Skip to content

Independently verify #7544’s conditional κ gain and owned1557 erasure boundary - #7556

Open
Clear1nsight wants to merge 3 commits into
CrocSwap:mainfrom
Clear1nsight:review-7544-boundary-20261114
Open

Clear1nsight wants to merge 3 commits into
CrocSwap:mainfrom
Clear1nsight:review-7544-boundary-20261114

Conversation

@Clear1nsight

Copy link
Copy Markdown
Contributor

Change and mathematical scope

This independently verifies the finite conditional result in #7544 at d6be7d620c9c8337104daeb09fe01196eced5fe3: κ 0.003647969120276, exact gain 0.000000888080587 over #7481. Construction credit stays with ToikY, DreamingOfClouds and the retained lineage. This review supplies no new numerical increase beyond #7544 and changes no selected frontier.

The separate rational checker compares all 74 owned values immediately before CLEAR. Exactly 73 agree with #7481. The one changed value, owned1557 at chart595, differs by −3/2(e902+e985). It checks the numerical effect of all five subsequent reads: each target row is restored immediately after its ADD. Both source words already have 73 nonzero erasure operands. The complete public map, COPY guard, ordered owned lifecycle/MOVE stream, grid, mass, actual fixed17caller invoice and both directed prices are checked with separate implementations.

The complete source replay exposed a runtime portability failure: its original 595-second aggregate command timed out after the native kernel passed. This failed run remains in evidence/. The review factors the unchanged native and signed kernels into separate required cold commands. Each uses a new isolated source export. The fresh native mathematical receipt must equal the authenticated frozen receipt consumed by signed verification; all three commands are required and emit the same source/receipt binding. Only the source’s exact top-level timing omission lists are used. Actual fresh/frozen receipt hashes and all adapter records are retained. Signed alone is explicitly insufficient.

The five inherited premises remain native_all_columns, owned_erasure, compiler, precision_tape, and analytic_recovery. In particular,1557’s operand is changed: the claim explicitly retains bounded arbitrary-content erasure. This is not an equal-operand restoration, compiled multiplier execution, independent human/formal-kernel approval, or maintainer admission.

Task and overlap

Base: canonical main e54523517d0f03e58352e647c403a5af3e525a57, priority revision8, task p104-side-correction, scope key 4045f5fea37856e1d039.

Scope changed by explicit donor choice to verification of #7544 and a separate review PR after that composition was published. No further numerical search or construction modification occurs. All-states PR7544 search, central review ledger and refreshed #7111/#1166 coordination were checked. Newer #7546, its role902 construction and other active/paused scopes remain outside this review.

The new audit reuses the unmodified generic dense run() and require() dependency from #7164 at 190cf7eadc8dd8002327bec4b63d1db7f978a0e9. Its historical fixed-size check() and fixed-clock omission branches are unused. The exact74file #7544 export and actual transitive replay inputs are committed for Git-free cold availability; SOURCE-MAP.json and ORACLES.json bind every imported source byte. This is actual replay input, not an attribution-only copy.

Two existing CI prerequisites are kept as separate commits: selected-replay navigation and source-pinned historical replay. Their patches are reused from the prior #7147/#7257 lineage; neither changes the selected result.

Validation

CPython3.12, NumPy1.26.4, assertions enabled. The supported registered invocation is:

python3.12 -B scripts/research_package.py --package research/p104-7544-independent-review

Required cold commands: verify.py --stage native (480s), --stage signed (495s), --stage audit (225s), with preparation and descendants included in each cap. Total cap1200s. Exact observed results and receipt binding are in EXECUTION.md.

All three cold commands passed: native351.326s, signed/final402.072s, independent audit60.936s; total814.334s. All emitted binding 3e9e4b83f85d4bc1a59d7f846dbfe826db414b8d83dd1d2c94b1aa81167ce4bc. The separate metadata/pin check, six selected-navigation tests, committed-byte/mode/source-history preflight and committed export/offline integrity check also passed. Publication head: e04613f0cdd715e4d6f44687234259d1b26368cc. Hosted checks will be reported against that exact head.

The source’s original exact command passed79.2659s; its original complete command remains a595s timeout. An initial independent-control selector failure is also retained: the994instruction occurs twice, while #7542’s original list.remove control selects the first. The repaired control binds that same occurrence40962 by its exact unique neighbouring instructions. All seven retained corrupt-word controls, dirty COPY, insufficient odd/dyadic grid and omitted-paid-job control must reject.

The reused full kernels check the3039map catalogue and absolute phases,104main boundary pairs, complete signed caller, ordinary bound782757964, precision coefficient2430, grid/mass, tape/stock/compiler,54strict+4functional outer inequalities and the fixed-profile ceiling. These are finite conditional checks, not a proof of the five premises. No selected record was edited.

Credit

ToikY/#7544 supplies the composition and complete application; DreamingOfClouds/#7542 supplies the helper rung merges (with disclosed Claude Code assistance); Clear1nsight/#7481 supplies the funded helper485 return; #7480/#7334 and earlier authors supply the native lineage; Rohan Arun/#7113 supplies the signed caller; ToikY/#7298 supplies the retained directed tariff. The dense oracle is earlier Clear1nsight work, not a newly invented separate author. All nested source notices are unchanged.

Prepared for Clear1nsight with substantial OpenAI Codex and verified GPT-6 Pro5/5 assistance. The completed Pro consultation inspected the precise factoring/consumer contract; it did not execute kernels or confer independent human approval. See NOTICE and PRO-CONSULTATION.md.

Open next questions

The smallest next step is maintainer assessment of this exact conditional application, especially the explicit owned-erasure reliance and required three-command receipt join. The five framework premises and maintainer selection remain outside this review’s finite verification. Stronger #7546 requires its own exact-source review. See TASK.json and the canonical finding and handoff.

Clear1nsight and others added 3 commits November 14, 2026 10:42
This is the exact documentation prerequisite retained by CrocSwap#7156. It points the onboarding check to the selected CrocSwap#7000/CrocSwap#7001 review. Credit to Rohan Arun and the source authors.

(cherry picked from commit bb36fc6)
The PR CrocSwap#6292 rejection census and package plan are statements about 2485805. Load its exact sources and planner in a private sparse checkout, keeping all source hashes, 19 expected rejections, nine package selections and the empty-plan control. Subsequent edits to the live Makefile and CI guide no longer invalidate this historical check.

The original audit is by unpaidmoon, building on Dennis’s historical replay gate. The pinned repository view follows the precedent of Matarener’s PR CrocSwap#5434. Repair prepared with substantial OpenAI Codex assistance.

(cherry picked from commit d2e54af)
@Clear1nsight

Copy link
Copy Markdown
Contributor Author

Exact-head delivery verified at e04613f0cdd715e4d6f44687234259d1b26368cc.

All hosted checks passed, run 43532990548, attempt1: plan132255019184, historical replay132256894245, and scientific replay132256894208. The scientific log contains all three cold-stage PASS markers and PASS_REGISTERED_RESEARCH_PACKAGE. Observed hosted stage times: native365.451s, signed/final414.989s, independent audit53.672s. Their receipt binding is exactly 3e9e4b83f85d4bc1a59d7f846dbfe826db414b8d83dd1d2c94b1aa81167ce4bc, matching local execution. Fresh/consumed native mathematical projection hashes match, and the cold audit receipt SHA256 is 94749e63e0c7ffc4c0dc2ae9591d77bb70c5cadb97d0552844cfd2a3f1fe765b.

This completes the PR-specific finite verification of exact #7544's conditional κ0.003647969120276 gain and owned1557 boundary under the documented factored replay contract. The original595s aggregate timeout and initial selector failure remain failed and preserved. All five framework premises remain; no new numerical gain beyond #7544, independent human/formal acceptance or selected-frontier admission is claimed. The PR is Ready for review, open/unmerged, with maintainer edits disabled.

Final canonical result and released review scope · Concise hub pointer.

The next step is maintainer assessment of this exact conditional proposal and its inherited erasure contract. Construction credit remains with ToikY/#7544, DreamingOfClouds/#7542, #7481 and the retained native/caller/tariff authors; substantial Codex/Pro assistance is disclosed. No further research or selected-record change is undertaken.

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