Skip to content

Raise the fixed phase-B rank-16 packing from 52 to 54 batches (no κ) - #7549

Open
petalas wants to merge 1 commit into
CrocSwap:mainfrom
petalas:research/w1-1113d-frontier-step
Open

petalas wants to merge 1 commit into
CrocSwap:mainfrom
petalas:research/w1-1113d-frontier-step

Conversation

@petalas

@petalas petalas commented Nov 13, 2026 •

Copy link
Copy Markdown
Contributor

Summary

A new finite certificate raises the feasible phase-B packing of the fixed historical gauge44 word from the 52 batches reported by Josie Elliston's #1817 to 54 in each orientation.

fixed skeleton + original rank-one climbs
    -> 54 disjoint groups of 16 at legal gaps
    -> same complete shared-role occurrence traces

Claim type: finite certificate with a scoped logical substitution proof. The new witness uses 864 of 1,072 phase-B rank-one climbs, preserves all 712 shared-role traces of each 24,767-event program, and leaves the exact maximum in [54,67]. The witness, deterministic discovery pass and checker were produced by Nick Petalas. The first improvement is shipped without pursuing an optimum.

Only six files in research/phase-b-54-batches/ are added. Proof, certificate contract and source credit. The advance claim sought an exact optimum or a larger packing; the scope update records the delivered lower-bound improvement.

Credit #1817 for the legal windows, retiming model, original witness and question; Jose Manuel / jmbarrancoml and publishing account 1337isnot1337 #1779 for the batching proposal; ToikY #1833 for the shared-role substitution and hosting discussion; Arham Tahir #1605, Douglas Colkitt #1240, Sagar Thakrar #1012, Ross Richardson, Greer Guthrie #679 and the retained gauge44/native/controller lineage. Their scientific results are cited premises and background, not new audit deliverables.

Evidence

  • Before: Research: diagonal rank-16 batches under register commutation, 134 not 8 (audit of #1779; no κ claim) #1817 leaves the phase-B maximum between its feasible 52 and the cardinality cap 67. After: the committed witness supplies 54 groups in both orientations. The checker authenticates the frozen archive/member, checks every batch and legal interval, reconstructs the event permutation, and compares every complete shared-role trace. It checks actual batch contiguity and rejects sixteen semantic mutations across the two orientations.
  • CPython 3.12.15 on macOS 27.0.1 arm64, standard library only. The two warning-strict registered commands were timed locally at 1.342 seconds and 0.291 seconds; the full registered replay passed in 1.793 seconds. These are check durations, not performance-improvement claims. PASS_PHASE_B_54_BATCHES and PASS_REGISTERED_RESEARCH_PACKAGE were observed. Normal and optimized hash-bound receipts match exactly.
  • Package metadata/pins, exact committed bytes and modes, source-history ancestry, official package export integrity, and read-only Git-free normal/optimized exported replay pass. The exported and working receipts agree; all exported bytes stay unchanged. Certificate regeneration matches the committed bytes and occupied-output refusal preserves the existing file.

Exact reproduction from the repository root:

/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -W error -B research/phase-b-54-batches/verify.py --check-discovery
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -O -W error -B research/phase-b-54-batches/verify.py
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/research_package.py --package research/phase-b-54-batches
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/check_research_commit.py --package research/phase-b-54-batches --check-source-history --check-modes

Another CPython 3.12 executable may replace the absolute path. Default verification is read-only and imports no predecessor executable. --check-discovery separately regenerates the deterministic witness. Source, member and package hashes are generated in the stdout receipt.

Limitations and unrun obligations

This is a historical scheduling side result. It groups original climbs and does not replace them with a native rank-16 child. The actual paid profile and selected #7000 frontier remain unchanged. No new kappa, optimal packing, favorable complete invoice, physical implementation or complete multiplication theorem is claimed.

Native compilation, gathering, frames, shared physical scratch, ownership, restoration, precision/tape, hosting, compiler admission and analytic recovery remain separate obligations. No predecessor scientific verifier, selected cold replay, Lean or Docker build was run. Independent human and formal proof-kernel review are unclaimed. A changed physical child must name its escape from the support/volume obstruction referenced by #1833.

Open next questions

  • Determine the exact maximum in [54,67], with an independently checkable upper-bound certificate.
  • Specify a changed native/encoding contract for one exact batch and price its full physical implementation, including arbitrary-dirty restoration, before extending the construction.

Related open work

  • #1836 studies higher-rank pairs at one ADD on the same historical word; this certificate groups original rank-one climbs and does not compose that proposed profile.
  • #1865 schedules commuting C3 fanouts on the same word; its three-destination shear groups differ from these sixteen-climb batches.

The closeout refresh inspected this account's seven other open PRs created since the CLAIM: #7547, #7548, #7550, #7551, #7552, #7553 and #7554. They concern principal demands, rounded commutators, cactus acquisition, dual-polar cuts, integer gauges, an ordinary four-cycle and modular cache memory. None studies this fixed phase-B climb packing or computes a competing batch count. No extra computation checked another contribution.

CI notes

CI did not certify this package. At unchanged head bd31251149dd5305c203064dd4627a3334ec1b57, run 43522269247 completed with plan job 132219246601 failed and package replay job 132221534021 skipped.

I read the completed checks once and inspected the plan job log. The Onboarding helper tests step reports exactly one failing test among 760, with 24 skipped:

test_research_current_status.CurrentCiNavigationTests.test_current_ci_section_matches_selected_record
AssertionError: '`make frontier7000-verify`' not found
FAILED (failures=1, skipped=24)

At canonical base e54523517d0f03e58352e647c403a5af3e525a57, certificates/selected-result.json names make frontier7000-verify, while docs/ci-verification.md still describes #6030 and make frontier6030-verify. I read those base files and confirmed that the selected record, CI document and failing test are unchanged by this PR. The failure is in the inherited repository navigation; the dependent package replay never ran.

I also inspected the failed logs of rohanarun's run 42905094314 and dillona's run 42903313790. Each has the same sole failing test, missing command, 760-test count and 24 skips. Run metadata confirms the respective contributor forks. No other comparison run is claimed checked here.

No fix, retrigger commit, push, job rerun or change outside the package was made during closeout. All scientific/package evidence above remains local; independent human and formal proof-kernel review remain unclaimed.

Merge Danger

Door: two-way. Six isolated research-package files are added.

Blast Radius: research. Shared code, frozen archives and the selected record are unchanged.

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