Skip to content

proofs(agda): Kernel group in Agda, reconciled with Lean (PWG-2i for 4 theorems) - #63

Merged
hyperpolymath merged 2 commits into
mainfrom
claude/new-session-up7579
Nov 16, 2026
Merged

hyperpolymath merged 2 commits into
mainfrom
claude/new-session-up7579

Conversation

@hyperpolymath

Copy link
Copy Markdown
Owner

First group of the Agda port: ROADMAP "Agda source of truth", owner decision 2026-11-09.

What changes

  • Agda source in proofs/agda/Marches/: Basic.agda (PathAlgebra, StrictlyInflationary, Inflationary, run, strict-implies-inflationary) and Kernel.agda (run-le, run-worsens, no-return).
    • A small local Prelude replaces the standard library. The import cone is the four files here plus Agda.Primitive, Agda.Builtin.Equality and Agda.Builtin.List.
  • Gate: scripts/check-agda.sh checks proofs/agda/All.agda cold under the PROOF-POLICY §2.1 invocation (--safe --without-K --cubical-compatible).
    • It scans the import cone for postulates with januskey's reference scanner, ported verbatim at b503488.
    • In the prover's own modules it allows only Agda.Primitive's four level primitives, each bound by its BUILTIN pragma.
    • It requires refusals: a command-line --safe control and two fixtures in tests/reject-agda.
      • FreewheelReturns shows the strictness hypothesis is load-bearing, as falsifier F2 does on the Lean side.
      • UsesK shows --without-K is in force on the command line.
    • It prints the receipt fields and the §8.5.1 inputs digest.
  • Setup: scripts/setup-agda.sh installs Agda 2.7.0.1, accepted only at echo-types' archive checksum.
  • CI: a new agda job in ci.yml. It uses only actions/checkout@v7.0.1, which the lock already records for this workflow, so actions.lock is unchanged.
  • Statement comparison: docs/agda-lean-correspondence.adoc, the §2.2 record. It was made by reading, with three residual assumptions: Agda's Set against Lean's Prop, equality without K, and the two trusted kernels.
  • PROOF-STATUS.adoc:
    • the Agda receipt;
    • the Lean re-check (no axioms for any of the four theorems);
    • the Kernel group RECONCILED, so its four theorems are PWG-2i;
    • census PWG-2i 4 and ungraded 207, of 211.
  • README, ROADMAP and STATE are updated to match.

Evidence (local, at 660e1c6)

  • scripts/check-agda.sh: [check-agda] OK, inputs-sha256 4f589dfe…1567. The policy's shell line reproduces the digest.
  • Lean v4.35.0, archive checksum OK:
    • lake build completed (19 jobs);
    • no sorry;
    • the BadGadget orbit is unchanged;
    • #print axioms reports "does not depend on any axioms" for all four theorems.

Still open

  • The Rocq leg (PWG-2ii).
  • The machine-readable proof receipt (§8.5.1 praxis deed), so no grade here is PWG-1.
  • The next group, Compose.

Generated by Claude Code

The estate proof policy makes Agda the source of truth for theory. This
adds the Agda source for the Kernel group: proofs/agda/Marches/Basic.agda
and Kernel.agda, with a small local Prelude. There is no standard library,
so the import cone is this repository plus Agda.Primitive and two
Agda.Builtin modules.

scripts/check-agda.sh checks proofs/agda/All.agda cold, under the §2.1
canonical invocation. It scans the import cone for postulates with
januskey's reference scanner, ported verbatim. Two fixtures must be
refused, and a runtime control checks that --safe on the command line
refuses a postulate. The script prints the receipt fields.
scripts/setup-agda.sh installs Agda 2.7.0.1 at echo-types' checksum. A new
`agda` job in ci.yml runs both. It uses only actions/checkout@v7.0.1, which
the lock already records for this workflow.

Agda-source session (estate proof policy), Kernel group.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VTeVWDEJgdYKsUNr4uhnXX
…theorems

docs/agda-lean-correspondence.adoc is the PROOF-POLICY §2.2 comparison
for the Kernel group. It was made by reading: definitions field by field,
theorems binder by binder. Three residual assumptions are listed: Agda's
Set against Lean's Prop, equality without K, and the two trusted kernels.

PROOF-STATUS.adoc records:

- the Agda leg receipt at 660e1c6: Agda 2.7.0.1 at its checksum, the
  canonical invocation, the import cone, the inputs manifest and digest
  4f589df…, and no axioms;
- a Lean re-check of the four theorems at the same commit, which reports
  no axioms;
- the Kernel group as RECONCILED, so its four theorems move to PWG-2i. The
  census is now PWG-2i 4 and ungraded 207, of 211.

The machine-readable receipt (§8.5.1) and the Rocq leg stay OPEN.
README, ROADMAP and STATE are updated to match.

Agda-source session (estate proof policy), Kernel group.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VTeVWDEJgdYKsUNr4uhnXX
Repository owner deleted a comment from coderabbitai Bot Nov 16, 2026
@hyperpolymath
hyperpolymath marked this pull request as ready for review November 16, 2026 19:35
@chatgpt-codex-connector

Copy link
Copy Markdown
Codex Review Summary: ❌ Failed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ❌ Failed 2026-11-16T19:35:48.406438Z
We couldn’t complete this review. Comment @codex review to try again.
e3cf2ad Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@hyperpolymath
hyperpolymath merged commit 08f6a79 into main Nov 16, 2026
15 checks passed
@chatgpt-codex-connector

Copy link
Copy Markdown

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

@hyperpolymath
hyperpolymath deleted the claude/new-session-up7579 branch November 16, 2026 19:35
@coderabbitai

coderabbitai Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

Explain Change →

Caution

Review failed

The pull request is closed.

  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants