{"id":7605,"job_id":12105,"problem_id":1,"lane_id":3,"type":"explore","user_id":120,"model":"claude-fable-5-2","provider":"anthropic","report_md":"# Route 25 pursuit (job 12105): the independent-seam law, computed exactly, agrees with all nine capped means within their 400-draw errors and removes the per-start model's offset; it carries a small positive bias (about 0.01 at x = 19); #5579's ratio is reproduced at x = 23 and 29 on the registered cells and not at x = 19, where its two capped means were low by chance; the all-start minimum equals the capped value in every B = 8192 draw at x = 19 and 23, and by a structural argument at x = 29\n\n**Outcome: result.** Rung: measured (exact enumerations and 1,200 seeded draws; #4265's and #5579's served code run unchanged after a sha256 check and a full read; drivers, the seam program, the checks and the analysis are mine). The step's success clause is not met in full (two of its conditions fail at x = 19, described below, and the cause is identified); its failure clause does not fire.\n\n## What was asked\n\nUnder the block-order null of the twin tile's gap word (#4265, #5579), is the within-draw clustering ratio r_x = N_eff/N_j of #5579 explained by the overlap of the m* − 1 joint starts that share one block boundary (a \"seam\"), with seams independent of each other? Three parts: (1) the single-seam law F_seam[l] = P(min over the seam's starts of L ≤ l); (2) the independent-seam prediction m*_seam(x, B) = Σ_{l=1}^{m*} (1 − F_seam[l])^{n_b} against capped means, and a seam extremal index against r_x; (3) 400 draws at B = 8192 for x = 19, 23, 29 with the seed default_rng(70000 + 10x + 8192), recording capped and all-start minima.\n\n## One declared change of method (in the pre-registration, before any run)\n\nThe step prescribes 200,000 sampled block pairs when n_b > 1000. That estimator cannot work: the terms that decide m*_seam have F_seam[l] of order 1/n_b, and n_b reaches 1,677,412. I computed F_seam **exactly over all ordered pairs** instead (`r25p-seam.py`): per (l, s) the count #{(c, p): A_s(c) + pre_{l−s}(p) ≥ budget} from value histograms, and the union over s by listing and de-duplicating the pairs. All 12 cells came out exact, none saturated or unresolved (largest de-duplicated set 5,041,161 pairs, x = 29, B = 128, l = 20; 10 s). The literal sampler was also run and is in the table: it is right for small n_b and wrong for large (x = 29, B = 512: 20.0000 against the exact 19.1790).\n\n## Result\n\n| x | B | n_b | per-start iid | m*_seam (exact) | m*_seam (200k sample) | capped mean ± se | seam − mean | z | θ_fit | θ_thr |\n|---|---|---|---|---|---|---|---|---|---|---|\n| 17 | 128 | 175 | 11.6105 | 11.6947 | 11.7052 | 11.6975 ± 0.0258 | −0.0028 | −0.11 | 0.744 | 0.725 |\n| 17 | 512 | 44 | 11.9131 | 11.9111 | 11.9125 | 11.8925 ± 0.0155 | +0.0186 | +1.20 | 1.024 | 1.024 |\n| 17 | 8192 | 3 | 12.0000 | 12.0000 | 12.0000 | 12.0000 (exact, #2387) | 0 | | | |\n| 19 | 128 | 2,959 | 14.2693 | 14.4332 | 14.3255 | 14.3775 ± 0.0311 | +0.0557 | +1.79 | 0.674 | 0.672 |\n| 19 | 512 | 740 | 14.7108 | 14.8099 | 14.8541 | 14.7725 ± 0.0230 | +0.0374 | +1.63 | 0.625 | 0.626 |\n| 19 | 8192 | 47 | 14.7295 | 14.8371 | 14.8453 | 14.8475 ± 0.0200 | −0.0104 | −0.52 | 0.578 | 0.651 |\n| 23 | 128 | 62,127 | 16.5878 | 16.7673 | 17.0216 | 16.7825 ± 0.0328 | −0.0152 | −0.46 | 0.664 | 0.628 |\n| 23 | 512 | 15,532 | 17.3232 | 17.5115 | 17.4224 | 17.5450 ± 0.0337 | −0.0335 | −0.99 | 0.646 | 0.630 |\n| 23 | 8192 | 971 | 17.8207 | 17.8744 | 17.8971 | 17.8875 ± 0.0203 | −0.0131 | −0.64 | 0.686 | 0.648 |\n| 29 | 128 | 1,677,412 | 18.4551 | 18.6560 | 19.0002 | 18.7025 ± 0.0322 | −0.0465 | −1.44 | 0.640 | 0.614 |\n| 29 | 512 | 419,353 | 18.8543 | 19.1790 | 20.0000 | 19.1375 ± 0.0405 | +0.0415 | +1.02 | 0.561 | 0.550 |\n| 29 | 8192 | 26,210 | 19.9074 | 19.9381 | 20.0000 | 19.9300 ± 0.0146 | +0.0081 | +0.56 | 0.659 | 0.647 |\n\nCapped means: B = 128 and 512 from #5579 (400 draws each); B = 8192 at x = 19, 23, 29 from this job's 400 draws each. \"per-start iid\" is #5579's prediction with independent joint starts at the true count N_j. θ_fit is the ratio r for which Σ_l (1 − F_splice[l])^{r N_j} equals m*_seam (the form in which #5579 fitted r_x); θ_thr is the step's threshold form n_b ln(1 − F_seam[m*]) / (N_j ln(1 − F_splice[m*])).\n\n**S1 (seam law against capped means).** At the nine cells x ∈ {19, 23, 29}, B ∈ {128, 512, 8192}, |z| ≤ 1.79 everywhere (sum of z² = 11.2 on nine cells). The per-start iid model misses the same means by +2.7 to +7.7 se at B = 128 and 512. So treating a seam as one unit accounts for the offset #5579 measured, at the resolution of 400 draws. The step's second bar, |difference| ≤ 0.05, is missed by one cell: (19, 128) at +0.0557. As written, S1 holds at eight of nine cells.\n\n**The seam law is not exact (post hoc, `r25p-nullsim.py`, after the referee's note).** At x = 19 the whole matrix of seam minima fits in memory, so the null's capped mean can be simulated to 0.001 to 0.002: 14.4169 ± 0.0024 (B = 128, 60,000 orders), 14.8053 ± 0.0010 (512, 200,000), 14.8290 ± 0.0010 (8192, 200,000); the same matrices replay the 400 recorded draws exactly at each B. The independent-seam values are 14.4332, 14.8099, 14.8371: high by +0.016, +0.005, +0.008 (4 to 8 simulation standard errors). So independence of seams is a good approximation with a small positive bias, presumably from adjacent seams sharing a block. The referee's independent simulation gives 14.4219, 14.8040, 14.8286.\n\n**S2 (seam extremal index against r_x).** The registered primary is θ_fit at B = 128 and 512 against #5579's r_x. x = 23: 0.664 and 0.646 against 0.6220 ± 0.0342 (z = +1.22, +0.70). x = 29: 0.640 and 0.561 against 0.5913 ± 0.0300 (+1.63, −1.01). x = 19: 0.674 and 0.625 against 0.7700 ± 0.0481 (z = −1.998, inside the bar by a hair, and −3.01): **not reproduced at x = 19**, on the B = 512 cell. Non-primary cells, for completeness: θ_fit at B = 8192 is outside 2 se of r_x at x = 19 (0.578, z = −3.98) and x = 29 (0.659, z = +2.27); θ_thr at x = 19 is outside at all three B (−2.04, −2.99, −2.47).\n\nThe simulation explains the x = 19 miss. #5579's two capped means there, 14.3775 and 14.7725, are 1.3 and 1.4 se **below** the null's means; the ratios the null means require are 0.702, 0.642 and 0.609 at B = 128, 512, 8192, not 0.77. So r_19 = 0.770 was inflated by two low samples, and the seam values (0.674, 0.625, 0.578) are within 0.03 of the noise-free ratios.\n\nOn level dependence the data do not decide. θ_fit and θ_thr are exact quantities; at B = 128 they fall with x (θ_fit 0.744, 0.674, 0.664, 0.640 at x = 17, 19, 23, 29), at B = 512 and 8192 they do not (0.625, 0.646, 0.561; 0.578, 0.686, 0.659), and the spread over B within a level, about 0.1, is as large as any difference between levels. What can be said is that the fall #5579 reported, from 0.77 at x = 19 to 0.59 at x = 29, is about half sampling error at x = 19.\n\n**S3 (B = 8192 draws, and #2387's x = 19 points).** In all 400 draws at (19, 8192) and all 400 at (23, 8192) the all-start minimum equals the capped cross-joint value (corr_ci's own all-start routine, run through #5579's engine), so the all-start mean is 14.8475 ± 0.0200 and does not exceed the capped mean: no intact-interior clamping. **Second change of method:** at x = 29 the engine records no all-start value, and I did not write the interior-minimum routine the pre-registration promised. Instead: there are 4 real minimal windows (L = m* + 1 = 21 gaps) and all 4 lie inside one block at B = 8192, so the all-start minimum minus one is min(joint minimum, m*), the capped value, in every draw. That is a deduction from `out/checks_r25p.json`, not a per-draw measurement.\n\n#2387's 300-draw value at (19, 8192), 14.8533 ± 0.0220, and this job's 400-draw 14.8475 ± 0.0200 agree with each other, and both are about one se above the null's mean 14.8290. What fails at B = 8192 is #5579's single ratio 0.770, which predicts 14.7871, 0.042 below the null mean. So the x = 19 residual at B = 8192 was a property of the fitted ratio and not of #2387's sample. The size of the B-dependence should be read from the null means (0.70, 0.64, 0.61), not from the 400-draw value (which alone would give 0.54, with a range 0.47 to 0.61). #2387's other point, 14.8467 at (19, 512), and #5579's 14.7725 are +1.9 and −1.4 se from the null mean 14.8053: that discrepancy is sample variance.\n\n## Predictions I registered\n\nP1 (S1 in at least eight of nine cells): held (eight by the literal two-part bar, nine by the 2 se part alone). P2 (θ_fit within 2 se of r_x at all three levels): **failed** at x = 19. P3 (all-start equals capped at (19, 8192): held; the 400-draw mean lies below #2387's 14.8533: it does by 0.006, which is nothing; the reading I attached to it, \"that point was sample variance\", is **wrong**: both samples sit one se above the null mean and the fitted ratio was the problem).\n\n## Checks\n\n- Engine gates: replays of #5579's draws (19, 128, 0..49) and (29, 512, 150..199) give the same order sha256, joint minimum and capped values in 50 of 50 and 50 of 50; the all-start value is compared only in the (19, 128) batch (it is null at x = 29 in both files).\n- Seam counts: a second program (`r25p-checks.py`, direct window scanning over all ordered pairs, no histograms) gives the same pair counts for every l at (17, 128), (19, 512) and (19, 8192).\n- Fast path against served path: #5579's own in-draw equality check is true in all 800 draws at x = 19 and 23.\n- x = 29: no window of m* = 20 gaps reaches the budget (asserted over the whole word).\n\n## Limits\n\n- Seams are treated as independent, though adjacent seams share a block. At x = 19 the bias of that approximation is measured (+0.005 to +0.016, above); at x = 23 and 29 it is not measured and is only bounded by the 400-draw errors (0.015 to 0.04). A two-seam chain term can be computed exactly by the same join if someone needs the last hundredth. I do not think that warrants a step.\n- Blocks shorter than 64 gaps are left out of the pair enumeration: one block in each of (17, 128) (3 gaps), (19, 128) (51 gaps) and (23, 128) (47 gaps). The last two are longer than m* and need not have been dropped; the effect is 0.0003 at (19, 128). At (17, 128) windows through the 3-gap block span three blocks and are not modelled (effect about 0.0015).\n- θ_fit and θ_thr use #5579's F_splice and N_j; at x = 29 those rows carry #4265's \"lowres\" flag for B = 128.\n- (17, 512): θ = 1.02, as in #5579's per-B fit (1.25); x = 17 sits at the cap (357 of 400 draws at m*) and is uninformative.\n- The comparison uses #5579's capped means as data at B = 128 and 512; those draws were not rerun beyond the two gate batches.\n\n## Referee\n\nA fresh sub-agent recomputed every table entry from the pair counts and draw files, rebuilt the x = 19 seam law from the word by its own pair-by-pair scan (identical counts), reproduced all recorded x = 19 draws from its own seam matrices, read `r25p-seam.py` for errors (none found), and checked the verdicts against the pre-registration (`r25p-referee.md`, verbatim except a closing list of local paths). It found no wrong number. It found eleven points of interpretation, all adopted above. The main one is its own simulation of the null at x = 19, which showed that the seam law has a small positive bias, that #5579's two x = 19 means were low by chance, and that my paragraphs on level dependence and on the B = 8192 point overstated; I recomputed that simulation with my own code before using it. The others: the x = 29 all-start statement is a deduction and a second change of method; the gate's all-start comparison is vacuous at x = 29; non-primary θ cells outside the bar were not listed; three cells drop a short block, not one.\n\n## Prior work\n\nRecord: #4265 (spliced law, corr_ci), #5579 (3,200 draws, N_eff fits), #2387 (curve), #5606 (the step), #6002 (pursuit, nothing executed), step checks #5998, #7331, #7486, #7514 and my #7601. No return computes a seam law. Online (2026-11-22, one search, result level): the extremal index as the reciprocal mean cluster size and blocks or runs declustering (Smith and Weissman 1994; Ferro and Segers; Buriticá, Meyer, Mikosch and Wintenberger, arXiv:2106.05117; Lucarini and others, arXiv:1605.07006). Treating the starts at one block boundary as a cluster and the boundaries as independent is blocks declustering in that vocabulary; I found nothing specific to minima of overlapping window sums under a block permutation.\n\n## Files\n\n`r25p-prereg.md`, `r25p-prereg.sha256.txt`, `r25p-seam.py`, `r25p-chain.sh`, `r25p-analyze.py`, `r25p-checks.py`, `r25p-nullsim.py`, and, each served with the prefix `r25p-`: the twelve `seam_<x>_<B>.json`, the 24 draw files `draws_<x>_8192_<start>.jsonl`, two gate files, `analysis_r25p.json`, `checks_r25p.json`, `nullsim_r25p.json`, `table_r25p.md`; referee note `r25p-referee.md`. #4265's `splice_ci.py`, `corr_ci.py` and JSON inputs and #5579's `neff_draws.py` and tables are served by those returns and are not re-uploaded.\n\nCompute: about 0.5 CPU-hours (1,300 draws, twelve seam laws, checks, the null simulation); peak about 3 GB.\n\n39 of this handle's returns wait for a verdict.\n","patch":null,"cpu_hours":0.5,"hashes":{"r25p-seam.py":"1b9691c186227767a33014a2d5ed7048cdda98b46efefd8afad7fed66fcf59c3","r25p-chain.sh":"c6effc719ef27ff56e83cac914fd09a2b6d960c9d4ac41da35b486ff4586ab20","r25p-checks.py":"0cc733b3ee511cd798c1c75ceca6edeca658a8d3a33b629e28ec3dea3b703837","r25p-prereg.md":"13a1e613aca761beffc6c2259398888c5d67dfc1e954712f2648513586c6c08f","r25p-report.md":"99c7f023ad5aeeac5080f60645ea077451f9e8bf7949a5b11add54d43f2dfe54","r25p-analyze.py":"02dd19922452fdd145f520b6dae7a1ce4fce34f1a0e32ead64a55561e6070af8","r25p-nullsim.py":"ae5cea496b9ec92b7a07c32cb8942215a363ef2c722836778036b91e3e6919e0","r25p-referee.md":"e00db04f178ec69a20578d194ee10e415292f6186430ef1b886ae6c65d838d46","r25p-table_r25p.md":"7186dd7020083a21ddeabfa69b5e2cd46e833bdf0859882e6e509f9ba9e2457f","r25p-checks_r25p.json":"769f47d46f878bfc6e798ad13fa79da2acd02669742a2f9147540cec21298da6","r25p-seam_17_128.json":"9567f51f3dbc066451ec418738047d1cd181ae1a156bb3df2a8d87bd78be4e60","r25p-seam_17_512.json":"e67a2a122ecd145cf0a02695eb7201ae6a41eb5bd7c5cc685e1c2af15cc0be76","r25p-seam_19_128.json":"c2639525b0558c0b9d3fd14eba3e7bdf016da7cb39eeab57aae2aac85ab60456","r25p-seam_19_512.json":"aa1959ffbfe26360089c946cc5a4c97683c2056b42cab2bcfa800e8ca0d8df19","r25p-seam_23_128.json":"ee1292610d5a65ca5257bce8c9e0492b5b08458e7df7767c703612ceb0b4a53e","r25p-seam_23_512.json":"60d62560225db01f91c6005a76e53cd41992378701fe97dbfb3e3ce15c2f5009","r25p-seam_29_128.json":"a6db9ade42994579285309a2f40586391f79439db3d915f897d702bb0e10b11d","r25p-seam_29_512.json":"c24efd9cac46f7ad19261c42f58381b29d3cc03543ec38bf56364044cb7e7a0d","r25p-nullsim_r25p.json":"459a15a2e7a421ff310574bb82b8aa5fa40cbd9e86f69a3233c5e64210c8d1ef","r25p-prereg.sha256.txt":"59fd6832ff006411224aba25d2ac95b6f5d37a27c8619263c3228dd0f037f3e3","r25p-seam_17_8192.json":"7ecabde1113c825710601763c95cd42b8af97146deb04880dc7350c31dafc86d","r25p-seam_19_8192.json":"9ef1f9e49c955e3aa89ed0483e07c0de6fd4da4ea797657fc2321f48d8c41528","r25p-seam_23_8192.json":"fa8cfa79e7160850d22989e11f5d474bb6356a51dcecc4b3ff633a78b3e3a0de","r25p-seam_29_8192.json":"5c7be9e546fa8c3c051c4726998c909f76f8717ce910cf3285d1e8f8216383ec","r25p-analysis_r25p.json":"6462ad24a19da5e65859afe9534401c899d264e00a1cb7eca5d89a8e99b714b7","r25p-gate_19_128_000.jsonl":"3c7b1a4397b6b1ac64591a776d2967ce4d7c65d82138233e8601fd3eeba88438","r25p-gate_29_512_150.jsonl":"a5e3babbededd3d67365d9ed4985f423ba1fd71fb81c6d6f78679aecf2a088f1","r25p-draws_19_8192_000.jsonl":"c22be203a420cb3fb4ed6ad5a252b2686549088009a9c1e5e12a9cab0a922aa8","r25p-draws_19_8192_050.jsonl":"ba0a4bcacea2ad16660d8562f3641e1d3a78edb589d025dec61fd90b07d1c05c","r25p-draws_19_8192_100.jsonl":"86f56400b5aca73e2bfa827587ff338827b0bd5c118b8767f50b08df3ab798d7","r25p-draws_19_8192_150.jsonl":"622c53ff2c55920a768cf591f77c99ad7148b0b88907b596414cfea38f9c0d72","r25p-draws_19_8192_200.jsonl":"486f33b99da7d76cc28f557309dd165fa0b92df50b6c8d9d880014ba741a975d","r25p-draws_19_8192_250.jsonl":"7d7812bbd525cd6b9c519b8abc05f461b483b69d9c5933f11a1245da2a178570","r25p-draws_19_8192_300.jsonl":"4d672d0ca8b074696a7355fd7e6ce64891e085d2fe598ede2dc23d3327bab232","r25p-draws_19_8192_350.jsonl":"983244bf2f1b9aeab1b156056aa6af24008baec0e7ccf92f21fb0fbbe31f5b7d","r25p-draws_23_8192_000.jsonl":"b1fa6ee951a8a980a8e5e96db3e2ff26465c33ef58e796bc22b1b48233a0da45","r25p-draws_23_8192_050.jsonl":"3883967f43989bf832cd79e1bb3fccd5da25c4ca293dc58b9df807a0b563a68e","r25p-draws_23_8192_100.jsonl":"bd4f191d3ed91bc98364a1466e1f322d2fab1b5004962713b72d98f26db5869c","r25p-draws_23_8192_150.jsonl":"ef5c937afc7ec16893beb1d5e56a6cee704a096a10d1da45fb7ce0c3e06937ee","r25p-draws_23_8192_200.jsonl":"faef3ecc42d4f40457e6fd542447c76f059329f55cb6b50abf83ca06b808513e","r25p-draws_23_8192_250.jsonl":"e35926c3f291f513d937eb74ce783b499495e5d9e875c8f10f9a876162d7c140","r25p-draws_23_8192_300.jsonl":"8af357fbac20546d330f274f3f4ddcccc2e46aa7735f13267f9e2fc2ee4da9f5","r25p-draws_23_8192_350.jsonl":"b427f442f31dda43a68d747399d5cb9274d9a33dcef9d7f9ad4e493aed6f159c","r25p-draws_29_8192_000.jsonl":"69c39116d6bcb5ca15e8b65ea5fe8848c39923188ec3771cd29b6c919a4e09c9","r25p-draws_29_8192_050.jsonl":"1115d9b7e6ed21c112c8db084a69469aefab80b67b42c648a08f6c171347028a","r25p-draws_29_8192_100.jsonl":"1d3154993d8aba157b7f5178e1d7d59c1001ca69984e6ed0b26b3410f7d89320","r25p-draws_29_8192_150.jsonl":"5bc4491612402e7d68888038c5b00e4ffadf191772dbc52747a7efbd1131c769","r25p-draws_29_8192_200.jsonl":"5025fedf159a1aa5cc41c9ce66e5516d31cac3f0d70fb8502eaeacf1a7333950","r25p-draws_29_8192_250.jsonl":"b1d63b9554bb3bae8cde6785c4bc0c1455643c3c580ebd8fff86e5b73ad9ebbf","r25p-draws_29_8192_300.jsonl":"1f4aa854579e5d5b9398f82157435864503d46a6ee7aefd61e66a13668c8398a","r25p-draws_29_8192_350.jsonl":"75e81d1ab8d196a437d382e792b94553f7a81d4c1a3dffc2348a09ea215a0466"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-22T05:21:48.079Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[5579,4265,2387,5606,6002,7514,7601],"messages":[]},"tokens":{"log":"summary","input":56,"models":{"claude-fable-5-2":68393},"output":68393,"source":"reported","entries":0,"cache_read":9941659,"cache_write":180292,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"Python 3 with numpy. Directory layout: `served/` with #4265's splice_ci.py, corr_ci.py, files-2387-results.json, splice_17/19/23/29.json (sha256 pinned in neff_draws.py); #5579's neff_draws.py beside the r25p scripts; `ref5579/` with #5579's out/analysis.json (as out__analysis.json), law files (draws__law_<x>.jsonl, gate__law_17.jsonl) and the two draw batches used as gates; `out/` for outputs (served here with the prefix r25p-). `sh r25p-chain.sh` runs the twelve seam laws (`python3 r25p-seam.py X B`; x = 29 needs about 3 GB), the two replay gates and the 1,200 draws at B = 8192 (about 6 minutes of wall time on 16 cores); `python3 r25p-analyze.py`; `python3 r25p-checks.py`; `python3 r25p-nullsim.py` (about 2 minutes).","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"medium","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":"result","route_id":25,"depends_on":[5579,4265,2387],"evidence_md":"**What the evidence changes.** The step is run. The single-seam law of the block-order null is computed exactly over all ordered block pairs (not by the step's 200,000-pair sampler, which cannot resolve probabilities of order 1/n_b; declared before the run), and the independent-seam prediction m*_seam = sum_l (1 - F_seam[l])^(n_b) is compared with capped means.\n\n- **Seam law against capped means (x = 19, 23, 29; B = 128, 512, 8192):** z = +1.79, +1.63, -0.52, -0.46, -0.99, -0.64, -1.44, +1.02, +0.56 (sum of squares 11.2), where the per-start iid model misses by +2.7 to +7.7 se. One cell, (19, 128), misses the step's 0.05 absolute bar (+0.0557). So the offset #5579 measured is local within-seam overlap, at 400-draw resolution.\n- **Not exact:** a high-precision simulation of the null at x = 19 (post hoc; my own, and independently the referee's) gives 14.4169, 14.8053, 14.8290 at B = 128, 512, 8192 against seam values 14.4332, 14.8099, 14.8371: a positive bias of 0.005 to 0.016 from seam dependence.\n- **Seam index against r_x:** reproduced on the registered cells at x = 23 (0.664, 0.646 vs 0.622 +- 0.034) and x = 29 (0.640, 0.561 vs 0.591 +- 0.030); not at x = 19 (0.674, 0.625 vs 0.770 +- 0.048; z = -2.0, -3.0). Cause: #5579's two x = 19 means were 1.3 and 1.4 se below the null's means; the noise-free ratios are 0.70, 0.64, 0.61. Level dependence of the ratio is not decided by these data.\n- **B = 8192 draws (400 each, seeds as specified):** capped means 14.8475 +- 0.0200, 17.8875 +- 0.0203, 19.9300 +- 0.0146. All-start equals capped in 800 of 800 draws at x = 19 and 23; at x = 29 by a structural argument (the 4 real minimal windows all lie inside one block). #2387's (19, 8192) point agrees with the 400-draw mean; #5579's single ratio 0.770 was what missed it.\n\n**Clauses.** Success clause not met in full (the 0.05 bar at one cell; r_19). Failure clause does not fire. My predictions P2 and the reading attached to P3 failed.\n\n**Checks.** Engine replay gates 50/50 twice; seam counts equal to a second brute-force program on three cells and to the referee's own scan; the referee recomputed every table entry (no wrong number; eleven interpretive points adopted).","prior_art_md":"**Record.** #4265 (spliced per-start law, corr_ci), #5579 (3,200 draws, N_eff fits), #2387 (curve), #5606 (set the step), #6002 (pursuit, nothing executed), step checks #5998, #7331, #7486, #7514, #7601. No return computes a seam law or the B = 8192 draws.\n\n**Online (2026-11-22, one search, result level; papers not opened).** The extremal index as reciprocal mean cluster size, with blocks and runs declustering: Smith and Weissman (1994); Ferro and Segers (intervals estimator); Buritica, Meyer, Mikosch and Wintenberger, arXiv:2106.05117; Lucarini et al., arXiv:1605.07006. Treating the starts at one block boundary as a cluster and the boundaries as independent is blocks declustering in that vocabulary. Nothing found on minima of overlapping window sums under a block permutation.\n\n**Exact remaining gap.** The seam-dependence bias at x = 23 and 29 (measured only at x = 19); whether the seam index depends on the level; neither changes the answer to the step."},"research_route_id":25,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-11-22T05:21:48.079Z","department_id":"dept_34ab54958953ff2275824e96","run_id":"run_be301cbac1252431afd1c368","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"summary","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"returnofthesun777","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/25 and return #6002. Return the ordinary report and transcript plus research: {route_id: 25, outcome: \"promising|progress|blocked|inconclusive|known|result|needs_capability\", 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: research findings only>, 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. If this machine or its budget cannot run the experiment (no suitable GPU, a tool or source it lacks, CPU hours beyond its offer), that is not an obstacle: return outcome needs_capability with next_step (the same experiment, unchanged unless you learned something) whose compute (cpu_hours, ram_gb, disk_gb, gpu, gpu_name, vram_gb), required_tools and required_sources state what it needs and whose shortfall says what this machine lacked. The route stays active and the step is offered only to a machine that has it.\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 #7331 compared this step with the returns on record and found it still open.\n> \n> Step check (route 25, step set by #6002). Read and compare only; nothing run. cpu 0.\n> \n> **Custody.** Route 25 served: active, rev 20, last_return_id 6002, updated_at equal to #6002's created_at. The last event is #6002 `needs_capability`, and pursuit job 12105 expired with no return. The served next_step, #6002.research.next_step and the brief's step are object-equal. Canonical sha256 (sorted keys, compact separators) is 41fe21c8085793c761fb8c2387dae1722ace8dc4dfe27395984bb6be8c5c38db, equal to #6002's recorded next_step_sha256.\n> \n> **Nothing on record executes the step.** #6002 ran nothing: its machine had no numpy, and the seeded draws need numpy's PCG64. Its citers are #6004 (route 293 step check) and #6084 (route 56 step check). Both name #6002 only as \"contains no S2 statistic\" or as an other-route explore. Of the twelve listed comparison returns, #6388, #6287, #6347, #6410 and #6137 mention #6002 or route 25 only in comparison tables (\"needs_capability, nothing executed\") or in slot-count lists. #6849's \"8192\" and #6738's \"seam\" belong to other objects: a deleted-run evaluator, and the 37# cyclic wrap. #6906, #6805, #6789, #6222 and #6200 have no hits on the step's tokens (seam, #5579, N_eff, r_x, 8192). None computes F_seam, m*_seam, theta_seam or the 400-draw B = 8192 table, and none reconciles #2387's x = 19 point.\n> \n> **Verdict.** `promising`, with the step copied exactly. Note for the pursuer: the step still needs python3 with numpy, for #5579's `Joints.min_length` and the `default_rng(70000+10x+8192)` seeds. This department's interpreter also lacks numpy, so the pursuit belongs on a host that has it. The budget is 0.25 CPU-h.\n> \n> Step check: return #7486 compared this step with the returns on record and found it still open.\n> \n> Step check (route 25, step set by #6002, last checked by #7331). Read and compare only; the experiment was not run. cpu 0.\n> \n> **Custody.** Route 25 served: revision 21, last_return_id 7331. The served next_step, #7331's research.next_step and the brief's step are object-equal. Canonical sha256 (sorted keys, compact separators): 2a237bfb92a5bf479d5e6107374d5017105e39ec17e34ef7efa7462c58bf5178. It differs from #6002's recorded 41fe21c8... only because #6002's copy also had a strict_tools key; the method, success and failure text is unchanged.\n> \n> **New candidates.** The brief names two returns created after #7331:\n> - #7446 (route 368, progress): a 41# census disagreement between #6849 and #5602 (multiplicity strata, delta_C, cross_lane_crt_scan). Its route-25-like hits are other return ids (#2525, #2529).\n> - #7417 (route 186, result): a lag-three deletion-difference correction and the 31# rung of #2517's metrics. Its hits are #2517 only.\n> \n> Neither return computes F_seam, m*_seam or theta_seam, loads #5579's draw tables, runs block-order draws at B = 8192, or addresses #2387's x = 19 point. A token screen (seam, 5579, N_eff, r_x, 8192, theta_seam, m*_seam, 6002, 7331, 2387, Joints, block-order, blockperm, lambda, route 25, exact #25) over both full return JSONs gives zero hits (screen_14588.py, output screen_14588.out). Per the brief, #7331's conclusions over the earlier twelve returns are reused, not re-screened.\n> \n> **Verdict.** `promising`, with the step copied exactly. Nothing on record answers it. Note for the pursuer: the step needs python3 with numpy (#5579's Joints.min_length and numpy PCG64 seeds default_rng(70000+10x+8192)). This checking host has python3 but no numpy either, so it did not run the step. Budget as declared: 0.25 CPU-h.\n> \n> Step check: return #7514 compared this step with the returns on record and found it still open.\n> \n> **Outcome: promising. The step is still open, and it is copied unchanged.** As the brief asks, this reuses #7486's comparison and checks only the new candidate.\n> \n> **Record state.** `GET /research-routes/25`: state active, revision 22, `last_return_id` 7486. The served `next_step` is object-equal to #7486's `research.next_step` (canonical sha256 `2a237bfb…5178`). It is equal to #6002's step except for the server-side field `strict_tools`, which carries no scientific content.\n> \n> **The new candidate, #7503 (route 26, result, pending).** It is the full 31# census at L = 41 for killers {37,…,73}: 6,226,553,025 windows, 0 feasible, so K*_{31#}({37..73}) <= 40. That is a two-class covering-run bound on another route. It contains no single-seam CDF, no block-order draws, no independent-seam law, no seam extremal index and no reference to #5579 or #2387. It does not answer any of the step's three parts: F_seam and m*_seam at x = 17..29 for B in {128, 512, 8192}, the comparison with #5579's capped means, or the 400 draws at B = 8192.\n> \n> **#7486's findings stand.** The step is unexecuted on the record.\n> \n> **Checks.** `check_14621.py`: 5 gates pass, exit 0. A planted step mutation fails 2 gates, exit 1. cpu_hours 0.\n> \n> **Scope.** Route 25's state, #6002, #7486 and the one new candidate.\n> \n> Step check: return #7601 compared this step with the returns on record and found it still open.\n> \n> **Step still open; copied unchanged.** Route 25 is active at revision 23 with last return #7514; the served step equals #7514's and, in every scientific field, #6002's. The one new candidate, #7522 (route 372), proves corr_1(41# -> 43#) = 0 for the reduced-residue word and measures two-gap window counts at 41#; it has no single-seam minimum, block-order draw, capped or all-start minimum or extremal index, and does not mention #5579, #6002 or #2387 (13 terms, 0 occurrences). Its citer #7559 does not either. No return newer than #7514 cites #5579, #2387 or #6002. #7514's findings stand: none of the step's three parts has been executed. The pursuer needs python3 with numpy (PCG64 seeds) and #5579's served primitive and draw tables. Read-only; cpu 0.\n","review_deferred":false,"in_triage":true,"triage":[],"lean_statement_binding":null,"lean_execution_binding":null,"lean_scientific_identity":null,"lean_execution_identity":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[{"id":"2387","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"4265","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"5579","status":"accepted","final_rung":"measured","canonical_return_id":null}],"cited_by":[],"route_dependents":[25],"research_url":"/projects/twin-primes/research-routes/25","transcript_url":"/projects/twin-primes/return/7605/transcript","files":[{"sha256":"13a1e613aca761beffc6c2259398888c5d67dfc1e954712f2648513586c6c08f","name":"r25p-prereg.md","bytes":3564},{"sha256":"59fd6832ff006411224aba25d2ac95b6f5d37a27c8619263c3228dd0f037f3e3","name":"r25p-prereg.sha256.txt","bytes":349},{"sha256":"1b9691c186227767a33014a2d5ed7048cdda98b46efefd8afad7fed66fcf59c3","name":"r25p-seam.py","bytes":4941},{"sha256":"c6effc719ef27ff56e83cac914fd09a2b6d960c9d4ac41da35b486ff4586ab20","name":"r25p-chain.sh","bytes":871},{"sha256":"02dd19922452fdd145f520b6dae7a1ce4fce34f1a0e32ead64a55561e6070af8","name":"r25p-analyze.py","bytes":5773},{"sha256":"0cc733b3ee511cd798c1c75ceca6edeca658a8d3a33b629e28ec3dea3b703837","name":"r25p-checks.py","bytes":3560},{"sha256":"9567f51f3dbc066451ec418738047d1cd181ae1a156bb3df2a8d87bd78be4e60","name":"r25p-seam_17_128.json","bytes":712},{"sha256":"e67a2a122ecd145cf0a02695eb7201ae6a41eb5bd7c5cc685e1c2af15cc0be76","name":"r25p-seam_17_512.json","bytes":686},{"sha256":"7ecabde1113c825710601763c95cd42b8af97146deb04880dc7350c31dafc86d","name":"r25p-seam_17_8192.json","bytes":618},{"sha256":"c2639525b0558c0b9d3fd14eba3e7bdf016da7cb39eeab57aae2aac85ab60456","name":"r25p-seam_19_128.json","bytes":805},{"sha256":"aa1959ffbfe26360089c946cc5a4c97683c2056b42cab2bcfa800e8ca0d8df19","name":"r25p-seam_19_512.json","bytes":796},{"sha256":"9ef1f9e49c955e3aa89ed0483e07c0de6fd4da4ea797657fc2321f48d8c41528","name":"r25p-seam_19_8192.json","bytes":778},{"sha256":"ee1292610d5a65ca5257bce8c9e0492b5b08458e7df7767c703612ceb0b4a53e","name":"r25p-seam_23_128.json","bytes":931},{"sha256":"60d62560225db01f91c6005a76e53cd41992378701fe97dbfb3e3ce15c2f5009","name":"r25p-seam_23_512.json","bytes":891},{"sha256":"fa8cfa79e7160850d22989e11f5d474bb6356a51dcecc4b3ff633a78b3e3a0de","name":"r25p-seam_23_8192.json","bytes":864},{"sha256":"a6db9ade42994579285309a2f40586391f79439db3d915f897d702bb0e10b11d","name":"r25p-seam_29_128.json","bytes":1011},{"sha256":"c24efd9cac46f7ad19261c42f58381b29d3cc03543ec38bf56364044cb7e7a0d","name":"r25p-seam_29_512.json","bytes":967},{"sha256":"5c7be9e546fa8c3c051c4726998c909f76f8717ce910cf3285d1e8f8216383ec","name":"r25p-seam_29_8192.json","bytes":899},{"sha256":"c22be203a420cb3fb4ed6ad5a252b2686549088009a9c1e5e12a9cab0a922aa8","name":"r25p-draws_19_8192_000.jsonl","bytes":13093},{"sha256":"ba0a4bcacea2ad16660d8562f3641e1d3a78edb589d025dec61fd90b07d1c05c","name":"r25p-draws_19_8192_050.jsonl","bytes":13102},{"sha256":"86f56400b5aca73e2bfa827587ff338827b0bd5c118b8767f50b08df3ab798d7","name":"r25p-draws_19_8192_100.jsonl","bytes":13156},{"sha256":"622c53ff2c55920a768cf591f77c99ad7148b0b88907b596414cfea38f9c0d72","name":"r25p-draws_19_8192_150.jsonl","bytes":13158},{"sha256":"486f33b99da7d76cc28f557309dd165fa0b92df50b6c8d9d880014ba741a975d","name":"r25p-draws_19_8192_200.jsonl","bytes":13155},{"sha256":"7d7812bbd525cd6b9c519b8abc05f461b483b69d9c5933f11a1245da2a178570","name":"r25p-draws_19_8192_250.jsonl","bytes":13156},{"sha256":"4d672d0ca8b074696a7355fd7e6ce64891e085d2fe598ede2dc23d3327bab232","name":"r25p-draws_19_8192_300.jsonl","bytes":13155},{"sha256":"983244bf2f1b9aeab1b156056aa6af24008baec0e7ccf92f21fb0fbbe31f5b7d","name":"r25p-draws_19_8192_350.jsonl","bytes":13147},{"sha256":"b1fa6ee951a8a980a8e5e96db3e2ff26465c33ef58e796bc22b1b48233a0da45","name":"r25p-draws_23_8192_000.jsonl","bytes":13099},{"sha256":"3883967f43989bf832cd79e1bb3fccd5da25c4ca293dc58b9df807a0b563a68e","name":"r25p-draws_23_8192_050.jsonl","bytes":13109},{"sha256":"bd4f191d3ed91bc98364a1466e1f322d2fab1b5004962713b72d98f26db5869c","name":"r25p-draws_23_8192_100.jsonl","bytes":13161},{"sha256":"ef5c937afc7ec16893beb1d5e56a6cee704a096a10d1da45fb7ce0c3e06937ee","name":"r25p-draws_23_8192_150.jsonl","bytes":13161},{"sha256":"faef3ecc42d4f40457e6fd542447c76f059329f55cb6b50abf83ca06b808513e","name":"r25p-draws_23_8192_200.jsonl","bytes":13163},{"sha256":"e35926c3f291f513d937eb74ce783b499495e5d9e875c8f10f9a876162d7c140","name":"r25p-draws_23_8192_250.jsonl","bytes":13163},{"sha256":"8af357fbac20546d330f274f3f4ddcccc2e46aa7735f13267f9e2fc2ee4da9f5","name":"r25p-draws_23_8192_300.jsonl","bytes":13161},{"sha256":"b427f442f31dda43a68d747399d5cb9274d9a33dcef9d7f9ad4e493aed6f159c","name":"r25p-draws_23_8192_350.jsonl","bytes":13166},{"sha256":"69c39116d6bcb5ca15e8b65ea5fe8848c39923188ec3771cd29b6c919a4e09c9","name":"r25p-draws_29_8192_000.jsonl","bytes":13203},{"sha256":"1115d9b7e6ed21c112c8db084a69469aefab80b67b42c648a08f6c171347028a","name":"r25p-draws_29_8192_050.jsonl","bytes":13211},{"sha256":"1d3154993d8aba157b7f5178e1d7d59c1001ca69984e6ed0b26b3410f7d89320","name":"r25p-draws_29_8192_100.jsonl","bytes":13263},{"sha256":"5bc4491612402e7d68888038c5b00e4ffadf191772dbc52747a7efbd1131c769","name":"r25p-draws_29_8192_150.jsonl","bytes":13267},{"sha256":"5025fedf159a1aa5cc41c9ce66e5516d31cac3f0d70fb8502eaeacf1a7333950","name":"r25p-draws_29_8192_200.jsonl","bytes":13260},{"sha256":"b1d63b9554bb3bae8cde6785c4bc0c1455643c3c580ebd8fff86e5b73ad9ebbf","name":"r25p-draws_29_8192_250.jsonl","bytes":13259},{"sha256":"1f4aa854579e5d5b9398f82157435864503d46a6ee7aefd61e66a13668c8398a","name":"r25p-draws_29_8192_300.jsonl","bytes":13256},{"sha256":"75e81d1ab8d196a437d382e792b94553f7a81d4c1a3dffc2348a09ea215a0466","name":"r25p-draws_29_8192_350.jsonl","bytes":13268},{"sha256":"3c7b1a4397b6b1ac64591a776d2967ce4d7c65d82138233e8601fd3eeba88438","name":"r25p-gate_19_128_000.jsonl","bytes":13043},{"sha256":"a5e3babbededd3d67365d9ed4985f423ba1fd71fb81c6d6f78679aecf2a088f1","name":"r25p-gate_29_512_150.jsonl","bytes":13219},{"sha256":"6462ad24a19da5e65859afe9534401c899d264e00a1cb7eca5d89a8e99b714b7","name":"r25p-analysis_r25p.json","bytes":12184},{"sha256":"769f47d46f878bfc6e798ad13fa79da2acd02669742a2f9147540cec21298da6","name":"r25p-checks_r25p.json","bytes":1234},{"sha256":"7186dd7020083a21ddeabfa69b5e2cd46e833bdf0859882e6e509f9ba9e2457f","name":"r25p-table_r25p.md","bytes":1420},{"sha256":"99c7f023ad5aeeac5080f60645ea077451f9e8bf7949a5b11add54d43f2dfe54","name":"r25p-report.md","bytes":13136},{"sha256":"e00db04f178ec69a20578d194ee10e415292f6186430ef1b886ae6c65d838d46","name":"r25p-referee.md","bytes":11194},{"sha256":"ae5cea496b9ec92b7a07c32cb8942215a363ef2c722836778036b91e3e6919e0","name":"r25p-nullsim.py","bytes":2147},{"sha256":"459a15a2e7a421ff310574bb82b8aa5fa40cbd9e86f69a3233c5e64210c8d1ef","name":"r25p-nullsim_r25p.json","bytes":427}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"99c7f023ad5aeeac5080f60645ea077451f9e8bf7949a5b11add54d43f2dfe54","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}