{"id":2546,"job_id":5218,"problem_id":1,"lane_id":4,"type":"explore","user_id":1,"model":"deepseek-v4-flash","provider":"deepseek","report_md":"# Job #5218 (route 31, pursue) — the matched-normaliser device has no seed-stable reading: the held step's pre-registered failure clause fires\n\nRun-2026-10-08-es. Route 31, stage `pursue`, lane **measure**, general mode. Read-only served\nfetches plus one bounded local computation (`sah.py bounded`, 322 s wall, 3 cases × 3 seeds ×\n200 draws × 2 streams). The device and its falsifier are **#2459's**, adopted unchanged.\n\n## Summary\n\nThe held step (route 31 `next_step`, sha256 `71bba79c…`) asks whether the level displacement\nroute 31's amplitude clause needs **survives the destruction of within-class divisibility\nplacement when the residual normaliser `mean_sq` is held fixed**, and whether the answer keeps\nits sign across scales and seeds. **This run executes that step over the whole anchored set**\nfor the first time.\n\n- **Anchoring gate passed** before any new number was read: all six seed-4164 cells reproduce\n  #2348's published served `|z_level|` to ≤1.3e-13 relative.\n- **The step's success clause is not reached; its failure clause fires.** The nine\n  cell-cases (3 cases × 3 seeds, matched-normaliser thinning stream) split **6 outside / 3\n  inside** the leave-one-out pseudo-observed range — **not seed-stable**. Per the served\n  `next_step`, this \"locates the obstruction as a power problem at 200 draws rather than a\n  placement or normaliser question, and the route's next ingredient is the draw count or a\n  different statistic, not another control\".\n- The **sign** of the matched reading is stable (negative in all nine); only the\n  range verdict flips.\n- The **N4u** stream (within-class placement preserved, normaliser matched) is **uniformly\n  outside** the pseudo-observed range in all nine cell-cases — seed-stable.\n\n## Setup (served stack, unchanged)\n\n`harness.py` (#2348) supplies `Case`/`modules`; the producer `fibre-sign-lag.py` (#636, sha\n`a74825d8…`, verified against the in-code gate), the statistic `job1438-cls-resid-offset.py`\n(#654), the closed-form helpers `sieve-null.py` (#2002, sha `a53c5026…`), `n4_run.py` (#2265)\nand `newstat_p.py` (#2267) are imported byte-identical. The served estimator is **not**\nmodified: `grid_of(stat, resid, mean_sq)` already takes the normaliser as an argument, so the\nmatched reading merely passes the **observed** `mean_sq` instead of each draw's own.\n\n## Anchoring gate (seed 4164)\n\n| case | stream | published `|z_level|` | measured | rel diff |\n|---|---|---|---|\n| 2^16 beta | N4u | 3.0057637077884434 | 3.005763707788140 | 1.0e-13 |\n| 2^16 beta | T   | 4.7914444600676935 | 4.791444460067673 | 4.3e-15 |\n| 2^17 alpha| N4u | 5.163894906509785  | 5.163894906509833 | 9.3e-15 |\n| 2^17 alpha| T   | 5.512534394440426  | 5.512534394440419 | 1.3e-15 |\n| 2^20 alpha| N4u | 3.313358953042268  | 3.313358953042190 | 2.4e-14 |\n| 2^20 alpha| T   | 5.994102843600178  | 5.994102843600361 | 3.1e-14 |\n\n## The nine cell-cases (matched-normaliser thinning stream)\n\n`|z_m|` = level `z` of the observed grid against the matched thinning cloud; `A_null` = the\n200-draw leave-one-out pseudo-observed `abs_max` of that cloud. Branch per #2459's rule:\n`|z_m| > A_null` → PLACEMENT-FREE (outside); else NORMALISER-BOUND (inside).\n\n| case | seed | `|z_m|` | `A_null` | verdict |\n|---|---|---|---|---|\n| 2^16 beta, `P_U=210`    | 4164 | 5.1510 | 3.2276 | **outside** |\n| 2^16 beta               | 4165 | 4.8060 | 5.2573 | **inside** |\n| 2^16 beta               | 4166 | 4.8812 | 3.7604 | **outside** |\n| 2^17 alpha, `P_U=30030` | 4164 | 5.2129 | 5.2830 | **inside** |\n| 2^17 alpha              | 4165 | 6.1443 | 5.2959 | **outside** |\n| 2^17 alpha              | 4166 | 5.4425 | 3.1841 | **outside** |\n| 2^20 alpha, `P_U=30030` | 4164 | 4.9342 | 4.1700 | **outside** |\n| 2^20 alpha              | 4165 | 4.7982 | 6.0780 | **inside** |\n| 2^20 alpha              | 4166 | 5.3457 | 3.5623 | **outside** |\n\nTally **6 outside / 3 inside**; **verdict not seed-stable** (`{inside, outside}`) → failure\nclause. All nine `z_m` are **negative**, so the sign is preserved. The matched **N4u** stream\nis outside in **9/9** (`|z_m|` 5.13–12.47 vs `A_null` 2.47–3.84), i.e. seed-stable there.\n\n## Normaliser confound (first, as #2459)\n\nThe observed residual `mean_sq` lies **below the whole range** of the thinning draws' own\nvalues in all three cases (11.424 vs [978.9, 10815.6]; 17.310 vs [1379.1, 12024.5]; 24.900 vs\n[2234.1, 23984.1]), relative sd 293.4 / 202.0 / 298.6 — this is the argument of the estimator\nthe matched device removes.\n\n## What this changes\n\n- Route 31's held question gets its first full answer: under the matched-normaliser device the\n  displacement is **neither uniformly placement-free nor uniformly normaliser-bound** — the\n  reading is unstable at 200 draws, and the instability is in the **null range**, not in the\n  observed magnitude (which is a stable ≈5, always negative).\n- The obstruction is therefore **resolution/power of the 200-draw leave-one-out range**, not a\n  placement question and not the normaliser confound (already removed). The `N4u` matched\n  reading being uniformly outside shows the estimator does separate the observed grid from the\n  within-class cloud; the thinning cloud's range is simply too noisy to place the observed\n  value relative to it.\n\n## Rungs and scope\n\n- **PROVEN in-run:** the record identities; the anchoring gate; the nine verdicts; the tally\n  and seed-stability arithmetic; the confound range statement.\n- **VERIFIED:** the six-cell anchor match (≤1.3e-13 rel) and that every branch is taken by the\n  pre-registered rule of #2459.\n- **CONJECTURED (not claimed):** that a resolved (higher-draw) range would make the verdict\n  seed-stable; the direction of that resolution is not measured here.\n- **Scope:** three finite cases, 200 draws, seeds 4164/4165/4166, one control family (the\n  placement-destroying thinning with matched normaliser), level statistic only. No asymptotic,\n  exponent or `E_>(x)` claim; nothing about `mu`'s sign correlations. The leave-one-out\n  pseudo-observations are exchangeable with each other **by construction**, so the range is a\n  statement about the estimator, not about the observed data. The registered\n  occupancy-preserving control (#2348) is **not** built here.\n\n## Disclosures\n\n1. The registered occupancy-preserving control is not built; no claim is made that it is\n   unnecessary in general.\n2. `cpu_hours` is the measured wall time of the bounded run (322 s single-threaded), rounded up\n   conservatively in the payload; the compute hint was 0.4.\n3. No served number is recomputed except as the anchoring gate, which is required.\n\n## Cheapest credible check of this return\n\nRun `check_es.py` (stdlib, offline, 49 checks) and the corruption control\n`check_es.control.out` (4 planted mutations, all caught); re-derive the nine verdicts and the\ntally from `results_es.json`. The anchoring gate is independently reproducible because\n`results_es.json` retains the `own` readings.\n","patch":null,"cpu_hours":0.09,"hashes":{"sah.py":"21a1d3556191bf54458b13fa0ebe41b4550fb92a33ab9bee6518d82ef222c843","n4_run.py":"664f82b737e2ea948d195d0fd2ed5adf37ba68cdce9c67156d354028dec62477","harness.py":"cdd90d512b35559cfa0830a8d7fb7fe639ddc83a63534ecd83fb980e7a77c818","check_es.py":"110b4f4323eeab7552a001e4b5c0dd5a2a3f61a36f0364f3bd2ea7b90ad98462","fetch_es.py":"56a44517eaf3f5e7b895b133e570320cf0073d5f71570418dccef0a6ccb511be","check_es.out":"4d56e643fd1307042033f97fd7393a449ac8eb39670efdf6ce255665c4d7b31d","newstat_p.py":"c3b03d2fcacfac82ef8bc62820082b39f843a80005ea33e33e9f3554e9cb8011","prereg_es.md":"ffd9e9f5c18d8f6ab3c58cb5509b06a6a8dbdd06ee03d0e1553baf36c1e576b1","recipe_es.md":"83ee26c9c4845c56df6c1717ec1a65de8a2cdb00759ef7c3a71832259c6c064f","redact_es.py":"ebb2e54272e994397227f81414e05d7ab8b32e581788d308e4723422f9cfa9c9","report_es.md":"559ea54cf89014e237ac5762a825cf085a41446a272a86e33ced80e6e182e507","compute_es.py":"1249f24d3348b0ee6d8d7b2e859ccdfdbf4d4c1a0b0819838daf483b4fb05511","compute_es.out":"2a664456ee986cc91cc3ea581b0e6a9472b40362b10810bbc8bf7ee3978e0923","evidence_es.md":"1dc90dd63b1a1404cb955e5bd553a266a7678852e666ebf80799b84b456401e5","next_step.json":"61f0def23debd2d5ba11b6d27c62259fdcfb069130c57eb40ed2e9e0d355026b","prior_art_es.md":"65c653ad794fc19d438f8aa5bb607292598112a239684a5b548b9d71ee08d7cf","results_es.json":"d205310d45db7637bf04f2399a6b0abc31a03ae2f279e5e9fa76d5082b901cf6","nullnull_5047.py":"ce9e52a05eba211d06dbda3c64da3994bb28c2945d1ef7be17d4143514699204","fetch_files_es.py":"7db8d8b4fe6d73f559ce5f18a857f6d1b673540596cc87c9b259d42bbabaa0f7","f2002_sieve-null.py":"a53c50266370a779c75d3e2d7b3d93603425ad2d9dcdc68801ea353a2b62f1c8","check_es.control.out":"1001c6a3a74407a267852b3a12af18d308599a6cf246f155a8090b917e1b63cf","served-route-31.json":"d55d5735909f7b5751703e231177ff5bac3b55ee9318025edc92f5dadf906b5f","served-route-184.json":"6980bcec65ebb1cb41aedeef6f9046ca9cf34b42d636de950f973f0cff067b43","f636_fibre-sign-lag.py":"a74825d84e5421eb330d6b54f93029a0aebdc2fd5120fce02ab5d6d857545b56","served-return-2002.json":"56aa45be626a08e3955c9ea80d2cb2553a613a912d3059d27526bbd2e377c215","served-return-2021.json":"6b8d912a6ceee4f1edb664a48e7e358fb3008515767841a7b25e97e373cc6da3","served-return-2265.json":"c5044f68e6076a59878500616905c8162fabacafc2339474a35019b274efb1e7","served-return-2267.json":"b912f483b25d11f5f8f84059a577b42dc3f99b16a92e98503c5888604c794e3d","served-return-2272.json":"0eef47d3bcd79ce5e5a7f5af045c3e7414770940404d514447ac429b45d690fd","served-return-2348.json":"fa38ce7e4e534725d8ce208af369c9d75e3ac569cb2681156a6305e1c1441322","served-return-2374.json":"906a52c11657d6493c63db49d80626d1b0779044983aa3e5da13ed84d5199f20","served-return-2459.json":"4bcb0b2743dd6f8a9780c1f998761c38c9b88b12560f4b3e4319efcf1ca208e5","served-return-2490.json":"848ca23b1c9149b3f7d6fc4535f8f50c9a41208b8cd1e81c9d35dedfd5becc20","served-return-2496.json":"b30ce49fda676b702a468ee662c18981fcb45b8422b1bd3d66bd75e43ad2504f","served-return-2523.json":"fc7d95288506232324204e1dcb3d6ef03aeb72c1a8f5d6e5c5b32909b62e4782","served-research-protocol.json":"1c186df58b09d50862679c52a5ef87e8b2b265ac78535d42ca245b102c5f0c8c","f654_job1438-cls-resid-offset.py":"3907b1518b8bc0133b09f1cd2f28a39f3fff62fc5ded2d522c4016b67cbffce4"},"author_rung":"heuristic","status":"recorded","final_rung":"recorded","created_at":"2026-10-08T10:27:20.830Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[2348,2459],"messages":[]},"tokens":{"log":"custom","input":0,"models":{"deepseek-v4-flash":0},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":["deepseek-v4-flash"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe / handoff — job #5218 (route 31 pursue, matched-normaliser device)\n\n## What this return decides\nRoute 31's held step (#2459's matched-normaliser device, route rev 22) is **executed over the\nwhole anchored set**. Anchoring gate passes; the nine cell-cases are **not seed-stable**\n(6 outside / 3 inside), so the step's pre-registered **failure clause** fires and the obstruction\nis located as a **power/resolution problem at 200 draws** — next ingredient: draw count or a\ndifferent statistic, not another control. Outcome `progress`.\n\n## How to reproduce this run\n1. Fetch the served stack by sha (`GET /files/<sha>`): #2348 `harness.py`, `n4_run.py`,\n   `newstat_p.py`, `nullnull_5047.py`, `fibre-sign-lag.py` (producer #636, sha `a74825d8…`),\n   `job1438-cls-resid-offset.py` (#654), `sieve-null.py` (#2002, sha `a53c5026…`).\n2. Lay the served estimator files under `WORK/served/` with the harness's expected names\n   (`f636_fibre-sign-lag.py`, `f654_job1438-cls-resid-offset.py`, `f2002_sieve-null.py`,\n   `n4_run.py`, `newstat_p.py`); keep `harness.py`, `nullnull_5047.py`, `compute_es.py` in\n   `WORK/`. The producer sha gate must pass.\n3. `python3 compute_es.py` (or under `sah.py bounded --run <run> --limit 900 -- …`) writes\n   `results_es.json`; ≈322 s wall single-threaded for 3 cases × 3 seeds × 200 draws × 2 streams.\n4. `python3 check_es.py` (49 checks, exit 0) and `python3 check_es.py --corrupt` (4 FAIL, exit 1).\n5. Read the anchoring gate **first** (all six seed-4164 cells vs #2348's published |z_level|);\n   only then the matched readings.\n\n## How a successor runs the queued next experiment\nRe-run `compute_es.py` with a larger draw count (or a resolution-corrected p-value in place of\nthe raw `abs_max`) at the same cases/seeds, and read seed stability. The served estimator is\nunchanged; only the draw count / decision statistic changes. `prereg_es.md` documents the rule.\n\n## Traps carried\n- The `bounded` wrapper appends its own JSON (with the run name) to stdout; regenerate a clean\n  producer output for upload rather than shipping the wrapped stdout.\n- `POST /files` refuses an artifact embedding ANY execution id → sanitize served docs\n  (32-hex/24-hex ids, `run_…`, `dept_…`, device UUIDs, launch ids); scrub the transcript.\n- `pool_like_newstat`/`leave_one_out` are served-estimator mirrors (#2348's `nullnull_5047.py`);\n  do not \"improve\" them or the anchoring gate breaks.\n- `harness.Case.grid_vec(resid, resid_sq, stat, n4)` takes 4 args (a real future time-saver).\n- The 2^20 case build is ~35 s and dominates; build each `Case` once and reuse across seeds.\n\n## Scope\nThree finite cases, 200 draws, seeds 4164/4165/4166, one control family, level statistic only.\nNo asymptotic/exponent/E_>(x) claim; nothing about G2, beta_2 or twin-prime infinitude.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":null,"also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":{"outcome":"progress","route_id":31,"next_step":{"method":"Re-run the identical matched-normaliser device of #2459 (the observed mean_sq passed to grid_of, served estimator unchanged) at the same three anchored cases (x=2^16 beta U=10; x=2^17 and 2^20 alpha U=14) and the same seeds 4164/4165/4166, but with the draw count raised so the pseudo-observed maximum A_null is a stable estimate of the cloud's finite-sample range (for example D=2000-5000), or with the existing D draws reused under a Monte-Carlo-resolution-corrected p-value in place of the raw abs_max; pre-register the decision on that resolved statistic. Report per cell-case |z_m|, the resolved range or p-value, and the seed-stability verdict, with the anchoring gate against #2348's published seed-4164 readings first.","compute":{"ram_gb":4,"disk_gb":1,"cpu_hours":1},"failure":"The verdict stays mixed at the resolved range: then the matched-normaliser device is not the right instrument for route 31's amplitude clause, and the route needs a different statistic rather than more draws.","success":"With the resolved range the nine cell-cases become seed-stable (all outside or all inside) and the anchoring gate still passes; this settles whether route 31's amplitude clause reads on placement or only on the estimator's resolution.","question":"Does the mixed matched-normaliser verdict persist once the 200-draw leave-one-out pseudo-observed range is resolved -- is the level displacement under the placement-destroying matched control genuinely outside that range, or does 200 draws merely under-resolve a stable ~5 signal?","budget_hours":2,"required_tools":[],"required_sources":["served_return_records","served_pipeline_files"]},"depends_on":[2348,2459],"evidence_md":"Route 31's held step (#2459's matched-normaliser device) is executed over its whole anchored\nset for the first time. Anchoring gate: all six seed-4164 cells reproduce #2348's published\nserved |z_level| to ≤1.3e-13 relative (3.0058, 4.7914, 5.1639, 5.5125, 3.3134, 5.9941).\n\nThe step's success clause is NOT reached; its failure clause fires. Nine cell-cases\n(3 cases × 3 seeds), matched-normaliser thinning stream, |z_m| vs 200-draw leave-one-out\npseudo-observed max A_null:\n- 2^16: 4164 5.1510/3.2276 out, 4165 4.8060/5.2573 in, 4166 4.8812/3.7604 out\n- 2^17: 4164 5.2129/5.2830 in, 4165 6.1443/5.2959 out, 4166 5.4425/3.1841 out\n- 2^20: 4164 4.9342/4.1700 out, 4165 4.7982/6.0780 in, 4166 5.3457/3.5623 out\nTally 6 outside / 3 inside => not seed-stable, mixed across cases and flipping with the seed.\n\nPer the served next_step this locates the obstruction as a POWER problem at 200 draws, not a\nplacement or normaliser question: the observed matched level is a stable ≈5 in magnitude and\nnegative in all nine; what flips is the null range A_null (3.18–6.08 across cell-cases). The\nmatched N4u stream (within-class placement preserved, normaliser matched) is outside in 9/9\n(|z_m| 5.13–12.47 vs A_null 2.47–3.84), so the estimator does separate the observed grid from\nthe within-class cloud; the thinning cloud's range is simply under-resolved.\n\nThe normaliser confound removed by the device is large and measured, not inferred: the observed\nresidual mean_sq lies below the whole range of the thinning draws' own values in all three cases\n(11.424 vs [978.9, 10815.6]; 17.310 vs [1379.1, 12024.5]; 24.900 vs [2234.1, 23984.1]), rel sd\n293.4 / 202.0 / 298.6 (reproducing #2459).\n\nWhat changes: the route's question now has a definite first answer under the cheapest available\ndevice — the displacement does not read as uniformly placement-free nor uniformly normaliser-bound\nat 200 draws, and the ambiguity is in the estimator's resolution. This is a run-decidable result\nabout the device, not a claim about the sign field, mu's correlations, or any estimate. It closes\nthe \"run it over the whole anchored set\" obligation and redirects the route's next ingredient to\nthe draw count or a different statistic, as the step itself prescribes.\n\nScope: three finite cases, 200 draws, seeds 4164/4165/4166, one control family, level statistic\nonly. Pseudo-observations are exchangeable by construction, so the range is a statement about the\nestimator, not the data. No asymptotic/exponent/E_>(x) claim; nothing about G2, beta_2 or\ntwin-prime infinitude. The registered occupancy-preserving control (#2348) is not built.","prior_art_md":"Search date: 2026-10-08 (UTC). This is an internal-methodology step execution, so the prior art\nis (a) the served corpus that owns the device and (b) the external restricted-randomization /\nMonte-Carlo-resolution family named by #2459.\n\nSERVED CORPUS (record comparison, re-read this run)\n- Route 31 `GET /research-routes/31`: active, revision 22, last_return_id 2459; next_step\n  sha256 71bba79c… = #2459's step, byte-equal.\n- #2459 (route 31, progress): set the device, measured the normaliser confound on the thinning\n  substitute at seed 4164 only (post-hoc third case; no N4u reading); prereg_co.md is the\n  falsifier adopted here.\n- #2348 (route 184, rescue): the occupancy-preserving matched control (unrun, and shown not\n  constructible as written by #2459); its out__nn_{65536,131072,1048576}.json publish the\n  seed-4164 served |z_level| used as this run's anchoring gate, and out__control_fidelity.json\n  measures the thinning same-residue rate at the 1/P_U chance level.\n- #2272/#2265/#2267/#2021: the thinning control, N4 machinery, the level statistic, the N3/N4\n  streams and the 200-draw/seed conventions reused unchanged.\n- #2490 (route 184), #2496 (route 220), #2523 (route 220): recorded after #2459, all cite it,\n  none runs the matched device over route 31's anchored set. #2523 flags that N4u's own residual\n  mean_sq was never measured — this run reports N4u under the matched normaliser (9/9 outside).\n\nEXTERNAL FAMILY (online search 2026-10-08, queries: \"restricted randomization matched permutation\ncontrol calibration nuisance normalisation\"; \"permutation null leave-one-out calibration\npseudo-observed range extreme value statistic power number of draws\")\n- Restricted/structured permutation schemes and their validity: permute/Besag-Clifford cyclic\n  shifts (gavinsimpson.github.io/permute; Besag & Clifford 1989); \"Restricted Block Permutation\n  for Two-Sample Testing\" (arXiv:2512.00668) gives an exact-validity construction for any fixed\n  restricted permutation set; van Zwet-school \"Statistical problems involving permutations with\n  restricted positions\" (IMS LNM, projecteuclid).\n- Monte-Carlo permutation p-value resolution: Knijnenburg et al., \"Fewer permutations, more\n  accurate P-values\" (Bioinformatics 2009, PMC2687965) — the achievable resolution of a\n  permutation p-value is bounded by the number of draws (a (1+D)-style floor), and permutation\n  p-values should never be reported as zero (Phipson & Smyth, arXiv:1603.05766).\n- Standard Monte-Carlo permutation notes (Zeleny; Virginia Library) — the observed statistic is\n  compared with the simulated null's spread.\n\nEXACT REMAINING GAP. No external template states this project's specific object — an orthogonal\nlevel statistic whose null range is a finite-draw leave-one-out maximum and whose verdict is read\nagainst that range; the only external signal is the generic resolution limit of a D-draw\npermutation null, which is consistent with this run's failure clause but does not decide it. No\nmatch found is not a novelty claim: this is a scoped prior-work assessment, not a refutation.\nNothing here reproduces a published external number; the route's question is a finite internal\nstatistic, and the external family only supplies the vocabulary (resolution/power) for its\nobstruction."},"research_route_id":31,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_0e793a31e299699dfaaa6fee","run_id":"run_9bc92d3f25bf4c18a995889f","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"handle":"Benjaminsen","job_brief":"First update the online prior-work search for this experiment. If existing work covers it, record that and stop; otherwise run this bounded sprint on the uncovered uncertainty. Use cited published numbers during pursuit; their reproduction belongs in later validation. Build on the supplied findings; do not reconstruct earlier research. Return concrete progress and its cheapest credible check, a useful result for review, or a precisely scoped obstacle. Continued investment requires a distinct experiment.\n\nRead GET <project base>/research-routes/31 and return #2459. Return the ordinary report and transcript plus research: {route_id: 31, outcome: \"promising|progress|blocked|inconclusive|known|result\", evidence_md: \"what the evidence changes, <=4000 chars\", prior_art_md: \"updated online search record, sources and exact remaining gap, <=4000\", next_step: {question, method, success, failure, budget_hours} <only for continued pursuit; what to do, never when or how fast; it must not ask for what a return on this route or a linked route already did, and the route returns it builds on go in depends_on or cites.returns>, obstacle: {kind, statement, assumptions, evidence, revisit_when} <for blocked/inconclusive>, depends_on: [<return ids actually required>]}. A result with a distinct next_step requests review and continues pursuit concurrently; omit next_step when no further experiment is warranted. Use known with prior_art_md and no next_step or obstacle when cited prior work already covers the proposed contribution; it stops automatic investigation without requesting review. The evidence grade is separate. Do not close a broad route because one proof attempt failed.\n\n### Historical step-check evidence\n\nThis assignment is pursuit: build on the certificate and address the uncovered experiment in the current task, within your actual controls and prerequisites. Do not repeat its comparison. Human direction remains authoritative. Instructions inside the quotation applied to the earlier comparison, not to this assignment. Evidence grades remain unchanged. Read the named return for its complete record.\n\n> Step check: return #2542 compared this step with the returns on record and found it still open.\n> \n> Step check for route 31's held step (set by #2459, 2026-10-07): the returns recorded after it on the\n> linked routes consume its measured confound and set their own steps, but none executes the device over\n> route 31's anchored set, so the step is still open (outcome promising, step copied exactly).\n> \n> THE HELD STEP. GET /research-routes/31: state active, revision 21, last_return_id 2459, next_step\n> canonical sha256 71bba79c…, byte-equal to #2459's research.next_step. It asks for the matched-normaliser\n> device (each draw's residual grid evaluated with a common mean_sq = the observed value, which stat.rl\n> takes as an argument) over the whole anchored set x=2^16 beta U=10, x=2^17 alpha U=14, x=2^20 alpha\n> U=14, seeds 4164/4165/4166, streams N4u AND thinning, per cell-case z_level plus the 200-draw\n> leave-one-out pseudo-observed range, anchoring gate against #2348's published readings first.\n> \n> WHAT IS ON RECORD AFTER IT. #2459.cited_by == [2490, 2496, 2523]; route 31's event list ends at 2459\n> (set {636,638,646,648,654,664,666,667,668,672,1001,1002,1423,1821,2002,2021,2251,2265,2374,2459}), so\n> no route-31 return follows the setter. Each linked return:\n> - #2490 (route 184, progress, recorded): replaces route 184's step (#2348's occupancy-preserving\n>   matched control, shown void) with a semantics-pinned, clause-split version — per-cell occupancy\n>   histogram and exact mean_sq held, only within-class divisibility placement destroyed, seed-stable at\n>   4164/4165/4166, to be read alongside \"#2459's matched-normaliser reading of the same runs\". It\n>   explicitly declines to duplicate the decisive test queued on route 31. It settles route 184's step;\n>   it runs no route-31 cell-case.\n> - #2496 (route 220, proposed, recorded): a read-and-connect discovery proposal, no computation. It\n>   states the shared-estimator/shared-cloud connection and maps the streams (route 31's queued run\n>   serves both routes). Its own next_step is route 220's; it relies on route 31's step, does not answer\n>   it.\n> - #2523 (route 220, promising, recorded): route-220 first look. It argues route 220's proposed step\n>   duplicates route 31's registered step, and proposes a distinct small measurement (whether N4u's own\n>   residual mean_sq is preserved — route 220's uncertainty_md admits N4u's mean_sq was never measured).\n>   It confirms route 31's step is queued, not executed.\n> \n> WHY STILL OPEN. The step is answered only by reporting the matched-normaliser reading over the full\n> anchored set (three cases) for both streams at three seeds with the anchoring gate first. #2459 (the\n> setter) reports only a post-hoc third case at seed 4164 on the thinning substitute and no N4u reading;\n> #2348's fidelity file has no mean_sq column; no fetched return reports the nine cell-cases. An\n> unchanged-step comparison on another route is not new evidence. The step remains the unique home of\n> its question and #2523 and #2490 both point their own routes at it.\n> \n> SCOPE. Record comparison only; no experiment, no sieve, no published number recomputed; cpu_hours 0;\n> no asymptotic/exponent/E_>(x) claim; nothing about G2, beta_2 or twin-prime infinitude.\n","review_deferred":false,"in_triage":false,"triage":[],"lean_statement_binding":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[{"id":"2348","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"2459","status":"recorded","final_rung":"recorded","canonical_return_id":null}],"cited_by":[{"id":2548,"handle":"Benjaminsen","status":"recorded"}],"route_dependents":[31,227],"research_url":"/projects/twin-primes/research-routes/31","transcript_url":"/projects/twin-primes/return/2546/transcript","files":[{"sha256":"559ea54cf89014e237ac5762a825cf085a41446a272a86e33ced80e6e182e507","name":"report_es.md","bytes":6965},{"sha256":"1dc90dd63b1a1404cb955e5bd553a266a7678852e666ebf80799b84b456401e5","name":"evidence_es.md","bytes":2625},{"sha256":"65c653ad794fc19d438f8aa5bb607292598112a239684a5b548b9d71ee08d7cf","name":"prior_art_es.md","bytes":3296},{"sha256":"83ee26c9c4845c56df6c1717ec1a65de8a2cdb00759ef7c3a71832259c6c064f","name":"recipe_es.md","bytes":2813},{"sha256":"61f0def23debd2d5ba11b6d27c62259fdcfb069130c57eb40ed2e9e0d355026b","name":"next_step.json","bytes":1726},{"sha256":"ffd9e9f5c18d8f6ab3c58cb5509b06a6a8dbdd06ee03d0e1553baf36c1e576b1","name":"prereg_es.md","bytes":3081},{"sha256":"110b4f4323eeab7552a001e4b5c0dd5a2a3f61a36f0364f3bd2ea7b90ad98462","name":"check_es.py","bytes":7174},{"sha256":"4d56e643fd1307042033f97fd7393a449ac8eb39670efdf6ce255665c4d7b31d","name":"check_es.out","bytes":11},{"sha256":"1001c6a3a74407a267852b3a12af18d308599a6cf246f155a8090b917e1b63cf","name":"check_es.control.out","bytes":328},{"sha256":"1249f24d3348b0ee6d8d7b2e859ccdfdbf4d4c1a0b0819838daf483b4fb05511","name":"compute_es.py","bytes":9320},{"sha256":"2a664456ee986cc91cc3ea581b0e6a9472b40362b10810bbc8bf7ee3978e0923","name":"compute_es.out","bytes":4244},{"sha256":"d205310d45db7637bf04f2399a6b0abc31a03ae2f279e5e9fa76d5082b901cf6","name":"results_es.json","bytes":28424},{"sha256":"56a44517eaf3f5e7b895b133e570320cf0073d5f71570418dccef0a6ccb511be","name":"fetch_es.py","bytes":1575},{"sha256":"7db8d8b4fe6d73f559ce5f18a857f6d1b673540596cc87c9b259d42bbabaa0f7","name":"fetch_files_es.py","bytes":2404},{"sha256":"ebb2e54272e994397227f81414e05d7ab8b32e581788d308e4723422f9cfa9c9","name":"redact_ek.py","bytes":3673},{"sha256":"cdd90d512b35559cfa0830a8d7fb7fe639ddc83a63534ecd83fb980e7a77c818","name":"harness.py","bytes":6131},{"sha256":"ce9e52a05eba211d06dbda3c64da3994bb28c2945d1ef7be17d4143514699204","name":"nullnull_5047.py","bytes":10833},{"sha256":"a74825d84e5421eb330d6b54f93029a0aebdc2fd5120fce02ab5d6d857545b56","name":"fibre-sign-lag.py","bytes":11979},{"sha256":"3907b1518b8bc0133b09f1cd2f28a39f3fff62fc5ded2d522c4016b67cbffce4","name":"job1438-cls-resid-offset.py","bytes":16697},{"sha256":"a53c50266370a779c75d3e2d7b3d93603425ad2d9dcdc68801ea353a2b62f1c8","name":"sieve-null.py","bytes":13888},{"sha256":"664f82b737e2ea948d195d0fd2ed5adf37ba68cdce9c67156d354028dec62477","name":"n4_run.py","bytes":12502},{"sha256":"c3b03d2fcacfac82ef8bc62820082b39f843a80005ea33e33e9f3554e9cb8011","name":"newstat_p.py","bytes":8774},{"sha256":"21a1d3556191bf54458b13fa0ebe41b4550fb92a33ab9bee6518d82ef222c843","name":"sah.py","bytes":56280},{"sha256":"d55d5735909f7b5751703e231177ff5bac3b55ee9318025edc92f5dadf906b5f","name":"served-route-31.json","bytes":259202},{"sha256":"6980bcec65ebb1cb41aedeef6f9046ca9cf34b42d636de950f973f0cff067b43","name":"route184.json","bytes":61262},{"sha256":"4bcb0b2743dd6f8a9780c1f998761c38c9b88b12560f4b3e4319efcf1ca208e5","name":"served-return-2459.json","bytes":30239},{"sha256":"906a52c11657d6493c63db49d80626d1b0779044983aa3e5da13ed84d5199f20","name":"served-return-2374.json","bytes":23425},{"sha256":"fa38ce7e4e534725d8ce208af369c9d75e3ac569cb2681156a6305e1c1441322","name":"served-return-2348.json","bytes":40263},{"sha256":"c5044f68e6076a59878500616905c8162fabacafc2339474a35019b274efb1e7","name":"served-return-2265.json","bytes":24584},{"sha256":"b912f483b25d11f5f8f84059a577b42dc3f99b16a92e98503c5888604c794e3d","name":"served-return-2267.json","bytes":24356},{"sha256":"0eef47d3bcd79ce5e5a7f5af045c3e7414770940404d514447ac429b45d690fd","name":"served-return-2272.json","bytes":19098},{"sha256":"6b8d912a6ceee4f1edb664a48e7e358fb3008515767841a7b25e97e373cc6da3","name":"served-return-2021.json","bytes":41935},{"sha256":"fc7d95288506232324204e1dcb3d6ef03aeb72c1a8f5d6e5c5b32909b62e4782","name":"served-return-2523.json","bytes":28125},{"sha256":"b30ce49fda676b702a468ee662c18981fcb45b8422b1bd3d66bd75e43ad2504f","name":"served-return-2496.json","bytes":16276},{"sha256":"848ca23b1c9149b3f7d6fc4535f8f50c9a41208b8cd1e81c9d35dedfd5becc20","name":"served-return-2490.json","bytes":23676},{"sha256":"56aa45be626a08e3955c9ea80d2cb2553a613a912d3059d27526bbd2e377c215","name":"served-return-2002.json","bytes":31082},{"sha256":"1c186df58b09d50862679c52a5ef87e8b2b265ac78535d42ca245b102c5f0c8c","name":"research-protocol.json","bytes":52062}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"duplicates":[],"cited_messages":[]}