The 8.6 Review: How a Universe Builds, Against the Sky

On 14 November 2026 the author asked for one run watched from start to finish, "and take notes of how the universe builds and where stars are etc, it needs to be a lot more visual accurate to the way the universe actually works". This Part records that run, what it showed beside the real universe, the causes found in the code, and what changed in v8.6 / Claude / 1 because of it.

The run

One universe at the default level, 202, played headless on the server (a software renderer, so about 5 to 12 frames a second against a desktop's 60, and the timings slower in the same proportion), its saves blocked so nothing reached the database. The model was level 5: a field 128 cells across, a life of 32 Tempos. Its state was read every five seconds (__pcState, app.js) and the screen photographed every ten.

Drawn: the bodies' points rose from 2,862 to 8,791; the galaxies' points stood at about 7,100 from the first moment to the last; the pulsars blinking rose from 12 to 310.

What the screen showed. The fill and the draw-in read as they should. From the big bang the young universe was a small bright torus inside the empty shell of space. By half its life the torus was dark, a thin scatter of white points on it, and what stood out were black discs with orange rings, the wells' horizons. At the final well the black discs and their rings covered the torus; the galaxies hardly read; no quasar ever blazed.

Against the sky

• Stars. The universe at our level holds about 10²² stars (book Part 2.10; the law's count at level 202 is 1.2 × 10²²). The run formed 1.26 × 10⁷, fifteen orders of magnitude short. The starlight skin is calibrated for the true counts (Part 8.7: a patch glows from a million stars to ten sextillion), so at about 700 stars a patch it stayed dark, and the torus looked empty.
• Black holes. A black hole gives no light of its own; the sky's are seen by what they do: the stars that orbit them, and, while they feed, as quasars and Seyferts (book Part 3.14). The heaviest holds a few thousandths of its galaxy's bulge (Kormendy & Ho 2013). The run's hearts held several hundredths and more of their galaxies, and their horizons were drawn several cells wide, black discs over the galaxies they sat in.
• Quasars. The sky's quasars are the bright phase of a central black hole's growth, near its Eddington rate, most common about a fifth of the way through the life, at the peak of star formation (cosmic noon; Madau & Dickinson 2014), a few percent of massive galaxies at any moment. The run had none: a heart of a galaxy's mass, fed a thousandth of its galaxy's star formation (book Part 3.14, the Eddington Ratio of a Well), stays far below a tenth of Eddington.
• Pulsars. About one neutron star in five sweeps its beam past us. The run drew them, one in five neutron-star lots, blinking, but at a flash of 0.05 against 0.009, too small to read across the torus.
• Galaxies and the web. The sky's galaxies assemble: many small ones early, merging into fewer and larger, spirals while they form stars and ellipticals after their mergers, strung along filaments with clusters at the knots and voids between. The run's galaxy points stood at the same number from the big bang to the final well. This is open (below).
• Smoothness. A sub-step a flip still owed, its matter and its stars together, held the screen for 213 ms on average in the run (matter.step 118, stars 58), the largest hold left.

Causes found, and what changed

• The Sun's floor (app.js:1372). sunCells = max(10⁻³, 0.77·Q/(D·0.39)): below about level 60 the floor held the game's Sun at a thousandth of a cell, so the stars a universe could form were capped near ten million. The floor is gone; the Sun is the law's. Star formation, winds, metals, supernova feedback and the stellar wells are all reckoned in mass (Parts 5.7–5.13), so only the counts change, and the drawing reads them: the starlight glows as Part 8.7 was written to, and the stellar wells come to about 10¹⁹ at level 202, as the sky's stellar black holes do. The first three Stellar Null Well notices are now counted by deaths, and a notice names one star's remnant, not a lot's.
• Every clump's heart is a Direct Collapse Seed (app.js:726). Book Part 3.14 makes a cloud collapsing whole a direct collapse, its well at most the seed law's, M_dc = Ṁ_in·t★ with Ṁ_in = 0.975·c_s³/G and never above a tenth of the gas. fillSite() held a clump to it only when the clump was born in a waiting site; a clump born anywhere else kept the whole clump as its own black hole. Now every clump's heart is the seed law's, with the warm gas's sound speed (7.4 km/s) where there is no halo to read: about 1.4 × 10⁵ Suns, the size of the direct-collapse seeds of the literature. The clump's whole mass still moves the well; the heart grows from its seed by its galaxy's gas (feedWellFromStars), and while it grows near its Eddington rate it is a quasar.
• A heart's horizon is its own (app.js:1447). wellList() drew a heart's horizon from the drawn mass that moves it, its galaxy's centre; it is now drawn from its own black hole, ownSuns. With the seed law, a heart's horizon is far smaller than a cell, and the heart is seen by its mark, its ring, and when it feeds its glow and jets (Part 8.5).
• Pulsars (scene3d.js:1193). The beam's flash is 0.11 against 0.012, for 15 % of each period: a pulsar reads across the torus.
• The stars on a tick of their own (app.js:5288). A sub-step a flip still owes now moves the matter on one tick and runs what follows it, the stars, the births and mergers of wells and the scene, on the next (pendingPost); nothing moves the matter in between (stepMatterTo waits). The physics and its order are unchanged.
• Earlier in 8.6, the same day: the skin, the Pulsecore's data and the starlight blended between steps (Part 8.9); the fill made about 32 s; the Data Nova's pour timed in seconds, so the draw-in reaches the pin head and the bang starts there (Part 8.4); no cloud of data through the full torus at the bang; the wells' sweep and the post-bang measurement on ticks of their own (app.js:5273).

The run after

A second run the same day, with these changes and the galaxies of Part 6.11, at the same level and settings: 5.6 × 10²¹ stars shining at the final well (5.8 × 10²¹ formed), against the law's 1.2 × 10²² and the first run's 7.9 million, the starlight filling the torus; quasars 5 just after the big bang, 64 to 82 between about 0.3 and 0.47 of the final size and 9 to 24 after, as the sky's quasar era peaks near a third of the present scale factor; the Seyferts rising to 64 at half the size; galaxies from 5 to 219, spirals through the middle of the life and lenticulars, dwarf spheroidals and the first ellipticals at its close; no errors. It showed two more things, changed the same day (Part 6.11): every quasar stood two jets, where about one in ten is radio-loud; and against the starlight the quiet hearts' black marks were blots.

Still open

• Galaxy assembly (the first run's finding): addressed by Part 6.11; the galaxies' merging is read from the wells' (Chapter 3), and its rate against the sky's is not yet measured.
• The quasar era is seen in the run after (above); its share of massive galaxies at a time against the sky's few percent is to be read from a record that reaches the database past the big bang (Part 9.3; the save script held records under 200 KB until 14 November 2026).
• The matter's own thread. The largest hold left is one sub-step's matter step. The worker that runs it off the page (MATTER_THREAD) has been off since 29 October 2026, when the stars starved with it on; its cause is not yet found.