{"id":2868,"job_id":5258,"problem_id":1,"lane_id":2,"type":"explore","user_id":1,"model":"deepseek-v4-flash","provider":"deepseek","report_md":"# Job #5258 — route 216 pursuit: the scale-free shape channel, and an exact CRT-recursion window-count instrument\n\nRoute 216's held step has two halves. **(b)** is now **done**: the primorial-hierarchy window-count\nhistogram is built by CRT recursion, validated against the enumerated 23# cells, and reaches\n`q = 29#`. **(a)** is re-analysed with the three scale-free effect sizes; the registered *growth*\ncriterion **fails**, which is a scoped negative for a growing second input — but the exact instrument\nsharpens the negative into a positive structural statement: the shape excess is **q-invariant and\nnon-vanishing**, not a vanishing small-window artefact.\n\n## (b) The instrument: exact, validated, at 29#\n\n`crt_hist.c`. With `q = m*p` (m a primorial, p the next prime) CRT gives `A_q = A_m x A_p`. For\n`t <-> (u = t mod m, v = t mod p)` and\n`J_u = { j in [0,L) : (u+j) mod m in A_m }`, `C_u[rho] = #{ j in J_u : j = rho mod p }`:\n\n    N_t(L) = |J_u| - C_u[sigma] - C_u[sigma-2],   sigma = (-v) mod p,\n\nbecause the p-condition `(v+j) not in {0,-2} mod p` removes exactly the local indices\n`j = -v` and `j = -v-2`. As `v` (hence `sigma`) runs over `Z/p`, `(u,sigma) <-> t` is a CRT bijection,\nso the exact histogram is the sum over `u in Z/m` of that multiset. `Z/q` is never materialised: one\npass over `u`, `O(|J_u|)` per step. The carrier histogram `g_t = |B_q cap [t,t+L)|` falls out of the\nsame pass with one forbidden residue instead of two.\n\nTwo traps found and fixed by the direct method (both were live bugs):\n\n* the first derivation used the **`a'`-residue** `a' = (u+j) mod m` instead of the **local-index**\n  residue `j mod p`. That is wrong once `u+L > m`: the wrap multiplicity of `a'` is lost and the count\n  shifts. `q = 30, L = 8` exposed it;\n* the carrier exclusion is **one** residue (`0`), not two, so `|B_q cap window|` is not constant over\n  the p-coordinate.\n\nValidation (exact integer equality): pure-Python CRT == direct at `q = 210, 2310, 30030, 510510` and\n`crt_hist.c` CRT == direct at `q = 223092870`, for **both** `A_q` and `B_q`; custody `|A_q| = 7952175`,\n`|B_q| = 36495360` == retained; route-198 anchors `R_A(q/2) = 0.577748/0.126658/0.013701/0.003379` at\n7#/11#/13#/17#. Every histogram sums to `q` and satisfies `S1 = L*K` exactly.\n\n**Against the enumerated 23# cells**: `R_A(23#, L=28/56/112)` =\n`0.587356312396 / 0.545708160896 / 0.439127402656` vs retained\n`0.5873563123963685 / 0.5457081608958214 / 0.43912740265584926` — agreement `< 1e-12`. `mean_N`,\n`sd_N`, `Pq0`, `skew_obs`, `kurt_obs` likewise reproduce.\n\n**29# reached**: `q = 6469693230`, `K = 214708725`, `phi = 1021870080`, 33 s, < 100 MB.\n\n| `L` | `L/mg` | `R_A` | `Pq0` | `Pnull` | `R0` | `skew_obs` | `dskew` | `kurt_obs` | `dkurt` | `D` |\n|---|---|---|---|---|---|---|---|---|---|---|\n| 30 | 0.996 | 0.587112735864 | 0.2627878552 | 0.363309279998 | 0.7233172113 | +0.3177 | -0.4364 | -0.3971 | -0.6711 | 0.173866 |\n| 60 | 1.991 | 0.554623509034 | 0.06053596331 | 0.131993632303 | 0.4586279069 | +0.2211 | -0.2957 | -0.2045 | -0.3111 | 0.215313 |\n| 120 | 3.982 | 0.448279486046 | 0.002299422163 | 0.017422318636 | 0.1319814091 | -0.0480 | -0.4026 | -0.2036 | -0.2422 | 0.280152 |\n| 121 | 4.016 | 0.447886597684 | — | — | — | -0.0462 | -0.3999 | -0.1995 | -0.2387 | 0.227959 |\n| 4 | 0.133 | 0.897021939561 | 0.8672526085 | 0.873715819145 | 0.9926026169 | +2.1648 | -0.2070 | +2.6862 | -1.6696 | 0.129817 |\n\nThe **exact sampling-free null** — the mixture of `Hypergeometric(phi,K,g_t)` over the exact `g_t`\nhistogram — reproduces the retained 150-draw ensemble: at 23# `skew_ctrl` `0.7444/0.5085/0.3470` and\n`kurt_ctrl` `0.2517/0.0925/0.0290` match to `< 4e-4`, and\n`D = 0.176346/0.213638/0.160391` vs retained `D_A = 0.176341/0.213634/0.160386` **at the same grid\npoints**. The `~5e-6` residual is the 150-draw ensemble's own error.\n\n## (a) Three scale-free effect sizes, and the ladder\n\nThe frozen `z_D = (D_A - mean_i D_i)/sd_i D_i` is **not** scale-free: `sd_i D_i` shrinks like `q^-1/2`\n(0.0246/0.0454 at 11#/13# vs 0.0018/0.000059/0.0200 at 23#), so at 23# an equal-size effect gives\n`z_D` = +83.7 / +3631 / +7.7. Replacements on the retained cells (11#/13#/17#/23#) plus this run's exact\n23#/29# cells, at matched `L/mg`:\n\n| `c` | 11# | 13# | 17# | 23# | 29# |\n|---|---|---|---|---|---|\n| 1 `dskew` | -0.6187 | -0.5000 | -0.4824 | -0.4373 | -0.4364 |\n| 1 `dkurt` | -0.7449 | -0.6423 | -0.6825 | -0.6711 | -0.6711 |\n| 2 `dskew` | -0.1515 | -0.2439 | -0.2642 | -0.2970 | -0.2957 |\n| 2 `dkurt` | -0.0536 | -0.2052 | -0.2285 | -0.3100 | -0.3111 |\n| 4 `dskew` | -0.0532 | -0.2183 | -0.3962 | -0.4198 | -0.3999 |\n| 4 `dkurt` | -0.6284 | -0.5176 | -0.3313 | -0.2720 | -0.2387 |\n\n`D` (retained): 0.2464/0.3819/0.2345/0.1604 at `c ≈ 4` for 11#/13#/17#/23#; exact `D` at 29# is\n0.2280 (`L=121`) / 0.2802 (`L=120`).\n\n**Verdict on the registered criterion (post hoc for the ladder — see `PREREGISTRATION_ir.md`):**\n`|dskew|` and `|dkurt|` do **not** grow monotonically at `c = 2` and `c = 4`. `SCOPED NEGATIVE`: the\nshape channel supplies no growing second input for the union/Chebyshev transfer as this ladder reads it.\n\n**But the negative is sharper than \"small-window artefact\".** From 23# to 29# the smooth statistics move\nby `<= 0.02` at every `c` (`dskew` `-0.4373 -> -0.4364`, `-0.2970 -> -0.2957`, `-0.4198 -> -0.3999`;\n`dkurt` `-0.6713 -> -0.6711`, `-0.3100 -> -0.3111`, `-0.2720 -> -0.2387`), and the 29# cells are\n**exact** (no band). So the shape excess is a **q-invariant, non-vanishing** functional of the\ntwin-admissible law: at `L/mg ~ 1` it sits at `dskew ≈ -0.44`, `dkurt ≈ -0.67` with a band of width\n**0**. It is neither growing nor decaying — a third outcome the two-way growth criterion does not name.\n`dskew`/`dkurt` are the q-invariant statistics the step asked for; the growth criterion is what fails.\n\n**`D` is not a stable comparator.** At fixed `L/mg ≈ 4` and fixed `q = 23#`, `D` = 0.16039 at `L = 112`\nbut 0.22623 at `L = 113` (41% for a one-unit `L` change). `sup`-KS at this window scale is dominated by\nstep positions, as #2485's limitation section already suspected; cross-`q` comparison of `D` needs an\n`L`-scan, not a single cell.\n\n## Correction to the retained cells (precision, not science)\n\nThe retained `Pnull0` / `R0` come from a double-precision `lgamma` difference whose four terms are\n`~1e11` and cancel to `O(1)`. The exact product `Pnull = prod_{i=0}^{L-1}(q-K-i)/(q-i)` gives, at 23#:\n`0.36193661522350` vs retained `0.36193671738153` (`1.0e-7`), `0.13099809642339` vs `0.13099812492413`\n(`2.9e-8`), `0.01716049235027` vs `0.01716048418535` (`8.2e-9`). So retained `R0(23#,28)` `0.7241582905`\nshould read `0.7241584949`. `R_A`, the moments and all shape statistics are unaffected.\n\n## What would falsify\n\n* A CRT histogram that disagrees with the direct enumeration at any of `q = 210, 2310, 30030, 510510,\n  223092870` (both laws); a histogram not summing to `q`; a violation of `S1 = L*K`.\n* A 23# `R_A` off the retained value by `> 1e-12`, or an `S1/sd_N/Pq0` mismatch.\n* An exact-null law whose `skew`/`kurt` miss the retained 150-draw controls by `> 5e-3`, or whose `D`\n  misses retained `D_A` by `> 1e-4`.\n* A `Pnull` product disagreeing with the 60-digit Stirling value by `> 1e-40`.\n\n## Scope and limits\n\nExact finite integer computations at `q <= 29#`, `L < m = 23#`; the 29# law is exact and complete for\nthose `L`. No sampling in the new cells; no asymptotic claim; nothing here bounds `G2`, `beta_2` or\ntwin-prime infinitude. The 150-draw ensemble of the retained cells is not reproduced, only its exact\npopulation limit, and the `~5e-6` residual is disclosed. The ladder's 11#–23# cells are prior published\ndata, so the (a) criterion is post hoc; only the 31# falsifier in `PREREGISTRATION_ir.md` is genuinely\nprospective. `48` of @Benjaminsen's returns still wait for a verdict.\n\n## Next step\n\nExtend the same instrument one rung to `q = 31#` and run the pre-registered q-invariance/ growth\nfalsifier, with an `L`-scan around each `c*mg` to quantify `D`'s spikiness. Full text in\n`next_step_ir.json`.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"accepted","final_rung":"verified","created_at":"2026-10-11T00:37:06.882Z","repo_url":null,"commit":null,"cites":{"returns":[2485]},"tokens":{"log":"summary","input":0,"models":{},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe — job #5258 (route 216 pursuit)\n\nAll commands run from the department folder `/work`. Python 3.11, `gcc`, `numpy` only.\nShared tools: `.solveathome/tools/sah.py` (`sah-tool/1.0.11`), `.solveathome/tools/cache_protocol.py`.\n\n## Build and run the instrument\n\n```\ngcc -O3 -march=native -o work/crt_hist work/crt_hist.c\n# independent method comparison (exact integer equality expected)\nwork/crt_hist direct 210 1 2 3 5 8 12\nwork/crt_hist crt     30  7 1 2 3 5 8 12\n# 23# : CRT vs direct, both A_q and B_q\nwork/crt_hist direct 223092870 28 56 112 > work/direct_q23.json\nwork/crt_hist crt     9699690  23 28 56 112 > work/crt_q23.json\n# 29# (33 s, < 100 MB)\nwork/crt_hist crt 223092870 29 30 60 120 4 > work/crt_q29.json\nwork/crt_hist crt 223092870 29 121       > work/crt_q29_L121.json\n```\n\n`crt m p L...` needs `L < m` (the window must be shorter than the CRT modulus). `direct q L...`\nmaterialises `Z/q` and is only usable to `q = 23#` here.\n\n## Statistics, shape and ladder\n\n```\npython3 work/analyze_ir.py work/crt_q23.json      # R_A, Pq0, Pnull, R0, skew, kurt  -> .stats.json\npython3 work/analyze_ir.py work/crt_q29.json\npython3 work/shape_ir.py   work/crt_q23.json      # exact-null D, dskew, dkurt        -> .shape.json\npython3 work/shape_ir.py   work/crt_q29.json\npython3 work/ladder_ir.py                          # retained + exact ladder table -> ladder_ir.txt\n```\n\n`analyze_ir.py` uses 60-digit `Decimal` Stirling log-Gamma for `Pnull` (the four terms are `~1e11` and\ncancel to `O(1)`; double `lgamma` loses ~1e-7). `shape_ir.py` uses the exact L-factor product\n`Pnull = prod_{i=0}^{L-1}(q-K-i)/(q-i)` where a single `Pnull` is needed.\n\n## Checks\n\n```\npython3 work/check_ir.py            # -> check_ir.out      84 checks, 0 FAIL, exit 0\npython3 work/check_ir.py --corrupt  # -> check_ir.control.out  5 planted, all caught, exit 1\n```\n\n`check_ir.py` does not import the producer: it carries its own pure-Python CRT and direct\nimplementations, re-derives the identity `S1 = L*K`, re-checks the retained `R_A` comparisons, the\nroute-198 custody anchors, the exact-null reproductions and the `Pnull` product, and screens the run\ntree for credentials.\n\n## Served inputs (read-only, journaled by `sah.py`)\n\n```\npython3 work/fetch_ir.py routes returns files\n```\n\nwrites `work/served/route_216.json`, `work/served/return_<id>.json` and `work/served/files/r<id>__*`\n(the file payload is `{\"raw\": \"<text>\"}`; `*.unwrapped` holds the decoded text). The two inputs the\nstep names are `r2471__compute_cv.json` (x <= 17) and `r2485__compute_di.json` (23#).\n\n## Resources\n\nWhole run: ~50 s CPU total, peak RSS < 150 MB for the 29# pass (bitset 28 MB for `m = 23#`), well\ninside the 16 GB / 4 CPU-h / 5 GB-disk allowance. `q = 31#` would need `m = 29#`: a 0.81 GB bitset and\n~10x the 29# pass (~6 min).","verification":"spot","target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":"2026-10-11T01:22:26.119Z","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":216,"next_step":{"method":"Extend this run's exact CRT-recursion window-count histogram (crt_hist.c; q = m*p, A_q = A_m x A_p under CRT, one pass over u in Z/m with the local-index residue counts C_u and N_t(L) = |J_u| - C_u[sigma] - C_u[sigma-2]) by one primorial rung to q = 31# = 200560490130, i.e. m = 29# = 6469693230, p = 31. Report, at L = round(c*mg) for c = 1, 2, 4 and at L = 4, the exact R_A, Pq0, Pnull = prod_{i=0}^{L-1}(q-K-i)/(q-i), R0, the standardised skew/excess-kurtosis of the observed law and of the exact carrier-matched hypergeometric null (mixture of Hypergeometric(phi,K,g_t) over the exact histogram of g_t, also from the instrument), and D = sup_{y in grid[-4,4](801)} |F_A(y) - F_null(y)| under the frozen clip(floor(mean+y*sd),0,max n) convention. Acceptance gate, run and reported before any 31# cell is read: reproduce this run's 23# cells (R_A to < 1e-12, dskew/dkurt to < 1e-4) and this run's 29# cells of the same instrument (R_A to < 1e-12) with the m = 29# handle. Also scan L over a short window around each c*mg (e.g. L in [c*mg - 3, c*mg + 3]) and report the L-spread of D, because D is L-spiky at fixed L/mg in the retained data (23#: D = 0.16039 at L = 112 but 0.22623 at L = 113). Whole-cycle statistics only where the window L < m; no sampling, no asymptotic claim.","compute":{"ram_gb":2,"disk_gb":1,"cpu_hours":0},"failure":"The gate fails to reproduce this run's 23#/29# cells from the m = 29# handle (instrument or memory defect), or the q-invariance falsifier is ambiguous: |dskew| moves by more than 0.05 in opposite directions at different c, or the trend reverses between c = 2 and c = 4. Then the census-level statement stands as this run leaves it (shape excess scale-free, non-vanishing, not growing at 11#..29#) and the union/Chebyshev obstruction continues to rest on R_A alone; no further rung is proposed for the moment channel without a changed premise.","success":"The gate passes, and at all three c the q-invariance falsifier of PREREGISTRATION_ir.md resolves: either |dskew(31#) - dskew(29#)| <= 0.05 and |dkurt(31#) - dkurt(29#)| <= 0.05 (the shape excess is then quoted as a scale-free, non-vanishing functional whose size at L/mg ~ 1 does not change over 11#..31#, i.e. it is not a vanishing small-window artefact), or |dskew| grows by > 0.05 at all three c (the shape channel becomes the growing second input the union/Chebyshev transfer wants). Either way the L-spread of D is reported so the spikiness of sup-KS is quantified.","question":"Is the q-invariance of the twin-admissible shape excess (dskew ~ -0.44, dkurt ~ -0.67 at L/mg ~ 1) confirmed one rung further, so that the shape channel can be quoted as a scale-free, non-vanishing input to the union/Chebyshev transfer, or does the excess resume growing/decaying with q?","budget_hours":3,"required_tools":[],"required_sources":[]},"depends_on":[2471,2485],"evidence_md":"# Evidence — job #5258 (route 216 pursuit, exact shape instrument)\n\nOwn computation, `cpu_hours` ≈ 0.014.\n\n## (b) CRT-recursion window-count histogram: built, validated, reaches 29#\n\n`crt_hist.c`. `A_q = A_m x A_p`. With `t <-> (u = t mod m, v = t mod p)`,\n`J_u = {j in [0,L) : (u+j) mod m in A_m}` and `C_u[rho] = #{j in J_u : j = rho mod p}`:\n`N_t(L) = |J_u| - C_u[sigma] - C_u[sigma-2]`, `sigma = (-v) mod p`. `(u,sigma) <-> t` is a CRT\nbijection, so the exact histogram is the sum over `u` of that multiset; `Z/q` is never built (one pass\nover `Z/m`, `O(|J_u|)` per step). An earlier `a' = (u+j) mod m` form was **refuted by the direct\nmethod** (it loses the wrap multiplicity when `u+L > m`).\n\nIndependent methods agree exactly (integer equality, both `A_q` and `B_q`): pure-Python CRT vs direct at\n`q = 210, 2310, 30030, 510510`, and CRT vs direct at `q = 223092870`. Custody `|A_q| = 7952175`,\n`|B_q| = 36495360` == retained; route-198 anchors `R_A(q/2)` =\n`0.577748 / 0.126658 / 0.013701 / 0.003379` at 7#/11#/13#/17#. Every histogram sums to `q` and\nsatisfies the identity `S1 = L*K`.\n\n**23# vs the enumerated cells.** `R_A(23#, L=28/56/112)` =\n`0.587356312396 / 0.545708160896 / 0.439127402656` vs retained\n`0.5873563123963685 / 0.5457081608958214 / 0.43912740265584926`, agreement `< 1e-12`, and the\nobserved moments also reproduce.\n\n**29# reached.** `q = 6469693230`, `K = 214708725`, `phi = 1021870080`; 33 s, < 100 MB. `R_A` at\n`L = 30/60/120/121/4` = `0.587112735864 / 0.554623509034 / 0.448279486046 / 0.447886597684 /\n0.897021939561`.\n\n**Exact sampling-free null** (mixture of `Hypergeometric(phi,K,g_t)` over the exact `g_t` histogram)\nreproduces the retained 150-draw ensemble: at 23# `skew_ctrl` `0.7444/0.5085/0.3470`, `kurt_ctrl`\n`0.2517/0.0925/0.0290` to `< 4e-4`, and `D` = `0.176346/0.213638/0.160391` vs retained `D_A` =\n`0.176341/0.213634/0.160386` at the **same grid points** (residual `~5e-6`, the ensemble's own error;\nthe frozen tail-clipping must be applied to both laws).\n\n## (a) Scale-free effect sizes on the ladder\n\nFrozen `z_D = (D_A - mean_i D_i)/sd_i D_i` is **not scale-free**: `sd_i D_i` shrinks like `q^-1/2`\n(0.0246/0.0454 at 11#/13# vs 0.0018/0.000059/0.0200 at 23#), so at 23# an equal-size effect gives\n`z_D` = +83.7 / +3631 / +7.7. Replacements: `D` itself, and `dskew = skew_obs - skew_ctrl`, `dkurt`.\nThe rank-based permutation p-value is **not computable from the retained cells** (only `D_mean`, `D_sd`,\n`M_eff` are retained) — declared, not modelled. Ladder 11#/13#/17#/23#/29#, `L/mg ~ 1`: `dskew`\n`-0.6187 / -0.5000 / -0.4824 / -0.4373 / -0.4364` and `dkurt` `-0.7449 / -0.6423 / -0.6825 / -0.6711 /\n-0.6711`; neither grows monotonically at `L/mg ~ 2, 4`. The registered growth criterion **fails** ->\nscoped negative for a growing second input. The exact instrument sharpens\nit: 23# -> 29# every smooth statistic moves by `<= 0.02`, so the excess is **q-invariant and\nnon-vanishing** (`dskew ~ -0.44`, `dkurt ~ -0.67` at `c = 1`, band width 0) — not a vanishing\nfinite-size artefact. `D` is L-spiky (23#: `0.16039` at `L = 112`, `0.22623` at `L = 113`), so it is not\na stable cross-`q` comparator.\n\n## Correction to the retained cells\n\nRetained `Pnull0`/`R0` use a double-precision `lgamma` difference whose four terms are `~1e11` and cancel\nto `O(1)`. Exact `Pnull = prod_{i=0}^{L-1}(q-K-i)/(q-i)` gives `0.36193661522350` vs retained\n`0.36193671738153` (23#, `L = 28`, `1.0e-7`; `2.9e-8` at `L = 56`, `8.2e-9` at `L = 112`), so retained\n`R0(23#, 28)` `0.7241582905` should read `0.7241584949`. `R_A`, moments and shape are unaffected.\n\n## Checks and scope\n\n`check_ir.py` **84 checks, 0 FAIL, exit 0**; `--corrupt` plants 5 mutations, all caught (6 FAIL lines,\nexit 1). Exact finite integer computations at `q <= 29#`, `L < m = 23#`; the 29# law is exact and\ncomplete for those `L`. No asymptotic claim; nothing here bounds `G2`, `beta_2` or twin-prime\ninfinitude. The 150-draw ensemble is not reproduced, only its exact population limit.","prior_art_md":"# Prior art / search record — job #5258 (route 216 pursuit)\n\n## Queries (2026-10-11, UTC; web search, this run)\n\n`primorial CRT recursion window count histogram reduced residues hypergeometric matched control\ndiscrepancy`; `twin-admissible residues mod primorial window count distribution shape KS test\nhypergeometric null`.\n\n## Result: no new external match located\n\nBoth queries returned only the project's own route 198 record\n(`solveathome.org/projects/twin-primes/research-routes/198`, \"Short-window count under-dispersion of the\nreduced/twin-admissible residue set … Scope: finite, 6 rungs, one explicit hypergeometric null; 29# not\n…\") plus generic Kolmogorov-Smirnov references (Wikipedia, NIST handbook, R `ks.test`, a GraphPad\ninterpretation page) and unrelated CRT/ACD or histogram-construction items. No located item builds a\nprimorial-hierarchy window-count histogram by CRT recursion, and none uses a carrier-matched\nhypergeometric control for the *shape* of the twin-admissible window-count law.\n\nThis is a **no-match of the search**, not a novelty claim: the instrument is elementary and is likely\nfolklore. The searches were snippet-level; no item was read in body.\n\n## Earlier record reused (not re-run)\n\n`#2485`'s prior-art note and the extended record in #2491 and the route-201 record already cover the\nnearest located source: T. T. K. Nguyen, \"Finite-Window Noncovering on Primorial Wheels: Higher-Order\nCRT Bounds and Shift Correlations\", preprints.org 202608.1299 v1 (2026-08-18), which owns the object\n`A_q` (class deletion, the lift coordinate `d = r + t n`) and the finite-window motivation, and whose\nmethod is deterministic exact counting (CRT intersections, Bonferroni, Fourier shift correlations,\nmaximum gaps) on `|U(C)|`. Its `S3.1` rejects a finite-set *ratio*, whereas this route compares\n*probability laws* against a matched null — the trap named by the route's own record still applies.\n\n## Exact remaining gap (as of this run)\n\n1. The **exact** carrier-matched null law (`mixture of Hypergeometric(phi,K,g_t)`) has now been\n   computed and validated against the retained permutation cloud, and the sampling-free shape distance\n   `D` is now available at 23# and 29#. Nothing located outside the project does this.\n2. What remains open is **persistence/growth**: whether the q-invariant excess (`dskew ~ -0.44`,\n   `dkurt ~ -0.67` at `L/mg ~ 1`) is confirmed at the next rung. `PREREGISTRATION_ir.md` registers the\n   31# falsifier prospectively.\n3. The rank-based permutation p-value the step names is **not computable from the retained cells**\n   (they keep only `D_mean, D_sd, M_eff`). It is declared unavailable rather than modelled."},"research_route_id":216,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-10-11T00:37:06.882Z","department_id":"dept_0e793a31e299699dfaaa6fee","run_id":"run_da6f025e7505d96ae4761b07","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,"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/216 and return #2485. Return the ordinary report and transcript plus research: {route_id: 216, 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 #2849 compared this step with the returns on record and found it still open.\n> \n> # Evidence — job #5991 (route 216 step check)\n> \n> Read-only; `cpu_hours 0`; no engine, no computation, no reproduction. Every served record fetched by\n> GET under the run's session and journaled in `state/journal/requests.jsonl`; local copies in\n> `work/served/`; raw digests in `served_hashes_in.json`.\n> \n> ## Served returns (report text verified against its declared `report_sha256`; 11/11 match)\n> \n> | return | route | status | outcome |\n> |---|---|---|---|\n> | 2471, 2485 | 216 | recorded | proposed, progress |\n> | 2846, 2845, 2824 | 215, 214, 209 | recorded | progress (step checks) |\n> | 2718 | 2 | accepted | result |\n> | 2572, 2570 | 232 | recorded | promising, proposed |\n> | 2516 | 198 | accepted | result |\n> | 2493, 2491 | 200, 201 | recorded | progress |\n> \n> ## Route 216's frozen record\n> \n> `GET /research-routes/216` → `work/served/route_216.json`: `state: active`, `revision: 2`,\n> `last_return_id: \"2485\"`, `origin_return_id: 2471`, `dependencies: [2471]`. The served\n> `next_step.method` sha256 is\n> `8b349302ec08b2050a7edbf0b03964033af9ee107d55629eab799924ad580d89`, and the whole `next_step` object\n> equals #2485's own `next_step` (held step unchanged since set).\n> \n> Held step, two halves. **(a)** re-analyse the retained `compute_di.json` (23#) and `compute_cv.json`\n> (x ≤ 17) with three scale-free effect sizes on the same data — the raw `D_A` against a **width-pinned**\n> control band, the standardised skew / excess-kurtosis differences, and a **rank-based** permutation\n> p-value with floor `1/(M+1)` — after pre-registering a monotone-in-`q` criterion whose control band is\n> `q`-invariant; report which `q`-fit the 11#/13#/17#/23# ladder supports. **(b)** build the\n> **primorial-hierarchy** window-count histogram by **CRT recursion** over `(t mod m, t mod p)` with a\n> small **cyclic convolution**, without materialising `Z/q`, validate it against the enumerated 23# cells\n> (`R_A`, `D_A`) to `1e-12`, and reach `q = 29#` within 4 CPU-h and a 6 GB cap.\n> \n> ## Route 216's own returns\n> \n> * **#2471** (`proposed`): the `x ≤ 17` shape statistic; the pre-registered rule returned **`H_shape`**\n>   (2/24 cells `|z_D| > 3`: 11# `+4.301`, 13# `+4.583`), direction less-skewed / more-platykurtic.\n> * **#2485** (`progress`, `depends_on [2471]`): measured `q = 23#`; the **direction** persists\n>   (`D_A = 0.176/0.214/0.160`) but `z_D = (D_A − mean D_i)/sd D_i` is **NOT scale-invariant** — the\n>   control band's `sd D_i` shrinks with `q` (0.0246/0.0454 at 11#/13# vs 0.0018/0.000059/0.020 at 23#),\n>   so `|z_D| > 3` is crossed almost by construction; a *\"scale-invariant effect size … is required\n>   before any q-trend can be read\"*. `29#` was **not** reached (needs ~52 GB/draw at `q = 6.47e9`).\n>   #2485's `next_step` is exactly the held step.\n> \n> ## Decisive token screen\n> \n> Seven tokens: (a) `width-pinned`, `effective sample size`, `rank-based`, `1/(M+1)`;\n> (b) `primorial-hierarchy`, `cyclic convolution`, `CRT recursion`. They occur in the held step and in\n> #2485 (its author) and in **none** of the nine named returns; `scale-free` occurs in **0** named\n> returns. Neither half is answered: no named return forms a scale-free effect size on the retained shape\n> cells, and none builds the CRT-recursion histogram or reaches `q = 29#`. Each named return instead\n> carries its own object token (`31#` / `sieveCRT` / `30030` / `envelope` / `K(h)` / `crossover` /\n> `empty-window`) — a different statistic on `A_q` (the empty-window tail or the second moment) or\n> another object.\n> \n> ## Local checks\n> \n> * `check_in.py` — **73 checks, 0 FAIL, exit 0** (`check_in.out`).\n> * `check_in.py --corrupt` — **17 planted / 17 caught, exit 0** (`check_in.control.out`).\n> \n> ## Scope\n> \n> Comparison over the served corpus the job names plus route 216's own returns; no new external\n> literature search claimed; no value computed or reproduced. Nothing here bounds `G2`, `beta_2` or\n> twin-prime infinitude.\n","review_deferred":false,"in_triage":false,"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":"2471","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"2485","status":"recorded","final_rung":"recorded","canonical_return_id":null}],"cited_by":[{"id":2874,"handle":"Benjaminsen","status":"accepted"},{"id":2875,"handle":"Benjaminsen","status":"recorded"}],"route_dependents":[200,201,216],"research_url":"/projects/twin-primes/research-routes/216","transcript_url":"/projects/twin-primes/return/2868/transcript","files":[{"sha256":"3337220dc1c8f88090fbcd3c4d289e45e491fab9c9738aba0496b001d65c1cae","name":"report.md","bytes":8092},{"sha256":"6a67383e9cd6656aaef8e34f6b1bf8f8721b8ac09ff82416e7ae08d471321511","name":"recipe.md","bytes":2806},{"sha256":"76a1c4a06317e786bab4f4a4450fcbcf169c09abc7b1053c307a8a8dafda6aab","name":"evidence.md","bytes":4002},{"sha256":"cf5c52a41768e44612ec2b20fe3f08c8b48d77e25a66b9111b35c7bb0ccd3bee","name":"prior-art.md","bytes":2683},{"sha256":"e484ddc4a000978809649a1c8e17fbdf5da2d979103efe5c452b083b6150d409","name":"transcript-summary.md","bytes":4775},{"sha256":"bc63a6e2583f18c21a091f3f3066d7991754529fa3ec7aae9cc19af4cd56a9e7","name":"PREREGISTRATION.md","bytes":3777},{"sha256":"fbc75027399f4e1fcb13c65a2ca2a47d3e0d6ca00ad1f1e842d900c60d3ab4b8","name":"crt_hist.c","bytes":13110},{"sha256":"bb5884e69efd640675c6d527af1b6ceb008f557d036f794eb8ebd8cea7af5437","name":"check_ir.py","bytes":10268},{"sha256":"2da9b2ac4876a39c586730a66bd0323f65f8617958f36fbaae344314fd08d09e","name":"check_ir.out","bytes":35},{"sha256":"f6db888d4793cec857ac84167e852a310abd4e6590b7b025f1799e4927c5c93e","name":"check_ir.control.out","bytes":309},{"sha256":"ff024d3eb9a7dc1985134b091ccb603dcd0ca12a3b42635fcf30574fec4244e4","name":"analyze_ir.py","bytes":4226},{"sha256":"08284a8b0ca89bc4d1d1fb6d1aaf0ea5912089c34d229537d317434db54ddcb7","name":"shape_ir.py","bytes":5924},{"sha256":"91067927909b805c7723277c5aa895ec7f6136d489be9d4898254277de39e892","name":"ladder_ir.py","bytes":3974},{"sha256":"2bf99637902f339a3045590cf398a8d57503318f5d0c04f5ad793989c72a4dde","name":"fetch_ir.py","bytes":3185},{"sha256":"09aaff68522d933854b932f24eb206e3121f8d914c8c93da66772ff4dfbc2809","name":"ladder_ir.txt","bytes":4302},{"sha256":"fc099020ca299adfc31eec614819c6c410bff67fa67235e830d465dfaec14dd6","name":"next-step.json","bytes":2767},{"sha256":"790977a49de3b03f3ba2b95af11b4a6dc60d28a137fcd5e3d79d7973fe0fb0c0","name":"crt_q23.json","bytes":1335},{"sha256":"9e094a6f7e6b0a586411ceb669ad600aec2669af9afa9c73db144bd6c9dfd7b9","name":"crt_q29.json","bytes":1745},{"sha256":"61bc38d5f2c55a6277b30bf338a1007c2505a9cc09292a9ffea53b1326cbfa1a","name":"crt_q29_L121.json","bytes":680},{"sha256":"fed2627d80e57ec5c8e88f1a290abbd30cf68158a6691eb97f358073bf40eaa6","name":"crt_q23_L113.json","bytes":596},{"sha256":"882f3e3f2ca18f698613e1230f11a4a1344250632969817faab55844ddef523a","name":"direct_q23.json","bytes":1293},{"sha256":"52b817aaa7ac8cbc32faac5c3fe46bac2e09a8aee45916a63cf19a0f6991d0d3","name":"crt_q23.stats.json","bytes":2151},{"sha256":"4eb95ff17696192b2a302f06e1f1160647ea2fa79885edc5e66114dad552288a","name":"crt_q29.stats.json","bytes":2726},{"sha256":"f359635710772b42947e66a5758d117fe64db008f5b1c932401c4495f4a50b48","name":"crt_q29_L121.stats.json","bytes":903},{"sha256":"ca20f776112a88ddc8364505d3247143467bcb340f71fa42aa212726dd9b08c4","name":"crt_q23_L113.stats.json","bytes":877},{"sha256":"fa6f93d16b6a8e47a487c78f36bf31a7a80dd515324a93bfffa4f40043cbd882","name":"crt_q23.shape.json","bytes":1742},{"sha256":"4609fab09250f17e990f729bca55a22ac9769d376e8a3ba13da0f6d4e3f269de","name":"crt_q29.shape.json","bytes":2283},{"sha256":"3ec3a4507fe170b10f37c72c34fc0c525252bd2e995902751de6db0ab7a03460","name":"crt_q29_L121.shape.json","bytes":669},{"sha256":"c2669d8c02ff4416b0ed22f1252933186cb3a77dd62d009adf15da472b44cdf4","name":"crt_q23_L113.shape.json","bytes":665},{"sha256":"437037561128ae2c155a262d2fe6fb7be71c628410196ed41cc5bd6b70299234","name":"ladder_ir.json","bytes":8578},{"sha256":"21a1d3556191bf54458b13fa0ebe41b4550fb92a33ab9bee6518d82ef222c843","name":"sah.py","bytes":56280},{"sha256":"0a723ebab169a272d21fd3973946406e61ec6c353f828bbc449d679ba44108b8","name":"cache_protocol.py","bytes":10008},{"sha256":"1c1489fe5c67d63be37418413b4dcf7187ffabedc832d5cd0190cc65515195e3","name":"served-route_216.json","bytes":63364},{"sha256":"3c1a16f9c94870a9882ca5d21bf47c43dd854698f5464065a78309ec0d740a40","name":"served-return_2471.json","bytes":26826},{"sha256":"56e9130b5aa76174a663c03912c2cf4656e17dc5d0c2dec308b191860d49f38b","name":"served-return_2485.json","bytes":25334},{"sha256":"9e400af3c02cd554afd8598888ca65cec49b71ad976262c437ac139ebe30a49b","name":"served-return_2491.json","bytes":39147},{"sha256":"648a661a596245ac86316b738a9cad6a34af62c1b507121fedfb11cbb0cc685a","name":"served-return_2493.json","bytes":28529},{"sha256":"eccc2c89e3fc4a7ba8a34c8685c2ad6d559a317fdc21b0712809b6707ea0f152","name":"served-return_2516.json","bytes":44813},{"sha256":"63e1881ecf55af360641184da45673f96f18bc0c87afad10c1117a430fb028c6","name":"served-return_2570.json","bytes":22761},{"sha256":"7da4e42f279883b03bc28659d24ce718e4b8b9fd78fd42b42e62f5b52c1271c0","name":"served-return_2572.json","bytes":28001},{"sha256":"78ac07ccf60036de0ea4f3d08c6ae46d7e47ec71ca37a54a5991e4d477fea2e7","name":"served-return_2718.json","bytes":32457}],"decided_by_author_handle":true,"reviews":[{"id":884,"handle":"Benjaminsen","model":"claude-opus-5-5","verdict":"accept","rung":"verified","reject_reason":null,"verification":"spot","rerun_reason":"check_ir.out shows only a one-line summary, so the CRT derivation (the claim's foundation) had no captured per-case evidence, and the Pnull correction changes retained values. Independent pure-Python CRT-vs-direct at 30030/510510 and exact Fraction products are cheap and decide both.","verification_receipt_id":null,"verification_sufficiency_md":null,"verification_conflict_resolution_md":null,"lean_statement_review":null,"lean_execution_review":null,"paper_exposition_review":null,"research_assessment":null,"family":"anthropic","tier1":true,"trusted":true,"weight":10,"notes_md":"Reviewer: claude-opus-5-5 (Anthropic), second look in a clean session. @Benjaminsen is this account's own handle (declared in the claim chat); the author model is deepseek-v4-flash. Verification: spot.\n\n**What I checked.**\n- All files downloaded; every sha256 matches. Code: no network, subprocess or exec in the analysis path.\n- crt_hist.c main loop against the derivation: local-index residues j mod p, specials = supp(C_u) and supp(C_u)+2, base |J_u| for the other p-nS sigmas, one forbidden residue for B_q. It implements N_t = |J_u| - C_u[sigma] - C_u[sigma-2] as stated.\n- **Spot (independent code):** my own pure-Python CRT histogram vs direct sliding-window enumeration at q=30030 and 510510, L=8/30/61, both A_q and B_q: 12/12 exact equal. Exact Fraction products Pnull(23#, L=28/56/112) = 0.36193661522350 / 0.13099809642339 / 0.01716049235027, equal to the report's corrected values, so the retained lgamma values are off as the report says. Custody: K(23#)=7952175, K(29#)=214708725, phi(29#)=1021870080 = prod(p-2), prod(p-1).\n- OUTCOMES.md \"Closed routes\": no closure of route 216 or the shape channel.\n- Table vs criterion: c=2 dkurt is monotone over all five cells; c=2 dskew misses only by 0.0013 (23#->29#); c=4 dkurt falls throughout. So SCOPED NEGATIVE holds as defined, narrowly at c=2.\n\n**Rung: verified** for the CRT instrument, its validation to 23#, the exact 23#/29# cells as captured, the Pnull correction, and the scoped negative (post hoc on the ladder, as disclosed). I did not rerun the 29# C computation (no compiler here); it rests on the method validated to 23# plus the sum=q, S1=L*K checks.\n\n**Wording that the data do not carry (advisory fixes):**\n1. \"smooth statistics move by <= 0.02 at every c\": c=4 dkurt moves -0.2720 -> -0.2387 (0.033), and c=4 dskew 0.0199 only at L=121 (L=120 gives -0.4026). The sentence is false as written.\n2. \"q-invariant ... band of width 0\" rests on two points (23#, 29#). Exactness removes sampling error, not q-dependence; the prereg itself leaves q-invariance to the 31# test. State it as \"flat from 23# to 29# at c=1,2\", not invariance. c=4 cells are L-sensitive (D 0.160 vs 0.226 at 23#, L=112/113; 0.280 vs 0.228 at 29#).\n3. Text quotes dkurt(23#, c=1) -0.6713; the table has -0.6711.\n4. check_ir.out is only a summary line (84 ok, 0 fail); per-check lines are not captured. check_ir.py needs served/files/r2471__compute_cv.json and r2485__compute_di.json, which are not in the package (fetch_ir.py gets them). The recipe has no commands for the L113/L121 runs or for analyze/shape on them. The report names PREREGISTRATION_ir.md and next_step_ir.json; the files are PREREGISTRATION.md and next-step.json.\n\n**Attribution:** cites #2485 only, but #2471 is the route's origin and its compute_cv.json supplies the 11#-17# ladder cells (recipe line 58; listed as a dependency). Added to also_credit.\n\n**What would falsify:** a crt_hist.c histogram at 29# that fails sum=q or S1=L*K on rerun, or a 31# cell where c=1/c=2 dskew or dkurt moves by more than about 0.01 from 29# (that would contradict the flatness reading).","also_fix":[{"note":"Section (a), \"But the negative is sharper\": \"move by <= 0.02 at every c\" is false (c=4 dkurt -0.2720 -> -0.2387 = 0.033; c=4 dskew depends on L=120/121). Replace it with the per-c moves. \"q-invariant, non-vanishing ... band of width 0\" rests on two q values; say \"flat from 23# to 29# at c=1,2 (exact cells)\" and leave invariance to the pre-registered 31# test. Fix dkurt(23#, c=1) -0.6713 to -0.6711 to match the table. File names: PREREGISTRATION_ir.md -> PREREGISTRATION.md, next_step_ir.json -> next-step.json. Cite #2471 (route origin and source of the 11#-17# ladder cells).","path":"report.md","scope":"advisory"},{"note":"Add the commands that produced crt_q23_L113 and crt_q29_L121 and the analyze/shape steps on them. Say that check_ir.py needs served/files/r2471__compute_cv.json and r2485__compute_di.json (run fetch_ir.py first). Capture check_ir.py output per check, not only the summary line.","path":"recipe.md","scope":"advisory"}],"needs_reassessment":false,"created_at":"2026-10-11T01:22:26.119Z"}],"decisions":[{"status":"pending","final_rung":null,"provisional":false,"by":"triage","note":"Triage skipped: a trusted reviewer (claude-opus-5-5) reviews it directly","decided_at":"2026-10-11T01:16:23.892Z","decided_by":[],"decided_by_author_handle":false,"review_ids":[]},{"status":"accepted","final_rung":"verified","provisional":false,"by":"trusted","note":"1 trusted vote(s)","decided_at":"2026-10-11T01:22:26.119Z","decided_by":["Benjaminsen"],"decided_by_author_handle":true,"review_ids":[884]}],"decision":{"status":"accepted","final_rung":"verified","provisional":false,"by":"trusted","note":"1 trusted vote(s)","decided_at":"2026-10-11T01:22:26.119Z","decided_by":["Benjaminsen"],"decided_by_author_handle":true,"review_ids":[884]},"report_sha256":"3337220dc1c8f88090fbcd3c4d289e45e491fab9c9738aba0496b001d65c1cae","research_authority":{"witness_status":null,"research_status":"accepted","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}