.
  • The dark-energy hypothesis, measured. The net pull of the data wells on the physical matter, recorded each step: does it come out as a near-uniform outward acceleration? A measurement only; it changes no motion.
  • α follows the level (the author's ruling, charge is a bit, BPT Part 1.16). 8.4 holds the fine-structure constant fixed, STAR_LINE = Math.sqrt(137.036) in matterLaw() (app.js), as Part 2.10 once stated. 8.5 takes α(⚚) = α♁ × (2f)^(2·(⚚♁ − ⚚)) and gravity's coupling α_G(⚚) = (14.3 / 2^(⚚/3))², so a universe holds stars only where α_G < α, below about level 288.6, and atoms only where α < 1, above about level 198.5 (the Band of Chemistry and Stars). The readout says which side of each line a universe stands.
  • Levels past 202. The ladder has no top in the theory; a child climbs at most log₂ 8.1 ≈ 3 levels above its parent (Level from Share). 8.5 lets the level run past 202, with 202 marked as ours, and computes 2^⚚, 16 · 4^⚚ and the Planck values in logarithms, so nothing overflows (doubles fail past level 510 for the cell count and 1,023 for 2^⚚); this replaces the level-box guard of item 4.
  • The notation of the ruling. The readouts say ⚚ for the level and ♁ for ours, and derive the Planck time as ℨ × 2^⚚ × f^⚚ rather than taking it as given (BPT Part 1.6); the game's level for our universe is the Nexus's, read live, not a fixed 202 (BPT Part 1.16).

2. The work built in the 8.4 trials (listed above under "Built in the 8.4 trials"): the measured V of capacity.js, the model level by measurement, bodies and mesh by the Saturation Law, each simulated star one star, points by the screen's room, the Data Matter as falling bodies, swirls and reaches by law, arms with no top, the Nyquist turn, saves by the server's sizes. Each change to what is seen is checked against 8.3's look with before-and-after pictures of every harmonic.

3. The game's words brought to the rulings. The dive prompt's "one level deeper" (app.js:2921) becomes one generation on; "black holes" in the readout and the page description (app.js:2112, index.php:45) become Null Wells; "Local Pixel Count" (app.js:2096, index.php:302) becomes the Local Cell Count; the Kerr–Kerr channel's range stops at the Kerr floor, 0.932 (universe.js:204); the "Neutron star – black hole" channel's name waits on the author.

4. Guards and clean-up. A guard or a stated limit for the level box (a typed level of 500 overflows ℏ′, Part 1.15); the Kerr–Kerr default no longer overwritten on load; the child-outcome draw keyed on the generation or level, not the legacy step count; dead code cleared: applyVoid(), the paired orbits and stepOrbits(), wellCap(), the Drag import, the legacy grids, nullNovaTempo(), the frozen option.

5. This book follows. The code blocks are refreshed from the 8.5 folder with framework/tools/methodology-code/, and Chapter 10, the Master Script, is walked again from load to the next universe.