Land #70: give stars orbits their own potential holds, and clear the last galaxy.rs grandfathering #85

New issue

Closed

opened 2026-08-14 05:00:05 +00:00 by coilyco-ops · 0 comments

coilyco-ops commented

2026-08-14 05:00:05 +00:00

Member

Item 6 of #70, unblocked once #72's luminosity floor shipped. Also clears src/rust/galaxy.rs from #78's excludes, which is the same file and wants the same commit rather than a second churn through it.

What ships

Four changes, all measured on #70 and now promoted from ablation switches to per-scenario parameters:

parameter	disks	elliptical
`birth_orbit_ratio_cap`	1.06	1.06
`birth_velocity_dispersion`	0.3	1.5
`collapse_radiation_resist`	80 (was 20)	20
`STAR_FIELD_AXISYMMETRIC`	true	true

Every switch inverts cleanly to its own control, so each number stays re-measurable without an edit.

Measured, size 500, t=2500, two seeds

vsig above ~1.5 is a disk, below ~0.7 is a mush:

scenario	before	after
irregular => spiral	0.48 / 0.31	2.50 / 3.30
bang => spiral	-	2.49 / 2.75
irregular => elliptical	0.52 / 0.49 (target)	0.58 / 0.45

The disk holds to t=5000 - vsig 1.83, old cohort 1.51, no crossover at any checkpoint against a baseline of 0.11-0.44 - and the resolved population settles near 20k. #70 removed a bad sink and #72's floor added a good one, and they nearly cancel exactly as predicted.

bcirc now reads 1.06 rather than 1.69-3.21, so the cap binds and the probe reports it.

Three findings worth recording

The probe had drifted from the call site again. birth_circular_ratio mirrored the ablation switch but not the new scenario parameter, so it reported 1.93 while births were capped at 1.06 - the exact failure its own comment warns about, one release after the retraction that prompted the warning. Fixed, and the comment is now a two-line DRAGON on both sides.

The radiation gate had to become per-scenario. At 80 the elliptical's extra supernovae sweep gas into an annulus and it fails its own ring-signature check (0.293 against a 0.25 bar). Isolated by running each switch separately rather than assuming. The elliptical keeps 20.

Only one golden hash moved. Ring, bang-spiral, and irregular-spiral are byte-identical after 100 ticks - the elliptical is the only scenario forming stars that early, so it is the only one whose gas sees the change at that depth. Recaptured.

Verification

91 Rust tests, clippy, rustfmt, check-js, and the full precommit gate green.
48 e2e green against the new physics.
All four scenario-identity tests pass at size 250 with every existing threshold unchanged.

galaxy.rs comment grandfathering, cleared

138 blocks / 715 comment lines compressed to the two-line cap, with the reasoning relocated into seven new docs: stellar-heating.md, star-metrics.md, metric-calibration.md, integrator.md, seeding.md, stellar-population.md, boundary-ridge.md.

The guardrails stay put, just shorter - a dragon does not need three paragraphs to be flagged:

/// DRAGON: DO NOT TUNE AGAINST THIS - it scores an eccentric orbit at
/// pericenter as a good disk. Use vsig. docs/star-metrics.md.

The dead-cross-link hook caught two docs I had referenced but not yet written, which is the hook doing exactly its job.

pyproject.toml is down to four survivors, each now carrying its measured rationale inline: perf-rewrite.md, the seven skill docs, the four YAML entries, and pkg/**. That closes #78's fixable surface.

Still open, deliberately

The elliptical has no luminosity floor, so its population is unbounded - 56k at t=5000. Needs its retired light rendered first (#72).
Central over-concentration below size 250 (#70 item 4), parked by the option-1 decision.
#65's gas-confinement retune, which is a separate change with its own golden regeneration.

Item 6 of #70, unblocked once #72's luminosity floor shipped. Also clears `src/rust/galaxy.rs` from #78's excludes, which is the same file and wants the same commit rather than a second churn through it. ## What ships Four changes, all measured on #70 and now promoted from ablation switches to per-scenario parameters: | parameter | disks | elliptical | |---|---|---| | `birth_orbit_ratio_cap` | 1.06 | 1.06 | | `birth_velocity_dispersion` | 0.3 | **1.5** | | `collapse_radiation_resist` | **80** (was 20) | 20 | | `STAR_FIELD_AXISYMMETRIC` | true | true | Every switch inverts cleanly to its own control, so each number stays re-measurable without an edit. ## Measured, size 500, t=2500, two seeds `vsig` above ~1.5 is a disk, below ~0.7 is a mush: | scenario | before | after | |---|---|---| | irregular => spiral | 0.48 / 0.31 | **2.50 / 3.30** | | bang => spiral | - | 2.49 / 2.75 | | irregular => elliptical | 0.52 / 0.49 (target) | **0.58 / 0.45** | The disk holds to t=5000 - `vsig` 1.83, old cohort 1.51, no crossover at any checkpoint against a baseline of 0.11-0.44 - and the resolved population settles near 20k. #70 removed a bad sink and #72's floor added a good one, and they nearly cancel exactly as predicted. `bcirc` now reads 1.06 rather than 1.69-3.21, so the cap binds and the probe reports it. ## Three findings worth recording **The probe had drifted from the call site again.** `birth_circular_ratio` mirrored the ablation switch but not the new scenario parameter, so it reported 1.93 while births were capped at 1.06 - the exact failure its own comment warns about, one release after the retraction that prompted the warning. Fixed, and the comment is now a two-line DRAGON on both sides. **The radiation gate had to become per-scenario.** At 80 the elliptical's extra supernovae sweep gas into an annulus and it fails its own ring-signature check (0.293 against a 0.25 bar). Isolated by running each switch separately rather than assuming. The elliptical keeps 20. **Only one golden hash moved.** Ring, bang-spiral, and irregular-spiral are byte-identical after 100 ticks - the elliptical is the only scenario forming stars that early, so it is the only one whose gas sees the change at that depth. Recaptured. ## Verification - 91 Rust tests, clippy, rustfmt, `check-js`, and the full `precommit` gate green. - **48 e2e green** against the new physics. - All four scenario-identity tests pass at size 250 with every existing threshold unchanged. ## galaxy.rs comment grandfathering, cleared 138 blocks / 715 comment lines compressed to the two-line cap, with the reasoning relocated into seven new docs: `stellar-heating.md`, `star-metrics.md`, `metric-calibration.md`, `integrator.md`, `seeding.md`, `stellar-population.md`, `boundary-ridge.md`. The guardrails stay put, just shorter - a dragon does not need three paragraphs to be flagged: ```rust /// DRAGON: DO NOT TUNE AGAINST THIS - it scores an eccentric orbit at /// pericenter as a good disk. Use vsig. docs/star-metrics.md. ``` The dead-cross-link hook caught two docs I had referenced but not yet written, which is the hook doing exactly its job. `pyproject.toml` is down to four survivors, each now carrying its measured rationale inline: `perf-rewrite.md`, the seven skill docs, the four YAML entries, and `pkg/**`. That closes #78's fixable surface. ## Still open, deliberately - The elliptical has no luminosity floor, so its population is unbounded - 56k at t=5000. Needs its retired light rendered first (#72). - Central over-concentration below size 250 (#70 item 4), parked by the option-1 decision. - #65's gas-confinement retune, which is a separate change with its own golden regeneration.

coilyco-ops referenced this issue from a commit

2026-08-14 05:00:41 +00:00

feat(sim): give stars orbits the potential they read can hold

coilyco-ops closed this issue

2026-08-14 05:00:41 +00:00

coilyco-ops referenced this issue

2026-08-14 05:01:12 +00:00

Stars orbit a potential they were never given orbits for: one root cause behind #65 and #66 #70

coilyco-ops referenced this issue

2026-08-14 05:47:46 +00:00

main is red on a rustup download timeout, not on anything in the commit - run 150 #84