Skip to content

spec(t27f): particle_field -- agent events as physical particles; sacred_lattice -- the formula as a codec with a chance label (Closes #10289) - #10290

Merged
gHashTag merged 1 commit into
masterfrom
claude/particle-field-10289
Oct 18, 2026
Merged

spec(t27f): particle_field -- agent events as physical particles; sacred_lattice -- the formula as a codec with a chance label (Closes #10289)#10290
gHashTag merged 1 commit into
masterfrom
claude/particle-field-10289

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 18, 2026

Copy link
Copy Markdown
Owner

Closes #10289

What

  • specs/t27f/particle_field.t27: the physical layer of T27F Live, between the work graph and the scene (docs/nona-02-organism/t27f/genui-research.md section 5).
    • Nodes (user, agent, tool) are particles. Edges are damped springs, and every pair of nodes repels.
    • Each agent event has a ternary charge (+1 toward the agent, -1 out of it, 0 inside) and a hue. It emits packets that fly along an edge and kick the receiving node. The events are: user message, think, tool call, tool result, agent message, error, done.
    • Fixed DT = 1/60 and seeded jitter, so a replay is bit-identical. Reduced motion keeps the packets and drops the shaking.
    • Output is sized for 64-vertex meshlets: 16 particles per meshlet, 32 primitives.
  • specs/physics/sacred_lattice.t27: V = n * 3^k * pi^m * phi^p * e^q as an address of five integers packed into one i64, with every fit measured against a null model.
    • At complexity 6 the lattice fits 62% of random numbers within 0.1%.
    • Seeded random targets: 42 of 64 fit. Real dimensionless constants: 6 of 10 fit. That is no excess over chance.
    • A fit is labelled SUGGESTED only if its chance times the number of targets tried is under 5%. None of the ten passes; m_p/m_e = 6 pi^5 (Lenz 1951) comes closest.

Checks (Railway lab, t27c test-report)

  • particle_field: 14 of 14 pass, 0 vacuous.
    • Verlet harmonic period within 1e-4.
    • Kepler energy drift under 1e-5 with Verlet; explicit Euler drifts over 1e-2.
    • Spring-repulsion balance matches Newton's method.
    • Bohr speed and radius match CODATA 2022 within 1e-9. The truncated hbar 1.054571817e-34 fails this, and a test pins that.
    • Replay is deterministic.
    • Energy after DONE falls under 1e-4 of its peak.
    • Reduced motion stays under 10% of normal motion.
  • sacred_lattice: 10 of 10 pass, 0 vacuous, 1 comptime invariant.
  • Planted faults, each caught: a flipped charge (2 tests fail), a halved density (3 fail), Euler in place of Verlet (1 fails).
  • tools/dupe_scan.py: no new duplicate body. Both files are ASCII.

🤖 Generated with Claude Code

@gHashTag @claude
spec(t27f): particle_field -- agent events as physical particles; sac…
a3d4744 
…red_lattice -- the formula as a codec with a chance label (Closes #10289)

particle_field.t27 is the physical layer between the T27F Live work graph
and the scene: nodes are particles, edges damped springs plus pairwise
repulsion, every agent event (user message, think, tool call and result,
agent message, error, done) has a ternary charge and a hue and kicks the
receiving node. Fixed DT = 1/60, symplectic steps, seeded LCG jitter, so a
replay is bit-identical; reduced motion turns kicks and jitter off. Output
is sized for 64-vertex meshlets. Integrator checks: harmonic period to
1e-4, Kepler energy drift under 1e-5 (Euler over 1e-2), spring-repulsion
balance against Newton, Bohr speed and radius from CODATA 2022 to 1e-9.

sacred_lattice.t27 keeps V = n * 3^k * pi^m * phi^p * e^q as an address
(five integers packed in one i64) and measures fits against a null model:
at complexity 6 the lattice fits 62% of random numbers within 0.1%, and
42 of 64 seeded random targets vs 6 of 10 real constants. No fit is
labelled above UNKNOWN unless its chance times the trials is under 5%.

Lab (Railway, t27c test-report): 14/14 and 10/10, 0 vacuous; three
planted faults (a flipped charge, a halved density, Euler for Verlet)
each caught.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Oct 18, 2026
Merged
Merged
Merged
Merged
Merged
@github-actions

github-actions Bot commented Oct 18, 2026

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-18 07:45:10 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 49
PRs with All Checks Green 1
READY 0
FAILING 49
PENDING 0
NO CHECKS YET 1

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 enabled auto-merge (squash) October 18, 2026 07:47
@gHashTag
gHashTag merged commit 3e3e2c7 into master Oct 18, 2026
28 of 30 checks passed
@github-actions github-actions Bot mentioned this pull request Oct 18, 2026
Merged
This was referenced Oct 18, 2026
Open
Closed
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_field -- agent events as physical particles; sacred_lattice -- the formula as a codec with a chance label

1 participant

@gHashTag