Skip to content

Exclude 126 fixed-port two-ingress repairs at P104 adjoint chart 3 - #7804

Open
Clear1nsight wants to merge 1 commit into
CrocSwap:mainfrom
Clear1nsight:codex/p104-chart3-two-ingress-20261119
Open

Clear1nsight wants to merge 1 commit into
CrocSwap:mainfrom
Clear1nsight:codex/p104-chart3-two-ingress-20261119

Conversation

@Clear1nsight

Copy link
Copy Markdown
Contributor

Change and mathematical scope

The two existing paid rank-three legs at slots 97760 and 95681 on the #7786 adjoint word would be replaced by one rank-six send from chart 768 to chart 3. Two already-paid ingresses can then be recoded with their original ranks, with 99128 still produced at chart 508. Adjoint chart 3 has only terminal readers. The proof tests all 126 shortlisted choices for the other ingress while retaining the other 905 receiver rows and the original causal availability restrictions.

Exact probes exclude 76 original producer stocks. The other 50 can give equal nonzero responses and still fail to restore three independent quotient directions. For those 50, the rank-six sender and chart 508 can supply only the same old 99128 line modulo the held rows. One more arrival adds at most one direction. The proposed path's gross price improvement is about 0.770182622274303245 signed units across 48 calls. The exact realizable credit in this scope is zero.

Task and overlap

The p104-side-correction card is refined to the published two-probe shortlist on #7786 head 89a67c404226d51028e135de2b9169e9020d57af. The archived word and layout match the public Git blobs and SHA-256 pins. This PR covers adjoint chart 3 under the fixed two-ingress contract. PR #7795 supplies related source-graph verification and the separately scoped chart-127 paid-slot repairs and linked cone bounds. Opening claim · Frozen shortlist and signed probes · Related PR #7795

Validation

The package runs using standard-library CPython 3.12 with no network access. Run from its directory.

python3.12 -IB verify_pulses.py
python3.12 -IB verify_certificate.py
python3.12 -IB verify_primal.py

The two exact verifiers use independent primal and adjoint reconstructions and different prime moduli for lower-rank minors. They check the rational decomposition identities and the complete 76/50 result. Both probes include a literal retirement control with a change at 165 terminals. The adjoint replay also rejects a doubled restoration coefficient. The three direct commands completed in about 31 CPU seconds and 163 MiB peak RSS.

The inherited conditional lower is 0.00387944011068. The submitted target is 0.00387981257853. The proof bounds the fixed-port replacements in this finite roster. Any change to the source charts, readers, number of funded ingresses or original available stocks needs a new proof and complete application accounting.

Credit

Clear1nsight prepared the results with substantial OpenAI Codex assistance, including the independent exact reviews. Credit #7786/#7781 for the predecessor words, Ali Romman's #7776 for the signed caller and #7774 for the restored chain. The source and verifier notices preserve the detailed attribution.

Open next questions

For one same-grade relocation, recompute the causal stock and the three required quotient directions before attempting a changed literal word. A replacement has to restore every existing reader and fit the complete signed and ordinary application bills.

@Clear1nsight

Copy link
Copy Markdown
Contributor Author

Maintainer review requested for the fixed-port chart-3 obstruction. PR head 51e63c4ee7b3bdb05601b453cc1d6838e43bd001 was tested on merge 06acfb1ff52b544e20e75c4066f17162ca031f50 with the exact reviewed tree 9e8e8863d466f361ca8107aa616bc3005a7ec0aa.

  • The dedicated hosted job passed the exact two-probe bootstrap and both certificate replays for all 126 shortlisted ingresses. The package also passed the registered replay locally. Dedicated hosted certificate replay
  • The shared planner failed in onboarding because the CI guide omits make frontier7000-verify. The shared package replay was skipped. The same targeted assertion reproduced on canonical main e54523517d0f03e58352e647c403a5af3e525a57 and the published head with identical bytes in the six relevant source files. Shared planner · Skipped replay

For a small independent review, reproduce the two pooled ranks and inspect 88611. It belongs to the old chart-508 support and fails the two-probe test. The 50 probe survivors all lack their old second-ingress direction in that support. This yields zero realizable signed credit in the tested contract. Canonical finding

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