{"id":7615,"job_id":13541,"problem_id":1,"lane_id":3,"type":"explore","user_id":120,"model":"claude-fable-5-2","provider":"anthropic","report_md":"# Route 425 pursuit (job 13541): the exact N3 covariance evaluator matches the recorded original-range Monte Carlo on identical rows (17 cells, χ² = 16.0) and gives a baseline table on #6555's 500 anchors. The model's own mean is 14 to 35 % below the corrected main term, so N3 with its bias retained is not a valid same-anchor baseline; recentred, the measured squared errors are 1.42, 1.26 and 1.06 times the N3 variance at u = 3, u*, 5\n\n**Outcome: result.** Rung: measured (exact evaluations of a stated model on frozen rows; no fitted constant). Nothing here is a census error bound, and #6555's refutation of the fixed 0.6204 kernel is untouched.\n\n## What was done, against the step\n\n1. **Row availability (the step's first instruction).** #6555 serves no row-level table. Its served driver `route425-firstlook.js` (sha256 e5814e4b…) was rerun unchanged (18 s): the stdout is byte-identical to the served one. A derived copy, identical except for an appended eight-line block that writes the rows, froze for each of the 500 anchors Q, Q′, C, T, nR and, at u = 3, u*, 5: K, X(K), the naive mean and the corrected main term Xcorr (`rows_6555.json`); its stdout is byte-identical too.\n2. **Exact evaluator** of #6809's equations (1) to (3) (`p425-eval.py`): L = the first K active primes, R = the remaining active primes up to Q, A = the window's candidates. Pair lags are multiples of 6 and J(a, b) depends on d = b − a only through the primes dividing d, d − 2, d + 2 (no prime ≥ 7 divides two of them), so Var = Cm(1 − m) + 2Σ_d M(d)(J(d) − m²) with nine sieve-accumulated correction arrays; one pass over the primes serves all depths. 500 anchors in 4 s.\n3. **Comparison with the existing original-range N3 Monte Carlo on identical rows** (added after the pre-registered part, once I had located the source: the served producer `measure-roughpair-null-0829.js`, sha256 04edd301…, whose recorded output is in its own comments; the producer was not run). Its 40-anchor subsample per band is deterministic (every ⌊n/40⌋-th anchor of the band), so the same rows can be evaluated exactly.\n4. **Matched-weight pooled table** on the new anchors with μ, V and (μ − Xcorr)² separate.\n\n## Gates\n\n- **G1.** The direct pair formula reproduces #6809's two fixtures as exact fractions (30/1001 and 29130/1002001; 40/1001 and 44446/1002001). The referee confirmed the model formulas by exact enumeration of all phases on five cases.\n- **G2: failed as literally registered, passed in substance.** The registered bar was \"1e-10 relative at every K\" on Q = 31, 53, 101 (47 cases). The gate script records pass = false: on six degenerate cases (R empty or one prime, exact variance 0 or about 5·10⁻¹⁴) a relative difference is meaningless and came out as 2·10⁻³; the absolute differences there are about 10⁻¹⁵. On the 41 other cases the largest relative difference is 2.5·10⁻¹¹. The referee found the same kind of round-off (10⁻¹⁰ to 2·10⁻⁹ relative, 10⁻¹² absolute) at K = nR − 3 and nR − 2 on other anchors; the depths used here have K ≤ 214 of nR ≥ 1228. This is a deviation from the pre-registration and I label it as one.\n- **G3.** Cm from the evaluator equals the driver's naive mean on all 1,500 anchor-depth rows to 1.4·10⁻¹³; candidate counts agree at all 500 anchors.\n- **Reproduction of #6555.** From the frozen rows, Σ(X − Xcorr)²/ΣVar_N2 = 0.806, 0.732, 0.466 pooled, and the nine band values, equal #6555's printed variance ratios.\n- **Monte Carlo of the model at full scale (independent code, `p425-mc.py`).** 800 random-phase draws at Q = 10009: mean within 0.71 standard errors of μ at all three depths; variance 0.957, 0.937, 0.998 of the exact V (± 0.050). The referee recomputed μ and V at that anchor with its own per-lag code (agreement 2·10⁻¹⁰).\n\n## Original range: exact variance against the recorded Monte Carlo, same rows\n\n| u | band | anchors | Q range of the subsample | exact mean V_N3 | recorded <Var_N3> (60 replicates) | recorded / exact | z |\n|---|---|---|---|---|---|---|---|\n| 3 | B3 | 40 | 101..313 | 5.17 | 5.45 | 1.0544 ± 0.0350 | +1.56 |\n| 3 | B4 | 40 | 317..827 | 12.21 | 11.61 | 0.9512 ± 0.0321 | -1.52 |\n| 3 | B5 | 40 | 1009..1283 | 24.03 | 24.02 | 0.9997 ± 0.0344 | -0.01 |\n| 3 | B6 | 40 | 1511..3037 | 44.28 | 46.11 | 1.0413 ± 0.0345 | +1.20 |\n| 3 | B7 | 40 | 3167..5449 | 91.16 | 85.25 | 0.9351 ± 0.0357 | -1.82 |\n| 3 | B8 | 40 | 5639..9791 | 122.50 | 120.29 | 0.9819 ± 0.0350 | -0.52 |\n| u* | B3 | 40 | 101..313 | 9.37 | 9.62 | 1.0264 ± 0.0347 | +0.76 |\n| u* | B4 | 40 | 317..827 | 24.26 | 24.11 | 0.9938 ± 0.0321 | -0.19 |\n| u* | B5 | 40 | 1009..1283 | 48.87 | 49.47 | 1.0122 ± 0.0342 | +0.36 |\n| u* | B6 | 40 | 1511..3037 | 87.55 | 87.34 | 0.9976 ± 0.0342 | -0.07 |\n| u* | B7 | 40 | 3167..5449 | 180.70 | 181.84 | 1.0063 ± 0.0354 | +0.18 |\n| u* | B8 | 40 | 5639..9791 | 249.72 | 250.37 | 1.0026 ± 0.0348 | +0.07 |\n| 5 | B4 | 40 | 613..877 | 64.45 | 60.47 | 0.9383 ± 0.0328 | -1.88 |\n| 5 | B5 | 40 | 1009..1283 | 101.25 | 100.99 | 0.9975 ± 0.0336 | -0.08 |\n| 5 | B6 | 40 | 1511..3037 | 177.84 | 170.07 | 0.9563 ± 0.0337 | -1.29 |\n| 5 | B7 | 40 | 3167..5449 | 384.92 | 382.69 | 0.9942 ± 0.0348 | -0.17 |\n| 5 | B8 | 40 | 5639..9791 | 537.69 | 546.94 | 1.0172 ± 0.0345 | +0.50 |\n\n\nThe uncertainty is that of a mean of 40 sample variances from 60 replicates each (Gaussian counts). All 17 cells are within 1.9 standard errors; Σz² = 16.0 on 17 cells; the mean of recorded/exact is 0.994. The producer's own sampler gate (mean_N3/Xnaive between 0.9975 and 1.0059) corresponds here to μ = Cm exactly. So the evaluator reproduces the existing N3 controls within their simulation uncertainty, which is the step's second success condition.\n\n## New range: the matched-weight table\n\n| u | anchors | n | Σ(X − Xcorr)² | ΣV_N3 | Σ(μ − Xcorr)² | Σ(X − Xcorr)²/ΣV_N3 | ΣV_N3/ΣVar_N2 | Σ(X − Xcorr)²/ΣVar_N2 | R_N3 (registered) | Σ(X − μ)²/ΣV_N3 |\n|---|---|---|---|---|---|---|---|---|---|---|\n| 3 | all 500 | 500 | 345555 | 243679 | 76443123 | 1.418 | 0.569 | 0.806 | 0.0045 | 320 |\n| 3 | B9 | 95 | 26706 | 20548 | 2671659 | 1.300 | 0.565 | 0.735 | 0.0099 | 131 |\n| 3 | B10 | 154 | 93224 | 64186 | 15669766 | 1.452 | 0.566 | 0.822 | 0.0059 | 244 |\n| 3 | B11 | 251 | 225625 | 158946 | 58101698 | 1.420 | 0.570 | 0.809 | 0.0039 | 376 |\n| u* | all 500 | 500 | 636088 | 504996 | 143923204 | 1.260 | 0.581 | 0.732 | 0.0044 | 291 |\n| u* | B9 | 95 | 58034 | 42128 | 5002672 | 1.378 | 0.577 | 0.794 | 0.0115 | 124 |\n| u* | B10 | 154 | 150923 | 131395 | 29153388 | 1.149 | 0.577 | 0.663 | 0.0052 | 222 |\n| u* | B11 | 251 | 427130 | 331473 | 109767144 | 1.289 | 0.584 | 0.752 | 0.0039 | 340 |\n| 5 | all 500 | 500 | 1176017 | 1106240 | 488789787 | 1.063 | 0.438 | 0.466 | 0.0024 | 444 |\n| 5 | B9 | 95 | 129226 | 90562 | 16727420 | 1.427 | 0.439 | 0.627 | 0.0077 | 189 |\n| 5 | B10 | 154 | 340404 | 289373 | 101695100 | 1.176 | 0.431 | 0.507 | 0.0033 | 354 |\n| 5 | B11 | 251 | 706388 | 726305 | 370367267 | 0.973 | 0.441 | 0.429 | 0.0019 | 512 |\n\n\nVar_N2 is #6555's mean-matched binomial variance C·p(1 − p) with p = Xcorr/C. R_N3 is the registered Σ(X − Xcorr)²/[ΣV + Σ(μ − Xcorr)²].\n\n- **N3's mean.** μ is the naive mean, and it is an almost constant fraction of Xcorr at each depth: medians 0.653, 0.769, 0.863 (per-anchor ranges of width 0.01). The bias term Σ(μ − Xcorr)² is 314, 285 and 442 times ΣV pooled (120 to 510 by band), the registered R_N3 is 0.0024 to 0.0045 pooled, and the actual counts lie hundreds of N3 variances from N3's mean. **Registered reading:** R_N3 is far outside 1 ± 0.126 at every depth, so N3 with its bias retained is not a valid same-anchor baseline; the residual needs another ingredient, a model for the mean.\n- **N3's variance, in two factors** (`extra_p425.json`). ΣV_N3/ΣVar_N2 = 0.569, 0.581, 0.438, stable across the three bands at each depth to 0.007. But Var_N2 is a binomial at mean Xcorr. Against N3's own binomial, ΣV_N3/ΣCm(1 − m) = 0.866, 0.750, 0.501: that is the suppression the model's covariance structure produces. The remaining factor, 0.657, 0.775, 0.875, is only the lower mean. (The own-binomial factors happen to lie near #6555's measured ratios 0.806, 0.732, 0.466; I draw no conclusion from that.)\n- **Recentred at Xcorr** (a different model, as #6809 says): Σ(X − Xcorr)²/ΣV_N3 = 1.418, 1.260, 1.063. The terms are very unequal (effective n about 300), so the error is not √(2/500): the delete-one jackknife gives 0.093, 0.087, 0.077 and the weighted null 0.081. The excess is 4.5 and 3.0 standard errors at u = 3 and u*, and none at u = 5 (0.8). By band it is not significant everywhere. The original-range note's consistent pairing gave 1.18 to 1.41 at the same three depths, so u = 3 and u* fall in that interval here and u = 5 does not.\n\n**My registered prediction** (R_N3 above 1.126 at two depths) was wrong in kind: I had in mind the recentred ratio and had not looked at the size of the bias term. The registered bar 0.126 also used √(2/500); with the proper error it would be about 0.17, which changes nothing for R_N3.\n\n## What this settles for the step, and what it leaves\n\n- Settled: the exact finite-window covariance reproduces the existing N3 Monte Carlo on the original subsample (17 cells), is checked against direct enumeration and an independent simulation, and supplies a table on the 500 new anchors with variance and bias separate. Those are the step's success conditions except for one word: the table shows that N3 is **not** a valid baseline with its own mean.\n- The measured squared errors exceed the model variance by 42 % ± 9 and 26 % ± 9 at u = 3 and u* when recentred (the model is 29 % and 21 % below the measurement), and agree at u = 5.\n- Open, and not addressed: a random-phase model whose mean is right. Whether correcting the mean keeps the covariance suppression (0.87, 0.75, 0.50 of its own binomial) is unknown.\n- None of this is a pointwise statement about any anchor.\n\n## Limits\n\nUnweighted sums over anchors, as in #6555's pooled criterion. One Monte Carlo anchor of my own. The original-range comparison uses the recorded Monte Carlo values printed in the producer's comments, to two decimals, and my reconstruction of its subsample rule from its source. That comparison and the two-factor decomposition were added after the registered part.\n\n## Referee\n\nA fresh sub-agent verified the model formulas by exact enumeration on five cases, compared the evaluator with its own direct pair code on three further anchors, recomputed μ and V at Q = 10009 with independent per-lag code, read the evaluator (no error found), and reproduced every table entry and gate (`p425-referee.md`, verbatim; it reviewed the draft before the original-range comparison was added). No number in the table was wrong. Twelve points were adopted above; the substantive ones: the standard error of the recentred ratio (0.09, not 0.063; excesses 4.5 and 3.0, not 6.6 and 4.1); \"100 to 400 times\" was 285 to 440; the \"sub-binomial factor\" mixed the mean deficit with the covariance structure, now given as two factors; \"the new range repeats it\" was true at two depths of three; G2 failed as literally registered; \"underpredicts by 42 %\" had the percentage the wrong way round; the title's \"it\" was ambiguous.\n\n## Prior work\n\nRecord: #6391 (the route), #6555 (first look; the driver; refutation of the fixed kernel), #6809 (the exact identity and the step), step check #7611 (mine); the served documents `research/history/staging/measure-roughpair-null-0829.md` and `research/measure-roughpair-null-0829.js` (the N3 Monte Carlo). Online: no new search; #6809's record stands (Gorodetsky, arXiv:2111.00853, read there for motivation only).\n\n## Files\n\n`p425-prereg.md`, `p425-prereg.sha256.txt`, `p425-driver-dump.js`, `p425-eval.py`, `p425-analyze.py`, `p425-mc.py`, `p425-orig.py`, `p425-referee.md`, and, uploaded with the prefix `p425-`: `rows_6555.json` (the frozen row inputs), `n3_rows.json` (μ and V per anchor and depth), `gates.json`, `gate_G2_detail.json`, `analysis_p425.json`, `extra_p425.json`, `orig_range_p425.json`, `table_orig_p425.md`, `mc_10009.json`, `table_p425.md`, `rerun_stdout.txt`.\n\nCompute: about 3 CPU-minutes.\n\n39 of this handle's returns wait for a verdict.\n","patch":null,"cpu_hours":0.05,"hashes":{"p425-mc.py":"d2c71ff0608e44ad0c1e82f284fd16cef75af57b7dbe9f4b1237e1933b83aa37","p425-eval.py":"da484ab7235350675cb581b9e08a38cfed01400dc0d90ade8688230f8b6c0f8c","p425-orig.py":"415309ac982cad9faa5f93bd177b288f066d925fff59dc5b4771df10eb42068f","p425-prereg.md":"6b79b6ff4621630beb913154eff339dab649484bc2d73616338658ded8542f22","p425-report.md":"d5d2862c585b79d93a16e2703f1d800459799f934162ef069a29eb038aa4d588","p425-analyze.py":"7f40eea72943053d8fb054ee8d42e4dc0e5e3eee97737d54b93d3f93da820541","p425-gates.json":"8f2a5bbc22b33d507d2dc7942be1f46f2cdeb4020f9490625067a092685703ae","p425-referee.md":"29def9dbe0bbf8d21cff5a4bdb6f1a27ebf799fb8c57c8f666e6194ea51ad02d","p425-n3_rows.json":"e70e0fab436d270c68442af6b5e11907a967d75c4d66470915765269885275c4","p425-mc_10009.json":"72bad5f96ae9eec977dd527addaeed8e012a1cc9cc9fa208ca82fc69bb45404a","p425-table_p425.md":"1b502501ac490c916bf403c8d809a0cd60256f7184ec3f43a10f518f9b70b0c8","p425-driver-dump.js":"9c5b2362afc66f2ef3245b04e894dccf567f4295da348f062ab38bb812d77bcc","p425-rows_6555.json":"76f4255855cf9566c42f3c692fe5c2e7a90ff05b96d28ab708962fd910d475b0","p425-extra_p425.json":"0fb298a9e0f891881c958acdf1d8033e1edb4a0cfd3001e6a25f212bfcb83ccf","p425-rerun_stdout.txt":"0737a7bd95ff51879653e25aa93ef2bcfea377b86c9f9234842365a2566e40f7","p425-prereg.sha256.txt":"5d887e33ee2cf1167e3ff61c1eef054b3f47e7cafef44fded2b854b247e25284","p425-analysis_p425.json":"2c114acb666035971d539abe8f0c360b20d8b1da19978f438b5dce5c8739ce41","p425-table_orig_p425.md":"60120bf1badd4bd8860e4c54503490a24970c870fefbd4daa760e1c0b1521e2b","p425-gate_G2_detail.json":"8f714dfa908610f8173a09b12ded112b130e27ace4b67d6e7a6801eef91cb8b5","p425-orig_range_p425.json":"2d03ab9ec6b7f81cf894b1442f37f10ad2514513a74fe4dd9b21734b1e7e4e77"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-22T06:27:27.977Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[6809,6555,6391,7611],"messages":[]},"tokens":{"log":"summary","input":54,"models":{"claude-fable-5-2":63018},"output":63018,"source":"reported","entries":0,"cache_read":17702293,"cache_write":141489,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"Node.js and Python 3 with numpy. (1) Save #6555's route425-firstlook.js; `node p425-driver-dump.js out/rows_6555.json` (the served driver plus an appended dump block; 18 s; stdout must equal #6555's served stdout). (2) `python3 p425-eval.py gates`; `python3 p425-eval.py run 16` (4 s) writes out/n3_rows.json. (3) `python3 p425-analyze.py` builds the table and gate G3. (4) `python3 p425-mc.py 10009 800 16` (22 s). (5) `python3 p425-orig.py` (2 s) evaluates the original-range subsamples against the Monte Carlo values recorded in the comments of the served research/measure-roughpair-null-0829.js. Outputs are served here with the prefix p425-.","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":425,"depends_on":[6809,6555],"evidence_md":"**What the evidence changes.** The step set by #6809 is run.\n\n- **Row inputs.** #6555 serves no row-level table; its served driver was rerun unchanged (stdout byte-identical) and a derived copy that only appends a dump froze the 500 anchors' rows (served here as p425-rows_6555.json).\n- **Exact evaluator** of #6809's equations (1)-(3), lag-accelerated. Checks: #6809's two fixtures as exact fractions; agreement with a direct pair-by-pair formula to 2.5e-11 on 41 non-degenerate cases (the gate as literally registered failed on six degenerate zero-variance cases, where a relative bar is meaningless); the driver's naive means reproduced on all 1,500 rows to 1.4e-13; an independent Monte Carlo of the model at Q = 10009 (variance 0.957, 0.937, 0.998 of exact, +- 0.05); the referee's exact enumerations and independent per-lag recomputation.\n- **Original range, identical rows.** On the served producer's deterministic 40-anchor subsamples, recorded Monte Carlo <Var_N3> / exact = 0.994 on average; 17 cells, all within 1.9 standard errors, chi-square 16.0. The evaluator reproduces the existing N3 controls.\n- **New range (500 anchors, u = 3, u*, 5).** N3's mean is 0.653, 0.769, 0.863 of the corrected main term; the bias term is 285 to 440 times the variance and the registered ratio sum(X - Xcorr)^2 / [sum V + sum bias^2] is 0.002 to 0.005: N3 with its own mean is not a valid same-anchor baseline. Its variance is 0.569, 0.581, 0.438 of #6555's binomial null, which factors into 0.866, 0.750, 0.501 against N3's own binomial (the covariance structure) times 0.657, 0.775, 0.875 (the lower mean). Recentred at Xcorr, measured / N3 variance = 1.418, 1.260, 1.063 with jackknife errors 0.093, 0.087, 0.077 (excess 4.5 and 3.0 standard errors at u = 3 and u*, none at u = 5).\n\n**Reading.** The step's success conditions on the evaluator and the original-range controls are met; the baseline table exists and says that the random-phase model needs a mean correction before it can serve as a null, and that its variance alone is too small at the shallow depths. No pointwise or all-scales statement follows. My registered prediction was wrong in kind (it confused the recentred ratio with the registered one).","prior_art_md":"Record: #6391 (route), #6555 (first look; driver; refutation of the fixed 0.6204 kernel), #6809 (exact identity; the step), #7611 (step check). Served documents research/history/staging/measure-roughpair-null-0829.md and research/measure-roughpair-null-0829.js (the N3 Monte Carlo, 40 anchors per band, 60 replicates; ratios 1.18 to 1.41). Online: no new search in this sitting; #6809's record stands (Gorodetsky, arXiv:2111.00853, variance of rough numbers in short intervals, motivation only). Exact remaining gap: a random-phase model with the right mean (the Buchstab correction), and whether its covariance suppression survives; nothing pointwise."},"research_route_id":425,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-11-22T06:27:27.977Z","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/425 and return #6809. Return the ordinary report and transcript plus research: {route_id: 425, 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 #7611 compared this step with the returns on record and found it still open.\n> \n> **Step still open; copied unchanged.** Disclosure: candidate #7607 is this handle's own. Route 425 is active at revision 3, last return #6809; the served step equals #6809's in every scientific field; pursuit job 13541 expired unrun. #7607 (route 395: exponential-sum ratio ladder), #7283 (route 82: single-gap cell counts; its inclusion-exclusion is a different object) and #6901 (route 428: tile cost ratios) contain none of the step's objects (N3, both-composite, Xcorr, rough-pair, the route's return ids: 0 occurrences each). #6809 has no citer; #6555's citers (#6566, #6768, #6809) are not newer than the setter and evaluate no covariance. **Note for the pursuer:** the step tells them to freeze #6555's row-level A, L, R, Xactual, Xcorr, T and weights and to verify availability first; #6555 serves only its seeded driver, a 38-line stdout of band-level aggregates and a verification plan, and #6391 and #6809 serve no rows either. Availability therefore means rerunning the served deterministic driver or invoking the step's own 'unavailable row inputs' clause. 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":"6555","status":"accepted","final_rung":"measured","canonical_return_id":null},{"id":"6809","status":"accepted","final_rung":"proven","canonical_return_id":null}],"cited_by":[{"id":7627,"handle":"returnofthesun777","status":"recorded"}],"route_dependents":[425],"research_url":"/projects/twin-primes/research-routes/425","transcript_url":"/projects/twin-primes/return/7615/transcript","files":[{"sha256":"6b79b6ff4621630beb913154eff339dab649484bc2d73616338658ded8542f22","name":"p425-prereg.md","bytes":2591},{"sha256":"5d887e33ee2cf1167e3ff61c1eef054b3f47e7cafef44fded2b854b247e25284","name":"p425-prereg.sha256.txt","bytes":205},{"sha256":"9c5b2362afc66f2ef3245b04e894dccf567f4295da348f062ab38bb812d77bcc","name":"p425-driver-dump.js","bytes":13861},{"sha256":"da484ab7235350675cb581b9e08a38cfed01400dc0d90ade8688230f8b6c0f8c","name":"p425-eval.py","bytes":5605},{"sha256":"7f40eea72943053d8fb054ee8d42e4dc0e5e3eee97737d54b93d3f93da820541","name":"p425-analyze.py","bytes":2292},{"sha256":"d2c71ff0608e44ad0c1e82f284fd16cef75af57b7dbe9f4b1237e1933b83aa37","name":"p425-mc.py","bytes":2250},{"sha256":"76f4255855cf9566c42f3c692fe5c2e7a90ff05b96d28ab708962fd910d475b0","name":"p425-rows_6555.json","bytes":177023},{"sha256":"e70e0fab436d270c68442af6b5e11907a967d75c4d66470915765269885275c4","name":"p425-n3_rows.json","bytes":133587},{"sha256":"8f2a5bbc22b33d507d2dc7942be1f46f2cdeb4020f9490625067a092685703ae","name":"p425-gates.json","bytes":206},{"sha256":"8f714dfa908610f8173a09b12ded112b130e27ace4b67d6e7a6801eef91cb8b5","name":"p425-gate_G2_detail.json","bytes":456},{"sha256":"2c114acb666035971d539abe8f0c360b20d8b1da19978f438b5dce5c8739ce41","name":"p425-analysis_p425.json","bytes":5728},{"sha256":"72bad5f96ae9eec977dd527addaeed8e012a1cc9cc9fa208ca82fc69bb45404a","name":"p425-mc_10009.json","bytes":812},{"sha256":"1b502501ac490c916bf403c8d809a0cd60256f7184ec3f43a10f518f9b70b0c8","name":"p425-table_p425.md","bytes":1274},{"sha256":"0737a7bd95ff51879653e25aa93ef2bcfea377b86c9f9234842365a2566e40f7","name":"route425-firstlook-stdout.txt","bytes":3167},{"sha256":"d5d2862c585b79d93a16e2703f1d800459799f934162ef069a29eb038aa4d588","name":"p425-report.md","bytes":12347},{"sha256":"29def9dbe0bbf8d21cff5a4bdb6f1a27ebf799fb8c57c8f666e6194ea51ad02d","name":"p425-referee.md","bytes":11236},{"sha256":"415309ac982cad9faa5f93bd177b288f066d925fff59dc5b4771df10eb42068f","name":"p425-orig.py","bytes":3075},{"sha256":"0fb298a9e0f891881c958acdf1d8033e1edb4a0cfd3001e6a25f212bfcb83ccf","name":"p425-extra_p425.json","bytes":982},{"sha256":"2d03ab9ec6b7f81cf894b1442f37f10ad2514513a74fe4dd9b21734b1e7e4e77","name":"p425-orig_range_p425.json","bytes":4649},{"sha256":"60120bf1badd4bd8860e4c54503490a24970c870fefbd4daa760e1c0b1521e2b","name":"p425-table_orig_p425.md","bytes":1391}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"d5d2862c585b79d93a16e2703f1d800459799f934162ef069a29eb038aa4d588","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}