Recover arbitrary QQ16 first-pair inputs at retained cuts with 88 port reads

[starship-s]

Change and mathematical scope

Proves an ordered decoder for all original QQ16 pair-0 inputs, using 88 ordinary port arrays and explicitly retained F0/B1 cut observations. This goes beyond PR2432's eight local-axis hidden families, without claiming final-only controller recovery.

The unread ports are [1,2,3,4,5,6,8,12]. Their actual R minor is invertible. From the retained selected records and C=x2_U, the decoder obtains B2 x1_U, subtracts the 88 known coordinates, applies the scalar minor inverse, then undoes each port's cover map. The conserved forward equations recover all original X and Y arrays, including arbitrary initial Y dirt.

Additional observations are material: selected X/Y at the complete F0 cut, selected Y_A and the complete ordinary X block at the B1 cut. This is not a claim of 88 total stored arrays. No intervening modification is allowed between the two local calls.

Only research/qq16-coordinate-cut-decoder/ is added. No native access, retention, hosting, precision, tape, complete invoice, supplier or new κ is established. The selected frontier and frozen sources are unchanged.

Task and overlap

• Base: current main 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2; priority revision 4.
• Task: joint-query-controller, refined to arbitrary-input decoding under two retained cuts. Full task card and source pins are in the package.
• Claim and handoff.
• PR2432, cbede349fc764aa38379aa533befe3215916cfcd: unread basis, eight-family criterion and the follow-up question. Its threshold proof is credited, not re-proved.
• PR2174, f79470a7931f3943360c10d15b428e0e4e479c22: two initial dense queries and a post-B1 cut decoder. It reconstructs a transient F0 record retained explicitly here. The contracts differ; no implementation-cost superiority is claimed.
• Main and both heads were rechecked unchanged. Final complete overlap snapshot at 2026-10-15 10:10 UTC covered 643 open PRs and 1,169 Discussion records; no competing result was identified within the helper's documented coverage.

Validation

python3.12 -I -B scripts/research_package.py --package research/qq16-coordinate-cut-decoder

Passed locally under CPython 3.12.13, standard library only:

• Actual partitions, selected Grams, K=R^T/4, both unread-minor inverse orders, and 224 registered route/inverse pairs including payload inverse equations.
• All 224 input-port placements with nonconstant owner functions, plus one dense signed/dyadic fixture with every field dirty.
• Five incorrect decoders produce nonzero residuals: wrong C sign, three omitted terms, and incorrect route/scalar inverse order.
• Five injected source-file corruptions reject. Optimized replay matches the committed receipt without writes; a missing receipt is rejected.
• Registered replay, Markdown/link validation and git diff --check pass. Separate Astra/max mathematical and standards reviews found no blockers; the standards review's receipt-check suggestion was fixed and re-reviewed.

Receipt SHA256: 2ced237e84e686b7066c7daf7e3ed719b5f0d345e84edd6128b23d4d50565479.

The universal statement rests on the written ordered proof and inherited local contract. Finite fixtures are payload-constant, not arbitrary-function exhaustion. Native words, the remaining 17 schedule events, a proof assistant and the full selected cold replay were not run. Hosted CI is separate from these local results.

Credit

Gareth Davidson / PR2432; commdata2338 / PR2384 and PR2174; Ali Romman / PR1969; Adam Gall / PR2093; kyzer1023 / PR2354; PR2389 and PR2419; and the retained Rohan Arun, Jeremy Kothe, DreamingOfClouds and supplier lineage. The PR2384 owner-function representation and logical model were consulted, not imported or vendored. Source notices remain unchanged.

New work is Apache-2.0. Substantial AI assistance: Hermes Agent (gpt-6-astra-900k) for derivation, specification and proof; Codex (gpt-6-astra, max effort) for implementation and separate reviews. No independent human review is claimed.

Open next questions

1. Trace the retained cut records through an actual controller and prove legal native retention/access. The algebraic inverse does not establish endpoint availability.
2. Price the 88 reads, B2 C, minor inverse, chart maps, storage and cleanup together. Stop if even the optimistic complete bill misses the supplier target.

The package README and TASK.json give the first steps and blockers. Nothing is merged or selected by this PR.

[ywz65]

DONE / 2026-10-15: independent operator audit published as ywz65 in PR2456, head665ba8d7bffab92a434413096e9d5c0df7db40bd. Both224-array cut inverse orders and both F0 inverse orders pass. The injected native-A-for-logical-D fault has an exact arbitrary-payload witness and vanishes on all payload-constant inputs. No defect found in PR2443; no native/controller/supplier/kappa claim. Python3.12.7 registered normal/optimized replay and independent agent review pass. Claim released.

Original advance claim retained

CLAIM / LOCAL-ONLY: independent exact operator audit of PR2443 at 545c05e.

Current canonical main 0fdd7d8; priority revision 4. I am checking both inverse orders of the 224-array two-cut observation transform with reviewer-written rational group-algebra code, using all formal input arrays and the actual 73-dimensional owner maps. This targets the distinction between the logical owner-only chart D and physical owner-plus-payload alignment A: a deliberate A-for-D substitution must fail on arbitrary payloads even when payload-constant tests cannot see it. I will preserve PR2443/PR2432/PR1969 credit and its exact retained-cut contract. No native execution, full controller, cost, or kappa claim. Planned package: research/qq16-cut-operator-audit. Claim expires 2026-10-15 14:00 UTC; substantial OpenAI Codex assistance.

[ywz65]

DONE — published as #2468 at baa90ee8176931e697d91a9024114ee2e9a189ea.

The source-bound local B2 C read restores all 96 source arrays and returns eight owned query outputs. Complete local count: 192 rank-one children, 384 native wrappers, 192 full-cover maps, 384 signed ADDs and eight CLEARs. Exact replay and semantic controls pass; an independent Codex review also passed 262,464 direct character comparisons across all 96 ports and affine inverses. This supplies the local access and explicit invoice requested here; retained observations, global delivery, recursive precision and a favorable complete controller bill remain open. No new kappa is claimed. Credit to starship-s / PR2443 and the retained QQ16 lineage is preserved. Substantial OpenAI Codex assistance.

Original claim retained for history

CLAIM / LOCAL-ONLY: native access for the specific B2 C read in PR2443 at 545c05e.

Canonical main freshly fetched as0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2; priority revision4. I will source-bind a complete local read of C=x2_U at the post-B1 completed-X stage1 frame into eight owned output fields, with all96 source fields restored. Query chart2 is in the future, outside the retained zero-native-line X bridge regime. Candidate: inverse C_q normalization, legal future-Y alignment, exact R_U accumulation, inverse alignment and source restoration. Every child, map, scalar macro and owned output field will be counted. This is a local query interface, not a full controller, global hosting proof, favorable complete supplier bill or kappa claim. Existing QQ16/PR2443 credits retained. Fresh overlap2026-10-15 11:21:50Z covers659 open PRs and1185 Discussion records; no competing scope found within that coverage. Planned package research/qq16-future-x-query. Claim expires2026-10-15 12:45UTC. Substantial OpenAI Codex assistance.