Skip to content

spec(t27f): particle_engine -- a particle engine on real physics, controlled through forces, every visual derived from state - #10310

Merged
gHashTag merged 1 commit into
masterfrom
claude/particle-engine-10309
Oct 18, 2026
Merged

spec(t27f): particle_engine -- a particle engine on real physics, controlled through forces, every visual derived from state#10310
gHashTag merged 1 commit into
masterfrom
claude/particle-engine-10309

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 18, 2026

Copy link
Copy Markdown
Owner

Closes #10309

What

specs/t27f/particle_engine.t27, a particle engine on real physics from which the visual quantities are derived.

  • Bodies: sixteen, each with an SI mass, charge and radius.
  • Forces:
    • standard gravity, 9.80665 m/s^2;
    • Stokes drag;
    • softened Coulomb pair forces;
    • Hertz contact, F = K d^1.5, with a Cundall-Strack dashpot;
    • a restituting floor and walls.
  • Integrator: velocity Verlet, eight substeps of 1/480 s per 1/60 s frame.
  • Control: agent events become commands to a critically damped PD controller (KD^2 = 4 KP, a comptime invariant). Control enters only through forces; no position is ever set by hand.
  • Visuals: the glow, the lacquer reflection in the floor (Schlick Fresnel, n = 1.5) and the motion streaks are all functions of position, velocity and charge. They are rasterised into a 50-op display list and a 108x192 pixel grid, and both are read by the tests.

Contact models tried

Each was measured on the lab before choosing:

Model Result
Lennard-Jones Diverged: fast bodies reach the r^-12 core.
Linear spring Lost 10-20% of the energy in an elastic bounce.
Hertz, K = 1e5, 8 substeps Loses 0.2-0.5%. This is the model used.

Checks (Railway lab, t27c test-report)

21 of 21 pass, 0 vacuous:

  • Integrator against exact results:
    • Free fall and drag are exact to 1e-12.
    • Terminal speed is within 1e-6.
    • Bounce height is e^2 within 0.01.
  • Conservation:
    • Momentum is conserved to 1e-12.
    • Energy drift is under 1e-3 with contact and under 1e-4 with Coulomb only.
    • A head-on elastic collision keeps its kinetic energy within 0.5%.
  • Equilibrium: a +1/-1 dimer settles at the root that Newton's method finds, within 1e-6.
  • Control: the PD controller does not overshoot.
  • Agent events:
    • An error burst is faster than anything the conversation does.
    • After DONE, every body lies within 2 mm of the floor and the kinetic energy is under 1e-4 J.
  • Optics: Fresnel R0 = 4% and the 60-degree point.
  • Rendering: a true black background, the reduced-motion streaks, and a deterministic frame hash.

Planted faults: each of the eight was caught.

Fault Caught by
Schlick exponent 4 1 test
Euler instead of Verlet 5 tests
Pair force without its reaction 7 tests
Coulomb sign flipped 3 tests
Mirror y sign flipped 1 test
KD = 4, underdamped the comptime invariant, which refuses to compile
Rest threshold that never absorbs 1 test
Exposure +2 1 test

Other checks:

  • tools/dupe_scan.py --like: nothing in the file is written elsewhere.
  • The file has 0 bytes above 0x7F.

Not in this PR

🤖 Generated with Claude Code

@gHashTag @claude
spec(t27f): particle_engine -- a particle engine on real physics, con…
8dc1cf0 
…trolled through forces, every visual derived from state (Closes #10309)

Sixteen SI bodies under gravity, Stokes drag, softened Coulomb, Hertz
contact with a Cundall-Strack dashpot, floor and walls; velocity Verlet at
8 x 1/480 s per frame. Agent events drive a critically damped PD controller.
Glow, Schlick floor reflection and motion streaks are computed from state
into a display list and a pixel grid. 21/21 tests, 0 vacuous; eight planted
faults caught.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions github-actions Bot mentioned this pull request Oct 18, 2026
Closed
@gHashTag
gHashTag enabled auto-merge (squash) October 18, 2026 08:17
@github-actions github-actions Bot mentioned this pull request Oct 18, 2026
Merged
@github-actions

github-actions Bot commented Oct 18, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-18 08:20:25 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 48
PRs with All Checks Green 2
READY 0
FAILING 48
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 48 + 0 + 0 = 48, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=0f8c66b9ed0e != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@gHashTag
gHashTag merged commit 9da2c5f into master Oct 18, 2026
29 of 31 checks passed
This was referenced Oct 18, 2026
Closed
Merged
gHashTag added a commit that referenced this pull request Oct 18, 2026
@gHashTag @claude
dupe-ledger: bless log_frame copied by particle_engine (Closes #10315) (
34a4316 
#10316)

master's duplicate-bodies check fails at 4414941: "a body written
under log_frame is now copied [2] time(s) and was in no ledger".
#10310 added specs/t27f/particle_engine.t27, whose log_frame has the same
body as specs/t27f/particle_field.t27's. #10312 blessed the ledger before
#10310 merged, so the copy was in no ledger.

The one new line, `log_frame 2`, is `python3 tools/dupe_scan.py --bless`
output, run on the Railway lab at master 4414941. The file's sha256 here
(8f2300b5) matches the lab's. No other group changed: 188 groups and 662
bodies, as before.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
@gHashTag gHashTag mentioned this pull request Oct 18, 2026
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

spec(t27f): particle_engine -- a particle engine on real physics, controlled through forces, every visual derived from state

1 participant

@gHashTag