[Clear1nsight]

Change and mathematical scope

Independently qualify PR #5757's complete selected P104 replay on Linux. Six alternating fresh processes reduce median wall time from 339.49 to 130.58 seconds: 2.60x faster, or 61.54% less time. Every complete scientific stdout hash agrees. The original full RESULT/WORD comparisons, native phases, physical caller, complete invoice, compiler, outer checks and rejection controls run unchanged.

Timed peak kernel RSS increases from at most 475.05 to 492.72 MiB. Normal qualification of both original registered scientific commands takes 402.88 seconds; separate optimized qualification takes 404.05 seconds. This extra literal-oracle work is outside the 2.60x timing claim. All ten scientific commands pass; no source bytes or permissions change.

This is verification engineering. Conditional κ remains .00310045 and the five framework assumptions remain. The κ construction campaign stays paused.

Task and overlap

Refine replay-runtime to Linux qualification of exact PR5757 head 53006fe6031dab5c23d5b7d1cf6a92f411a04a68. Canonical main is de5ec2e7a6074068452f45b1a191b5c72832df4a, priority 8. The released source investigation explicitly left Linux measurements open. Our bounded claim preceded the runs. Targeted pre-submission overlap found no competing PR referring to this source; coverage is bounded.

The scientific contribution is one isolated 59-file package. The branch also carries the exact three-file test repair from PR #5761, with Ali Romman’s authorship and source commit preserved. It retains the original official 54-file source export, 31 live inputs and original notices without changing the engine or source proof. The original pending Git head is not needed at runtime. Scientific evidence commit: 5195ab6d93425bb47b2bb8f9acba05d2ae307413. Current PR head, including the test-only CI dependency: dd59757bd1d938e2aed75e73f24e7ae7b8054211.

Validation

Ubuntu 24.04.5 LTS, Linux x86_64, CPython 3.12.3. Literal timings are 349.64, 331.02, 339.49 seconds; quadratic timings are 130.58, 127.27, 132.14 seconds, in literal/quadratic/quadratic/literal/literal/quadratic order. Fresh processes, warm OS caches, variable host load; no sample is discarded.

Both normal and optimized full adapter checks pass. Both normal and optimized actual-state oracles compare all 17,651 distinct inputs and 9,037,312 amplitude entries per original literal interpreter, including zeros and denominators. Their complete outputs agree byte-for-byte. All thirteen source assertions remain explicit failures in the derived compiler; independent AST inspection finds none inside the replaced state method.

python3.12 -B scripts/research_package.py --package research/selected-quadratic-linux-validation

This short registered command checks recorded evidence and source custody, with 17 synthetic evidence rejections and three actual source-entry corruption controls. It does not launch the expensive positive scientific experiment. Complete cold reproduction is explicit:

python3.12 -B research/selected-quadratic-linux-validation/reproduce.py all --workdir /tmp/pr5757-linux-fresh

The original guarded phases took 2,217.07 seconds total, with 600 seconds per command, sampled aggregate process-group RSS capped at 1 GiB, and a 4,500-second shared budget. No stop or retry occurred. The official export integrity/input checks, exact committed bytes/modes and immediate pin ancestry pass. The optimized recorded-evidence checks also pass from a Git-free, read-only export with empty PATH and all 60 files/modes unchanged. Full-history ancestry preflight is unavailable in the shallow checkout.

Hosted CI is separate. The initial run failed on revision-8 catalogue expectations in unchanged shared assignment tests. Commit dd59757bd1d938e2aed75e73f24e7ae7b8054211 cherry-picks PR #5761 at ba8a9e640e1983f3c29993fefa3f46e8ed03313a: it preserves legacy cases as explicit frozen fixtures and independently checks current catalogue routing. All 28 focused normal assignment tests and 12 optimized current-assignment tests pass locally; recorded-evidence controls and exact committed bytes/modes pass. All 59 scientific package files are unchanged. The updated hosted run reports CI separately. Details, raw outputs, resource measurements, proof coverage and limitations are in the package README, RESULT and evidence directory.

Credit

Ali Romman supplied PR5757, the PR #5704 quadratic engine, original proofs/controls and macOS measurements. Retain the PR #5599 construction, native/controller/compiler/helper/BIT and classical stabilizer lineage in NOTICE and the unchanged source export. New Linux execution, review, reproduction and evidence checks were prepared for Chris R / Clear1nsight with substantial OpenAI Codex assistance. No subagents or independent human/proof-kernel acceptance. Apache-2.0.

Open next questions

Review whether this opt-in acceleration is ready for adoption after its existing mathematical obligations are checked. Any newer word needs a new source/consumer binding and full output/control comparison. Reducing repeated oracle cost needs a precise dependency and invalidation contract; cached PASS results are insufficient. This submission releases the bounded Linux qualification scope and resumes no κ research.

[Clear1nsight]

HANDOFF — bounded Linux qualification complete; scope released.

Research discussion / PR / artifact: PR #5816, research/selected-quadratic-linux-validation/, supporting PR #5757.

Exact commit: 5195ab6d93425bb47b2bb8f9acba05d2ae307413; original subject 53006fe6031dab5c23d5b7d1cf6a92f411a04a68; main context de5ec2e7a6074068452f45b1a191b5c72832df4a.

What is established: all ten guarded Linux scientific commands pass on CPython 3.12.3. Median complete literal/quadratic replay is 339.49/130.58 seconds (2.60x), with identical full scientific output. Both normal and optimized adapter and all-actual-state oracle checks pass. Every one of 17,651 distinct states is compared against both original literal interpreters, 9,037,312 amplitudes each. The complete original source proof/caller/invoice/compiler/outer checks and controls remain active. Timed peak kernel RSS is 475.05/492.72 MiB. All 54 runtime source files remain unchanged and read-only.

What remains uncertain: one host and three timing samples per engine do not establish universal performance. Normal registered scientific qualification costs an additional 402.88 seconds; optimized qualification 404.05 seconds. Five framework assumptions remain; no new kappa, physical multiplier, successor transfer, independent human or proof-kernel acceptance. Full-history preflight is unavailable in the shallow checkout; immediate declared-pin ancestry and exact bytes/modes pass. Hosted CI is recorded separately.

Smallest useful next task: inspect the exact source binding, Linux evidence and memory tradeoff, then review adoption of the existing opt-in acceleration. A changed source word or engine needs a new applicability/invalidation check. This handoff resumes no kappa research.

Reproduction command and expected runtime: python3.12 -B scripts/research_package.py --package research/selected-quadratic-linux-validation checks recorded evidence, 17 synthetic rejection cases and three actual source-entry corruption guards in a few seconds. It does not replay the positive scientific experiment. python3.12 -B research/selected-quadratic-linux-validation/reproduce.py all --workdir /tmp/pr5757-linux-fresh explicitly runs the complete cold experiment, about 37 minutes on this host, serially under the stated limits with no retries.

Predecessor credit: Ali Romman's PR5757 integration and PR5704 engine, PR5599 and the retained native/controller/compiler/helper/BIT/stabilizer lineage and notices. New Linux work for Chris R / Clear1nsight with substantial OpenAI Codex assistance; no subagents. Apache-2.0.

[Clear1nsight]

Hosted CI at exact head 5195ab6d93425bb47b2bb8f9acba05d2ae307413: run 40431567656 failed during shared onboarding tests; package selection and replay were skipped. The plan log records 511 tests in 158.027 seconds, with eight failures, two errors and 22 skips. All ten failure/error entries are in tests/test_research_current_assignment.py, including the revision-8 receipt/legacy-routing expectations.

The test module blob 860f6b9b9eea2ccfcd1131f49693581eba92f645 and catalogue blob 3404fbdd5e7529a0674eef26114b5071e4586cec are identical at this head and its canonical base. This is the existing shared failure addressed by PR #5761, whose separate hosted plan passed; this submission adds no duplicate repair. A hosted package pass is not claimed.

Local evidence remains separate: all ten positive scientific commands passed, the registered recorded-evidence checks and 17+3 rejection controls passed, and the sealed Git-free optimized checks preserved all 60 files/modes. The next CI step is to rerun after the shared repair is incorporated. No source or expected mathematical output was changed to address CI. The bounded Linux contribution is submitted and handed off; kappa research remains paused.

[Clear1nsight]

HANDOFF — PR #5816 workflow repair complete; verification-only scope released.

1. Research discussion / PR / artifact: PR #5816, supporting PR #5757, with the exact shared test repair from PR #5761.
2. Exact commit: current PR head dd59757bd1d938e2aed75e73f24e7ae7b8054211; Linux scientific evidence commit 5195ab6d93425bb47b2bb8f9acba05d2ae307413; imported repair source ba8a9e640e1983f3c29993fefa3f46e8ed03313a.
3. What is established: hosted run 40434259033 completed successfully at this head. The plan passed 513 tests in 188.700 seconds with 22 existing platform skips and selected only this Linux package. The package replay passed recorded-evidence validation, 17 synthetic rejections, three actual source-entry corruption controls and the committed-input immutability check. The original failure came from historical routing expectations against catalogue revision 8; the repair preserves legacy fixtures and current-catalogue coverage. All three imported Git blobs match the credited source. Local normal assignment tests (28), optimized current-assignment tests (12), exact committed bytes/modes and diff whitespace checks pass. All 59 scientific package files are byte-for-byte unchanged.
4. What remains uncertain: hosted CI validates this package’s recorded evidence and controls; it does not freshly rerun the ten expensive positive scientific commands, whose complete Linux evidence remains preserved. Timing generality, the five framework assumptions and all original scientific limitations remain. No new κ, selected promotion, merge or independent human/proof-kernel acceptance is claimed.
5. Smallest useful next task: review the frozen Linux evidence and opt-in replay acceleration, including its memory tradeoff and source binding. No further execution is dispatched by this handoff; κ research remains paused.
6. Reproduction command and expected runtime: python3 -B -m unittest discover -s tests -p 'test_research_*.py' -v runs the hosted helper suite (about 189 seconds here). python3 -B scripts/research_package.py --package research/selected-quadratic-linux-validation checks recorded evidence and controls in a few seconds. The original full cold reproduction command and approximately 37-minute cost remain documented in the existing scientific handoff; it was not repeated for this test-only repair.
7. Predecessor credit: Ali Romman’s PR #5761 is cherry-picked with original authorship and the source commit reference retained. Preserve the existing PR #5757, PR #5704, PR #5599 and retained mathematical/software lineage. Integration, checks and this report were prepared for Chris R / Clear1nsight with OpenAI Codex assistance; no new subagents.