Skip to content

Classify middle factors after an idempotent of any rank (no κ) - #7706

Open
aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-042
Open

aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-042

Conversation

@aliromman

@aliromman aliromman commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

All middle factors are now parameterized for an idempotent first factor of any rank over F2. For fixed A with A²=A, the ordered identity (I+xA)(I+xB)(I+x(I+A+B))=I+xI is equivalent to a construction from the A-invariant kernel of B and two maps on fixed complements. The proof gives an exact count and the full rank distribution at every dimension.

Read the proof, exact receipt, checker and source credit.

For dimension four, the totals at first-factor ranks 0–4 are 802,296,188,184,316. At rank two, 114 of the 188 solutions have both B and the third factor non-idempotent. The rank-one exclusion therefore does not extend. The named rank-one count 296 and its fingerprint are prior results; they are repeated only as a source comparison.

No new kappa. These are formal polynomial matrices, not admitted paid calls. The selected value stays 363005817/100000000000 with all five assumptions. Native realization, preparation, storage, ownership, restoration, the complete signed invoice and recursive consumer remain unresolved. The result does not rebuild a paired-star word.

Canonical base: e54523517d0f03e58352e647c403a5af3e525a57. Card joint-query-controller changes scope to this algebraic alphabet classification; the claim records why. Phrase search identified the already completed rank-one question in PR #7223 before work began. The new contribution is the all-rank kernel parameterization. Live Discussions, the closed cache and a fresh open-PR update delta were checked. The final full open-PR refresh timed out; the separate updated-order page covers every update after the retained snapshot. Sparse local notes and unpublished work remain overlap limits.

With CPython 3.12 and no third-party dependency:

python3 -B scripts/research_package.py --package research/idempotent-first-factor-kernel
python3 -I -O -W error -B research/idempotent-first-factor-kernel/verify.py

Expected direct output:

PASS_IDEMPOTENT_FIRST_FACTOR_KERNEL raw_matrix_cases=395316

Both registered modes pass. The 395316 raw matrix cases cover every middle matrix for all 14 canonical first factors through dimension four plus the named predecessor matrix. Direct polynomial multiplication agrees with a separate complete subspace-and-map construction, including each kernel, count stratum, rank distribution and idempotence class. The source fingerprint agrees exactly. Changed-C, non-idempotent-A, impossible-injection and wrong-kernel controls pass. The all-dimension conclusion comes from the written bijection, not extrapolation from this census.

AST and indentation checks, committed bytes and modes, source history, export input coverage, and an optimized sealed Git-free replay pass. The sealed source hashes and read-only modes stay unchanged. Validation details separate the evidence from unrun checks. No complete predecessor suite, selected cold replay, native, Lean or Docker replay, independent human review or formal-kernel check was run. Hosted CI did not certify this package. The plan job failed in the shared test_current_ci_section_matches_selected_record: the base guide names frontier6030 while the selected record requires make frontier7000-verify. The log reports 760 tests, one failure and 24 skipped; package replay was skipped. This package changes no shared CI file. Local replay remains the scientific evidence.

Credit @unpaidmoon for PR #7223 at 7a17378d02117c956ca76b8106f35b15aa34c0dc: its proof supplies the coefficient criterion and named rank-one classification. Credit the same contributor's PR #7040 at 2bac9bfafe5fa009bce2141c07b703c977b18456 for the earlier orthogonal and one-sided classification. Their notices disclose substantial Cursor Grok assistance and retain the earlier source lineage. Standard linear algebra supplies the subspace and complement counts. No predecessor code is copied. Prepared by Ali Romman (@aliromman) with substantial OpenAI Codex assistance, Apache-2.0.

Open next question: can one classified higher-rank factor receive a fully charged native implementation in an actual caller? Its missing implementation and consumer contract block any saving claim. No follow-up is reserved.

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