Docs: map which QQ16 observation contracts are already checked

[commdata2338]

Change and mathematical scope

The onboarding brief suggests testing a changed QQ16 first pair against the original lost-query witness, but a public fork catalogue (pending, not on main) already covers all 252 retained first pairs. Other follow-ups distinguish a repaired witness, recovery granted an intermediate B1 cut, and recovery from final records. Those contracts are easy to conflate when following the historical entry point.

This adds docs/research/qq16-controller-scope-map.md and one sentence to the QQ16 paragraph of docs/research/bottleneck-brief.md; the paragraph's existing guidance is unchanged. The six-source map states the exact observations, timing, result, excluded hypotheses and publication status. In particular:

• The 9-by-28 first-pair catalogue changes route choices inside fixed logical maps; it is not an obstruction to every first map.
• PR2079 detects eight old witness directions; PR2093 provides a different invisible family even after every nonzero-chart initial read.
• PR2174 recovers pair 0's original selected inputs for arbitrary four-pair inputs with a recorded post-B1 cut. PR2346 excludes replacing that cut by its specified 64 initial coordinates plus final observations on its ordinary-tail schedule. It does not exclude every 64-coordinate record or early selected interventions.

Kind: source-bound navigation audit. No new mathematical obstruction, implementation, paid-profile improvement, supplier or κ is claimed. The selected result, frozen packages and task priorities are unchanged. Open submissions and the public fork are explicitly distinguished from retained main evidence.

This is documentation only, so no new reusable verifier/package was created. The repository package generator was considered; duplicating the six existing evidence packages would not improve this scope map. Their own package metadata, source pins and registered checks were used instead.

Task and overlap

Base and final main refresh: 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2, priority revision 4. Canonical main was fetched at entry and again at 2026-10-15 06:44 UTC; main, guidance, selected record and all four pending PR heads below were unchanged. Documentation commit: 157143ab500812dd16cf50d12c1e5749bd6ea482; main was rechecked unchanged at 06:54 UTC before submission.

Exact evidence heads (no pending implementation is adopted):

Source	Head
Ali Romman, PR1969, retained	a6e6ad3936f67b8f63b5b68c5e75d540610a65e7
Philippe Van Velzen, first-pair catalogue	261eb732c229cf523304e1c3cf7123251260c992 in philippeVV/integer-mult-bounds
Cinna Davis, PR2079	49df8190b158e3028e702b2c6b798117006f7249
Adam Gall, PR2093	8b21a900eb520e4f11ea177b310f66316f8e8983
commdata2338, PR2174	f79470a7931f3943360c10d15b428e0e4e479c22
alephnan, PR2346	653520d7f18f686ea3a274b4963981f5be563477

The four-bank predecessor 680d0997621a2cecfef2186ab143377658687825 is credited as context for PR2346, not newly audited here. The selected source remains PR1094 0fc329d0bec93ca7cd288eaea1295172e8789810.

Task ID / proposed scope key: scope-navigation / b18e0e173a4430a9564b; refined QQ16/observation-contracts/first-pair-versus-recorded-cut-versus-final.
Base commit; exact predecessor and dependency heads: listed above; no pending supplier dependency is introduced.
Concrete object and question: the onboarding joint-query paragraph and its historical fixed-first-pair witness route; which later hypotheses are covered, and when does cut access prevent transferring a recovery result to final-only observations?
Scope status: refined; one bounded family index allowed by the original card, with a concrete onboarding gap. No mathematical premise is changed.
Existing restriction being changed (or exact claim being audited): audit the scope implications of the six named contracts, preserving their exact observation times and evidence status.
Method; explicit parameter ranges/candidate alphabet: retained A/T/U96 model, 9-by-28 first-pair choices, named fixed-pair B2/all-nonzero-chart/two-query/final-record variants. Read proofs and source bindings; replay the original bounded checks.
Nearest overlapping PRs and why this adds something: PR2079/2093/2174/2346 supply the mathematical results rather than this onboarding map. PR2037 maps R439 semantic-D; PR2183 maps bootstrap stock. Their brief edits concern other paragraphs.
Optimistic full cost budget; required observable improvement: no cost claim. Improve the decision route by distinguishing witness separation, full selected-input recovery, intermediate-cut access and final-only records.
Success evidence; falsification test; time/compute cap: exact immutable source links, correct hypothesis table, original package replays and controls, link resolution and patch checks; reject any generalization to all maps, all controllers or arbitrary 64-coordinate records. One bounded session, at most four busy processes and 12 GiB.
Predecessor authors; intended artifact and reproduction command: credited below. Two documentation files; predecessor commands are listed under Validation.

Chooser: nonce bafb6bd89e7033358716a71143cc59c7, track consolidation; the other suggestion was overlap-receipt / 5ae821d1b52092164092. Scope navigation was chosen because the onboarding gap was concrete; no comparable small integration-disposition gap was identified. No reroll or central-card override.

Coordination: the claim is in Discussion #2074/comment18942780. The overlap helper searched the card, QQ16 scope/navigation/observation terms, the exact new path and the bottleneck-brief path. Final shared snapshot: 574 open PRs including drafts at 2026-10-15T06:43:17.659075+00:00, 1,089 discussion records at 2026-10-15T06:45:17.647647+00:00, and 572 changed-file records at 2026-10-15T06:33:04.255876+00:00; all reported complete pagination. Offline warnings and possible later changes remain explicit. No matching QQ16 navigation map was found. Path overlap with PR2037 and PR2183 concerns different paragraphs. Active coordinate-access, dense-record and native-mechanism work is outside this index's mathematical scope; no unclaimed status is inferred.

Validation

CPython 3.12.3, Linux x86_64, standard library. Exact package bytes were copied from each pinned Git head into an isolated temporary tree, and input bytes were checked against each predecessor manifest. No predecessor source was edited. The existing runner checked metadata/pins and executed the registered commands.

From an isolated checkout of the corresponding exact head:

# PR2079
python3 -I -B scripts/research_package.py --package research/joint-query-extra-b2-observation-seed3e62
# PR2093
python3 -I -B scripts/research_package.py --package research/qq16-b2-observation-kernel
# PR2174
python3 -I -B scripts/research_package.py --package research/qq16-two-query-kernel
# PR2346
python3 -I -B scripts/research_package.py --package research/qq16-final-record-obstruction

All four returned PASS_REGISTERED_RESEARCH_PACKAGE. Main markers and coverage:

• PASS_QQ16_EXTRA_B2_OBSERVATION_SCREEN: eight displayed directions and the original conditional profile checks; about 4.3 s.
• PASS_QQ16_B2_OBSERVATION_KERNEL: three-port witness, eight rejecting controls; about 1.2 s.
• PASS_QQ16_TWO_QUERY_KERNEL: three free ordered identities, 17 tail events, 160 basis fixtures, twelve controls, normal/optimized replay and three artifact tests; about 6.1 s.
• PASS_QQ16_FINAL_RECORD_OBSTRUCTION: 1,008 actual routes, six relative-route pairs, 19 events, fifteen controls, normal/optimized replay and three integrity tests; about 13.7 s.

At fork head 261eb732c229cf523304e1c3cf7123251260c992:

sh research/joint-query-first-pair-catalogue/reproduce.sh

The underlying commands passed PASS_QQ16_PAIR_CATALOGUE_WITNESS and PASS_FIRST_PAIR_TRAJECTORY_ALL 252; both ctl_47 and ctl_noA rejected with their intended assertion failures. The regenerated full JSON matched pair_catalogue_result.json exactly: 28 distinct-route checks, all 252 first-pair fixtures, and the inherited tail. These runs took about 20 seconds total. The trajectory uses exact rational permutation fixtures; the actual-route group-algebra checks and written proof are separate evidence. Run without -O because the predecessor's trajectory uses assertions.

Also passed: python3 -B scripts/research_baseline.py (PASS_RESEARCH_BASELINE_BINDINGS_ONLY), resolution of 23 local or immutable Git link targets, git diff --check, and patch application against an isolated index of final origin/main.

These are source inspection and reproductions of the authors' checkers, not independent implementations or formal proof review. The first minimal export omitted PR2079's runtime selected-record file (declared provenance in that package); adding its exact pinned bytes made the unchanged checker pass. This was an export omission, not a mathematical failure. No source pin was weakened.

Before submission the brief edit was reduced to one added sentence, keeping the paragraph's original guidance, and every local and immutable link in the map was rechecked: the four PR heads and the fork commit resolve.

Not run: the selected full cold replay, a new PR1969 local native replay, full regular-cover enumeration, or a proof assistant. No new controller, native invoice, ownership/hosting, precision/tape, all-size admission or optimal transcript theorem is established. No tests were added for this documentation change.

Credit

Ali Romman supplies PR1969's local maps, source packet, native contracts and retained witness/review lineage. Cinna Davis supplies the B2 witness-repair test. Adam Gall supplies the nonzero-chart hidden-input witness. Philippe Van Velzen supplies the complete retained first-pair catalogue and earlier four-bank final-record lead, with the fork's AI-assistance disclosure preserved at the immutable source. This account's PR2174 supplies the recorded-cut decoder; alephnan's PR2346 supplies the actual-cover final-record obstruction. Ross Richardson's three-query rank method and the Rohan Arun, Jeremy Kothe, Hao Yan, geometry and compiler lineage remain credited through those sources and the unchanged QQ16 notice.

Prepared for commdata2338 (@commdata2338), with substantial OpenAI Codex assistance. New documentation is Apache-2.0; no predecessor package or notice is copied or changed.

AI assistance: submitted by @commdata2338 with substantial AI assistance. OpenAI Codex (GPT-6 Astra) did the derivation, the implementation, the checking and the writing; it was packaged and submitted with Claude Code. No independent human review is claimed.

Open next questions

The map's next-question section asks a successor to identify a hypothesis outside the first-pair catalogue or explicitly recover the B1 cut from its available records. Start with the actual ordered observation operator and test the complete selected-input kernel, then price native access, retention and cleanup. These are criteria for defining a useful next task, not claims that a specific mechanism is available. Check the current board before taking a follow-up.

Workflow feedback (optional)

The task chooser usefully directed this session to a bounded navigation gap. The main friction was that frozen packages correctly retain historical open questions, while later closures and changed observation cuts are discoverable primarily in long discussion threads. A short successor scope-map link at onboarding addresses this without rewriting historical evidence or undertaking a global survey.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG