Skip to content

Show the section-2 record inventory fits constant 1 at the two p=2 CRT corners - #100

Draft
unpaidmoon wants to merge 2661 commits into
mainfrom
cursor/crt-p2-record-inventory-cbe1
Draft

unpaidmoon wants to merge 2661 commits into
mainfrom
cursor/crt-p2-record-inventory-cbe1

Conversation

@unpaidmoon

Copy link
Copy Markdown
Owner

Change and mathematical scope

Finite certificate. No new κ, and the selected frontier stays 605457647/200000000000.

For the full shared-arena CRT batch, replace one schoolbook product per tape action by the PR CrocSwap#4192 section-2 record comparisons. Charge each of the 4AJ key-field comparisons, and each of the V = N^2 - N - Nh/2 pairwise comparisons, at the same cap 8w^2. Keep the inherited tape, rotation and CRT charges. At both excluded corners the resulting batch is strictly below E = 2^{4 ell}(p+2)^{16}:

  • (2,1,1): B = 822078336 < E = 68719476736
  • (2,2,2): B = 12370604416 < E = 1099511627776

The inherited majorant stays above E there (75548846976 and 1241895956864). The new scalar term is strictly smaller than that majorant on the whole admissible domain, so the predecessor classification lifts the fit to every integer p >= 2 and 1 <= h <= ell <= p. One schoolbook cap per arena cell per charged access still fits the two corners. These are explicit majorant comparisons, not a machine lower bound, a supplier admission, or a multiplication theorem.

This branch is based on upstream CrocSwap main 32890bc5de034ab8fc93ac88ead7279a6a416d04. The reviewable change is only research/crt-p2-corner-record-inventory/.

Task and overlap

Task uniform-transfer-audit, priority revision 7, scope refined. Base commit 32890bc5de034ab8fc93ac88ead7279a6a416d04. Live inputs are research/crt-arena-root-axis/PROOF.md, research/crt-arena-batch-threshold/PROOF.md, and the vendored PR CrocSwap#4192 proof. The self-contained object is the two corners; the all-domain sentence cites the predecessor classification and does not re-prove it.

Nearest work: Nick Petalas's research/crt-arena-batch-threshold leaves this as open question 1. PR CrocSwap#4650 fits one axis and leaves the batch open. PR CrocSwap#4377 supplies the CRT and rotation tariffs. An overlap search on 2026-10-25T07:21Z covered 398 open pull requests, 4819 discussion records and 7404 local notes. No open pull request or discussion states this section-2 corner inventory. Closed pull-request bodies were searched only through local notes.

Validation

CPython 3.12, standard library, from the repository root:

python3 -I -W error -B research/crt-p2-corner-record-inventory/verify.py
python3 -I -O -W error -B research/crt-p2-corner-record-inventory/verify.py
python3 -B scripts/research_package.py --package research/crt-p2-corner-record-inventory
python3 -B scripts/check_research_commit.py --package research/crt-p2-corner-record-inventory --check-source-history

Observed on this commit: PASS_CRT_P2_CORNER_RECORD_INVENTORY, PASS_REGISTERED_RESEARCH_PACKAGE research/crt-p2-corner-record-inventory outcome=completed, and PASS_RESEARCH_COMMIT_BYTES at c1e15939664b47cfb92748ed2cbd34ca0817df8c.

The corner identities are exact arithmetic. The 45,759-triple box corroborates T_record < T_scalar; the written comparison has no p <= 64 cap. Not run: Lean, Docker, the Beneš simulator, and native replay.

Credit

Prepared by unpaidmoon with substantial Cursor Grok assistance.

Nick Petalas proved the inherited full-batch threshold and left these two corners open. The one-axis bill and its schoolbook cap are unpaidmoon's PR CrocSwap#4650. The CRT and rotation tariffs are unpaidmoon's PR CrocSwap#4377. ToikY's PR CrocSwap#4192 supplies the arena census, the four-field scan and the pairwise count. Cinna Davis supplies the shape of E in PR CrocSwap#4100. Predecessor files are unchanged. New work is Apache-2.0. Independent human and formal proof-kernel review are unclaimed.

Open next questions

  1. Compose this allowance with reservoir placement and the paid repair-key service of PR Execute paid CRT key service at the actual repair port CrocSwap/integer-mult-bounds#4795. Begin with one complete entry, call and exit program, retaining every source-state query and restoration charge. The corner fit does not answer that composition.
  2. Emit a literal scalar transition table for one four-field scan and price those steps at (2,2,2), keeping T_tape. The schoolbook-cap margin does not substitute for that table.

Workflow feedback (optional)

CrocSwap#1200

Fork main has diverged from upstream. Please review the package directory, or diff this head against 32890bc5de034ab8fc93ac88ead7279a6a416d04.

Open in Web Open in Cursor 

1337isnot1337 and others added 30 commits October 23, 2026 08:48
The level-1 allowance in the bit-stock first step fails by at least q^4096, while the geometric ancestor chain plus parameter bits stays within 26041 V for every stream of volume at least e.

Co-authored-by: unpaidmoon <unpaidmoon@users.noreply.github.com>
The exact count on the selected nested-width slab explains the published
nonmonotone drop. Substituting it does not reach 2944/1000 or 3069/1000,
so the certified bracket and selected kappa stay unchanged.

Co-authored-by: unpaidmoon <unpaidmoon@users.noreply.github.com>
Root-linear factorizations of 2I_8 cost exactly 9t on the 120 lines, so nine or more terms cannot undercut the eight-term floor. A rank-9 excursion shows the rank hypothesis is necessary. No selected-frontier change.

Co-authored-by: unpaidmoon <unpaidmoon@users.noreply.github.com>
…r frames, with dual-source legality and actual native recompilation
… independent full 632-column forward/dual replay
…eter, exact ordered frame/warp invariants, negative control and source-pinned patch
…ct witnesses, whole-cost scope and attribution
zxv and others added 30 commits October 24, 2026 00:04
Exact rational certificates and quantified gross-integer mixed-budget proof; public PR5418 model with separate origin. No pricing, physical data or full compiler admission.
Retain the full current-source Q120 interface and use a source-preserving
correction instead of eight old-h calls. Include independent component
verification, pinned native premises and a scoped standalone cost bound.

No complete caller, funded kappa or selected metadata change is claimed.
Point current CI guidance to frontier5413 and retain the4610 historical
checkpoint link. Reuse the reviewed documentation bytes; scientific
package and tests remain unchanged.
…er free scalar reordering (negative result, no new kappa)

Every admissible re-routing of the 2,526 droppable intermediate levels of the
CrocSwap#4610 word is enumerated (518,315) under the retained frame contract with the
chronological order freed, joint frame chains and movable centre reads; an exact
acyclicity check shows no level can be dropped at a gain (the eight feasible drops
move a centre read to the full frame, +0.983 each). Control without the centre
interface finds 51 improving drops; kap-family ceilings quantify what Lambda would buy.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…xt files

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…wap#5420)

Canonical main still names make frontier4610-verify while the selected record
requires frontier5413-verify, so the shared onboarding test fails before any
package replay. This reuses byte-for-byte the docs/ci-verification.md repair
of CrocSwap#5420 (head baa8fbb, Clear1nsight);
no scientific file changes.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…wap#5420)

Canonical main still names make frontier4610-verify while the selected record
requires frontier5413-verify, so the shared onboarding test fails before any
package replay. This reuses byte-for-byte the docs/ci-verification.md repair
of CrocSwap#5420 (head baa8fbb, Clear1nsight);
no scientific file changes.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…rd (no scientific change)

The frozen PR2809 loader inside inputs/pr2809.tar authenticates certificates/selected-result.json against the record pinned at this package's base, so every frontier promotion broke the replay (working source drift, flagged in CrocSwap#5417). Vendor those exact bytes (blob b1fea0c, sha256 bb2eaa58...) as inputs/selected-result-pinned.json and let source.py serve them to the loader from a private repository view; every other declared input is copied unchanged and still authenticated. SOURCES.json marks the selected record as provenance and VERIFY.json drops it from the live inputs. Archives, checkers and RESULT.json are unchanged; all eight registered commands reproduce the stored receipt digests.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…lifts

Prove a general necessary condition and certify one fixed-projection exclusion. Preserve incomplete searches and exact verifier; f(8) remains unresolved.

Prepared for Nicholas Stone with substantial OpenAI Codex assistance.
…fter CrocSwap#5421 (negative result, no new kappa)

On the reverse PR589 BIT word, a matroid-union certificate shows that every
relation subspace V removes paid rank mass w(V) <= 7.5 dim V modulo CrocSwap#5421's
nine relations, so each extra relation dimension loses at least 0.44 raw
moment per local call. Calibration recovers exactly CrocSwap#5421's nine relations
as the only tight subspaces of the source word.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…wap#5420)

Canonical main still names make frontier4610-verify while the selected record
requires frontier5413-verify, so the shared onboarding test fails before any
package replay. This reuses byte-for-byte the docs/ci-verification.md repair
of CrocSwap#5420 (head baa8fbb, Clear1nsight);
no scientific file changes.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
… word after CrocSwap#5383

On the complex CrocSwap#4610 backward word (local word of the selected CrocSwap#5383 ->
CrocSwap#5398 -> CrocSwap#5413 chain), CrocSwap#5383's eight relations free all eight centres and
eight zero-site climbs; modulo them the 155 remaining pure-cold climb rows
are linearly independent over Q with rank mass <= 3, so every further
relation subspace loses at least 5.94 raw moment per dimension.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…wap#5420)

Byte-for-byte the docs/ci-verification.md repair of CrocSwap#5420 (head baa8fbb, Clear1nsight); canonical main still names frontier4610-verify, which fails the shared onboarding test before package replay. No scientific file changes.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…lambda/2)+6 bits, attained (no new kappa)

The CrocSwap#5383 eight-relation body (selected chain) and its CrocSwap#4610 source are
exact on a fixed dyadic Gaussian grid with ceil(9 lambda/2)+6 extra
fractional bits backward and +5 forward; every coefficient is dyadic and a
single-occurrence witness (coefficient -1/64, Clifford Fourier rank 9)
shows the guard is attained. Replaces the 2^192 envelope check for this word.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…wap#5420)

Canonical main still names make frontier4610-verify while the selected record
requires frontier5413-verify, so the shared onboarding test fails before any
package replay. This reuses byte-for-byte the docs/ci-verification.md repair
of CrocSwap#5420 (head baa8fbb, Clear1nsight);
no scientific file changes.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…ew scope [skip ci]

Scientific validation and selected-frontier admission remain open.
Replace one schoolbook product per tape action by the PR4192 four-field
record comparisons and show both excluded full-batch corners lie strictly
under E. The selected frontier is unchanged.

Co-authored-by: unpaidmoon <unpaidmoon@users.noreply.github.com>
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.