Skip to content

Certify QQ16's separate entry-frame adapter tariff - #2678

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
petalas:research/selected-interface-audit-20261015
Oct 16, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
petalas:research/selected-interface-audit-20261015

Conversation

@petalas

@petalas petalas commented Oct 15, 2026

Copy link
Copy Markdown
Contributor

Summary

PR #2540 recovers all 224 original logical arrays in F_U1 but leaves their consumer frames open. This proves and constructs the sharp native tariff for separate per-field adapters to the original physical entry gauges:

224 recovered arrays in F_U1
  -> 112 rank-16 + 112 rank-17 native children
  -> original x in C_q, original y in I

The adapter rank mass is 3,696. Appending the unchanged added decoder gives 224 H1 + 16 H8 + 112 H16 + 112 H17, mass 4,048. The less demanding F0 gauges cost 1,904; common F_L costs 1,792. These change the frames of original x,y, not their logical contents.

The lower bound uses binary phase polarization and includes arbitrary products of single-field convolution kernels and affine payload permutations/translations, at every positive common tensor depth. Explicit orthogonal charts and translations attain it and minimize the separate native moment for 0 <= a < 1.

Cross-field synthesis, shared children, borrowed observations, payload phase masks and fusion with the decoder or a later consumer are outside the theorem. A caller that directly accepts F_U1 can avoid these adapters. Physical wrappers, movement, hosting, precision, F0/B1 costs and global supplier admission remain open; no new κ or selected-frontier change.

Evidence

  • Before: PR Complete the QQ16 native selected-sum decoder #2540's exact retained proof specifies common F_U1 outputs and leaves the original physical-frame/consumer obligation unresolved.
  • After: all 112 actual port labels, 448 adapters, 68,992 adapter phase checks, 17,248 source-hole phase checks and six rejecting semantic controls pass. A separate Gaussian-integer convolution check covers all 131,072 addresses for the first actual entry-X adapter.
  • The registered Python 3.12.15 replay passes normally and with -O, plus six artifact tests for source/result corruption, read-only Git-free relocation, overwrite refusal, compilation and physical convolution. A separate internal mathematical review and artifact review found no material issue; no independent human review or proof-assistant verification is claimed.
python3.12 -I -B scripts/research_package.py --package research/qq16-entry-frame-tariff

The package contains the scoped proof, literal adapters, exact source pins and open next questions. No predecessor executable or full selected cold replay is run. Base b4560d9e, priority revision 5; canonical-main compatibility and live-overlap details are recorded in TASK.json. Canonical investigation and claim.

Credit ywz65/#2540 and #2511/#2512; commdata2338/#2651 for parallel bank reuse with the same endpoint; Ali Romman and the retained QQ16 geometry/native/owner lineage, including starship-s, Sagar Thakrar and Ross Richardson. Source bytes and notices are retained. Apache-2.0. Substantial AI assistance was used.

Merge Danger

Door: two-way.

Blast Radius: package.

Adds only research/qq16-entry-frame-tariff/ and its opt-in replay manifest. Existing selected and frozen evidence remain unchanged.

@Mister-Meeseeks
Mister-Meeseeks merged commit de880ca into CrocSwap:main Oct 16, 2026
55 of 61 checks passed
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.

2 participants