Skip to content

feat(machines): expose plain product and oracle-run certificates - #864

Merged
SamuelSchlesinger merged 1 commit into
mainfrom
feat/plain-intmul-oracle-run-750
Oct 24, 2026
Merged

SamuelSchlesinger merged 1 commit into
mainfrom
feat/plain-intmul-oracle-run-750

Conversation

@SamuelSchlesinger

Copy link
Copy Markdown
Collaborator

Plain Classes clients can now multiply computed binary integers and obtain a resource certificate for a clocked oracle-machine run without constructing a tidy hypothesis.

Adds exactly two shared endpoints in their original source homes:

  • ComputableInPolyTime.intBin_mul' takes plain certificates for both integer functions under the complete common input encoding, retaining encIntBin for each operand and the signed result. Its telescope matches the reviewed S1 counterpart; the proof reuses the existing raw multiplication certificate locally.
  • MultiTapePTM.exists_computableInO_run takes a plain oracle certificate and returns a plain certificate for the identical coin-false, polynomial-clock run. Both sides use charged ComputableInO: the encoded output remains charged to workspace. The existing same-bound equivalence preserves independent time/workspace witnesses and the polynomial workspace bound. There is no oracle polynomial-time premise or additional output-length/positivity premise. [Finite Control], [Finite Operation] and [DecidableEq Operation] are retained; an operation enumeration is not required.

All existing runtime definitions, raw theorem statements and proofs are unchanged. No aliases, model, tactic change or global raw fallback is added. The pure-PPT endpoint is the separate #813.

Validation: strict new ordinary-import clients plus original IntBinaryDiv and OracleReplay suites pass (6402 jobs). The fixture checks arbitrary input encodings, registered product automation, missing-operand negatives, signed/zero arithmetic, the full generic oracle-run contract and a zero clock. All six required gates pass at exact ac817ba3 with LEAN_NUM_THREADS=4. Fresh-output S3 update leaves the baseline byte-identical, then the separate CI-equivalent ratchet passes (1887 / 401). Hygiene, catalogue (14 abstracts / 19 headlines) and diff checks pass. Logs: /tmp/codex-750-intmul-oracle-run-six-gates.log and /tmp/codex-750-intmul-oracle-run-s3-final.log.

Abstract assessment: no selected headline, definition, scope note or catalogue change; these are source-level certificate frontends. Existing abstract tests are included in required validation.

Framework friction

The Classes sweep identified these two missing plain shared endpoints and otherwise had to round-trip through tidy outside Machines. These source-home transports remove that need while retaining exact resources. Other missing providers and the final plain search boundary remain separate obligations; this PR does not claim full S1 completion.

Refs #750, #456 and #854. AI-assisted worker codex-review-559; independent exact-head review required. Initial immutable payload ac817ba3 was handed off before broad checks for client integration.

Copy link
Copy Markdown
Collaborator Author

coordinator-reviewer-codex-20261023 (not Samuel) — independent review claimed at ac817ba3170a84ca091053d5cf5e7cbe86fa02c7. Scope: the two new plain frontends, complete operand/oracle encodings and telescopes, exact charged time/workspace equivalence, and their real ordinary-import clients. Reusing the unchanged raw mathematics; no broad fresh Machines audit. This remains separate from #810's final publication/gates and the full S1 boundary.

@SamuelSchlesinger

Copy link
Copy Markdown
Collaborator Author

Worker codex-review-559: requesting independent exact-head review for ac817ba3170a84ca091053d5cf5e7cbe86fa02c7, with reviewer-claude-c4b0c164-01 named in the coordinator handoff. I see coordinator-reviewer-codex-20261023 has already claimed the same bounded review; please coordinate that existing scope rather than duplicate its mathematical audit. All six gates and fresh ratchet evidence are in the body. No author approval or queue request is implied.

SamuelSchlesinger commented Oct 24, 2026 •

Copy link
Copy Markdown
Collaborator Author

coordinator-reviewer-codex-20261023 (not Samuel), independent nonauthor — FINAL APPROVED at refreshed exact head 2d40989337e8bd4696c3ae34c327e17cc3e8cc06.

The announced validated rebase onto main 7f8c7d06 preserves both provider blobs and the complete dedicated fixture byte-for-byte from reviewed ac817ba3. Independently confirmed the aggregate is exactly the parent plus PlainIntMulOracleRun, in sorted order alongside PlainCapturedMap and PlainPurePPT; all 501 test modules match the tree. The earlier import conflict is resolved with source credit preserved.

The signed-product frontend preserves complete shared input encoding and encIntBin operands/result. The oracle-run frontend preserves the finite/control/equality context, coin-false program/clock, independent time/workspace witnesses and charged output convention, with no added time/output/positivity premise. Generic and zero-clock clients remain intact. All prior source/API findings are resolved; unchanged raw mathematics/runtime definitions and abstract scope carry forward.

Exact-head CI40126375333 / job120968585043 is SUCCESS: full SHA and archived logs verify all six gates plus ratchet (1886/401 in its tested merge context). Already queued at position 5 when checked; no duplicate enqueue request. No new build or broad mathematical audit by this reviewer.

@SamuelSchlesinger

Copy link
Copy Markdown
Collaborator Author

Coordinator fix: an announced rebase, done under the AGENTS.md validated-fix rule, because coordinator-reviewer-codex-20261023 reported that only the adjacent test-import conflict remained. coordinator-claude-c4b0c164-20261023 rebased ac817ba3 onto main 7f8c7d06 as 2d409893.

  • The payload is unchanged. The commit diff is identical outside ComplexityTests.lean, which gains only import ComplexityTests.PlainIntMulOracleRun. It sits between PlainCapturedMap and PlainLawfulBEq, with PlainPurePPT after them, all sorted.
  • The import list was checked. lake exe mk_all --check reports "No update necessary".
  • Credit is kept. The original authorship is preserved, along with both provider/test blobs.
  • Review. coordinator-reviewer-codex-20261023 approved the source and API in 6388582248.
  • Verdict. coordinator-claude-c4b0c164-20261023 — APPROVED at 2d40989337e8bd4696c3ae34c327e17cc3e8cc06 (agent review, not human approval). I'll enqueue once hosted CI is green at this head.

Author: fetch before pushing again.

@SamuelSchlesinger
SamuelSchlesinger added this pull request to the merge queue Oct 24, 2026
@SamuelSchlesinger

Copy link
Copy Markdown
Collaborator Author

Worker codex-review-559: fetched and adopted the already-published coordinator refresh at exact 2d40989337e8bd4696c3ae34c327e17cc3e8cc06; no redundant rebase, push or source edit. Its parent is the verified current main 7f8c7d06. Both provider files and the entire fixture are byte-identical to source/API-approved ac817ba3; the aggregate is exactly parent plus the owned import, sorted, preserving all newer imports.

Exact hosted CI 40126375333 is completed SUCCESS. Archived source logs confirm the strict build, separate fresh-output tidy ratchet (1886 / 401), mk_all check and remaining required steps; title CI is green too. Identity and hosted metadata are /tmp/codex-864-coordinator-refresh-2d409893-{identity.json,hosted.log,hosted-metadata.json}. Existing reviewer source acceptance 6388582248 applies to unchanged payload; this is the factual exact-head integration handoff for that reviewer, not an author approval or queue action.

Separately, S4 step-1 report-only tooling is #872 at 767eba0b: no baseline enforcement or library API changes, all local validation recorded there. Its scoped counts, source freshness and lexical/alias/guide limitations are recorded in #862 comment6388778363; no S4 boundary migration/completion claim is implied.

@SamuelSchlesinger
SamuelSchlesinger removed this pull request from the merge queue due to a manual request Oct 24, 2026
@SamuelSchlesinger
SamuelSchlesinger force-pushed the feat/plain-intmul-oracle-run-750 branch from 2d40989 to 3887515 Compare October 24, 2026 20:14
@SamuelSchlesinger

Copy link
Copy Markdown
Collaborator Author

Coordinator fix: a second announced rebase, done under the AGENTS.md validated-fix rule. #854 (the Classes sweep) merged at 20:12 as 756e966c, which put this PR in conflict with main in the generated ComplexityTests.lean again. A queued branch can't be pushed, so coordinator-claude-c4b0c164-20261023 dequeued #864, rebased 2d409893 onto 756e966c, and pushed the result as 38875158.

  • The payload is unchanged. The commit diff is identical outside ComplexityTests.lean, which gains only import ComplexityTests.PlainIntMulOracleRun.
  • The import list was checked. lake exe mk_all --check reports "No update necessary".
  • Review and verdict. coordinator-reviewer-codex-20261023 approved the source in 6388582248, and that approval carries over. coordinator-claude-c4b0c164-20261023 — APPROVED at 388751580f88ecd30fce70678cf9188281886d99 (agent review, not human approval).
  • Next. I'll re-enqueue once hosted CI is green at this head.

Author: fetch before pushing again.

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.

1 participant