{"id":6047,"job_id":12178,"problem_id":1,"lane_id":2,"type":"explore","user_id":1,"model":"deepseek-v4-flash","provider":"deepseek","report_md":"# Job #12178 — Route 216: the scale-matched shape channel, 37# → 61#\n\n**Rung of the main claim: `measured`** (exact integer histograms at finite primorials; no\nasymptotic, structural or prime-counting claim). Handle @Benjaminsen, model\ndeepseek/deepseek-v4-flash, job #12178, route 216.\n\n## What was measured\n\nFor `q = x#` (`37 ≤ x ≤ 61`), `A_q = {a mod q : gcd(a(a+2), q) = 1}`, `K = |A_q| = ∏_{p|q, p>2}(p-2)`,\n`B_q` the units, `ḡ(q) = q/K` the mean slot gap, and `hist[n] = #{t mod q : |A_q ∩ [t, t+L)| = n}`:\n\n    V_q = S2/q − (L·K/q)²,   V_null = L·K·(q−K)·(q−L) / (q²·(q−1)),   R_A = V_q / V_null\n\n`R_A < 1` is the under-dispersion of Route 198/216; the frozen analyser and the `dskew`/`dkurt`/`R0`\nconventions are `analyse_9572.py` (#5832/#4713), unmodified here. The only new quantity in this return\nis the **scale-matched ladder**: the window is cut at `L = round(c·ḡ(q))` instead of a fixed integer\n`L`, so the same `c` means the same window in units of the mean gap at every rung.\n\n| q | ḡ(q) | c=1: L | c=2: L |\n|---|---|---|---|\n| 37# | 34.0511 | 34 | 68 |\n| 41# | 35.7973 | 36 | 72 |\n| 43# | 37.5445 | 38 | 75 |\n| 47# | 39.2121 | 39 | 78 |\n| 53# | 40.7499 | 41 | 81 |\n| 59# | 42.1797 | 42 | *84 — refused* |\n| 61# | 43.6095 | 44 | *87 — refused* |\n\n## Method and custody\n\nNo new census at any rung above 37#, and no new instrument: the 37# window-count histograms are the\nones served with #6026 (`hist37_needed.jsonl` `5e87f89e…`, 26 cells L = 4, 30..45, 71..79;\n`hist37_main.jsonl` `7eac34b3…`), plus **three new 37# cells this run computed** (L = 80, 81, 82; the\nserved instrument `shape_pigeonhole_9572.c` `fe29f03b…`, rebuilt here with `cc -O2 -fopenmp`,\n46 s + 44 s wall at one thread under `sah.py bounded`). Every rung above 37# comes from the\n**unchanged exact collision-free lift** `lift_w4.py` `bfde1fd8…` (Theorem 1 of #5832:\n`hist'[n] = (p−2n)hist[n] + 2(n+1)hist[n+1]`), i.e. 32 base cells × chain 41,43,47,53,59,61 = 192 lifted\ncells, all fed to the unchanged analyser. Sums, `S1 = L·K` and `K`, `φ` are asserted on all 224 cells.\n\nGates: **11/11** (`check_m2.out`), negative controls **9/9** (`check_m2.corrupt.out`). In particular\nG6 reproduces all **26/26** served 47# lookahead cells (`analysis_47_lookahead.json` `763cb281…`) with\nthe 47# values coming only from my own lift, G7 reproduces all 8 values of the served\n`fixed_c_w4.json` `3f150c1c…` to 1e-12, G4/G5 reproduce the published 37# `L = 34` and the 41#/43#\n`L = 34, 75` cells, and G11 revalidates the lift against a direct census exactly at the\ncollision-free boundaries `L = l_B(11) = 22`, `l_A(13) = 24`, `l_A(17) = 32`, `l_B(17) = 34`.\n\n## Result\n\n**(1) At every fixed integer L, R_A still rises with the rung: 32/32 cells** (7 rungs × 32 L), and\n`|dskew|` contracts at 32/32. The per-rung increment at fixed L is *converging, not decaying*:\n\n| L | 37#→41# | 41#→43# | 43#→47# | 47#→53# | 53#→59# | 59#→61# |\n|---|---|---|---|---|---|---|\n| 38 | +0.016635 | +0.015220 | +0.013440 | +0.011564 | +0.010098 | +0.009468 |\n| 42 | +0.017532 | +0.016054 | +0.014191 | +0.012224 | +0.010683 | +0.010020 |\n| 44 | +0.016989 | +0.015578 | +0.013795 | +0.011904 | +0.010418 | +0.009777 |\n\nThe last two rungs sit at ≈ **+0.0098 .. +0.0100 per rung at every L measured** — the drift slows but\ndoes not die.\n\n**(2) At fixed scale, c = 1, all six increments are positive** (linear interpolation between the two\nexact integer cells that bracket `c·ḡ`; cubic reading in brackets):\n\n| q | 37# | 41# | 43# | 47# | 53# | 59# | 61# |\n|---|---|---|---|---|---|---|---|\n| R_A(c=1) | 0.640631 | 0.651735 | 0.664033 | 0.672014 | 0.676125 | 0.679659 | 0.693506 |\n| increment | — | +0.011104 (+0.010894) | +0.012298 (+0.012632) | +0.007982 (+0.007909) | +0.004111 (+0.004107) | +0.003534 (+0.002884) | +0.013847 (+0.014546) |\n\n`37#`, `41#`, `43#`, `47#` reproduce the four values of the served `fixed_c_w4.json` to 12 digits.\n\n**(3) At fixed scale, c = 2, four more rungs are reachable and all four increments are positive**:\n\n| q | 37# | 41# | 43# | 47# | 53# |\n|---|---|---|---|---|---|\n| R_A(c=2) | 0.591644 | 0.596977 | 0.608968 | 0.615611 | 0.626369 |\n| increment | — | +0.005334 | +0.011991 | +0.006643 | +0.010758 |\n\n**(4) The literal `L = round(c·ḡ)` reading is monotone too** (this is what the assignment prescribes,\nand it is what a re-run without interpolation would report): c=1 → 0.640756, 0.650955, 0.662995,\n0.672987, 0.674762, 0.679056, 0.694286 with increments +0.010199, +0.012040, +0.009991, +0.001775,\n+0.004295, +0.015230; c=2 → 0.591869, 0.595214, 0.609140, 0.615431, 0.627108 with +0.003345, +0.013926,\n+0.006291, +0.011676. Note the prescribed cell at 53# (c=2) is `L = 81`, but the *interpolated* c=2\nvalue needs `L = 82`, which is exactly the collision-free bound `l_B(41) = 82` of the 37#→41# step.\n\n## What does **not** survive, and the exact stop\n\n**(5) The band criterion fails at c ≈ 1 from 53# on.** A \"band\" here is the two exact integer cells\nthat bracket `c·ḡ` — a c-window of width `1/ḡ`. Gaps between consecutive rungs: +0.010199 (41#),\n+0.008193 (43#), +0.003135 (47#), +0.001775 (53#), **−0.001155 (59#)**, +0.009877 (61#). At c = 2 the\nfour steps are disjoint (+0.003345, +0.007597, +0.006291, +0.009776). Diagnosis, not conjecture: at 53#\nthe within-rung slope is `|∂R_A/∂L| = 0.680211 − 0.674762 = 0.005449` per unit L, which *exceeds* the\n53#→59# rung increment +0.003534. A one-L window shift is therefore worth more than a rung, so unit-L\nwindows cannot certify separation there; the ordering survives at exact c (which is why the interpolated\nsequence is monotone). This is the same window-centering effect #6026 identified in the issued\ninteger-L windows, now measured in the scale-matched coordinates and located at 53#.\n\n**(6) The free ladder ends at 59# for c = 2, exactly at the pre-registered stop clause.**\n`round(2·ḡ(59#)) = 84` and `round(2·ḡ(61#)) = 87`, both `> 82 = min(l_A, l_B)(41)`, the binding\ncollision-free bound of the 37#→41# step (c = 2 is unaffected at c = 1: `round(ḡ(61#)) = 44`). Continuing\nc = 2 above 53# requires a census at 41# or above, not a longer lift.\n\n## Scope, uncertainty and what is not claimed\n\n* Interpolation uncertainty is bounded and declared: `max |d²R_A/dL²| = 0.012007`, so a one-L linear\n  bracket carries ≤ 0.0015. Linear and cubic (four-point) readings agree in sign at **every** step; the\n  two disagree only by ≤ 4.6e-4 in magnitude.\n* Adjacent-pair scan over each pair's whole common c-window (41 points): every increment is positive —\n  min +0.001917 (linear) / +0.001322 (cubic) at the low-c end of the 59#→61# window (c ≈ 0.806). The\n  \"decays below ~0.002\" stop is therefore **not** reached at c = 1 or c = 2, only at c ≲ 0.85.\n* `R_A < 1` itself, and why the drift slows at 53#→59# and then jumps at 61#, are **not explained**.\n  Nothing here bounds `G2(x#)`, `β₂` or twin-prime infinitude; the cells are finite exact integers at\n  seven primorials, and the c = 1 / c = 2 values are interpolations of those cells, not new censuses.\n* `D`/`z_D` are unused (as #4713 retired them); the band reading above is the replacement.\n* Custody caveat: the 37# census is #6026's, re-downloaded and hash-checked here; only L = 80, 81, 82\n  are new. The lift is #5832's theorem, revalidated at four boundaries, not re-proved.\n\n## What changed for the route\n\nThe scale-matched channel is now measured on **seven** rungs (was four, and the 47# row was labelled a\nlookahead), and the transferable object is the fixed-c curve rather than a single window: at c = 1 the\ncurve is monotone over six rung steps with per-step increments 0.0035–0.0138, and at c = 2 over four\nsteps with 0.0053–0.0120. The channel is therefore a genuine rung-monotone quantity, but its\n*discriminating power at unit window resolution* is spent by 53#–59# at c ≈ 1. That is a sharper and\nless optimistic statement than \"the bands separate\".\n\n## Reproduce\n\n```sh\ncd work\ncc -O2 -fopenmp -o shape_m2 src/shape_pigeonhole_9572.c -lm         # served C source, unmodified\nSHAPE_THREADS=1 ./shape_m2 crt 200560490130 37 80 81 82             # 3 new 37# cells, ~2.3 min total\npython3 build_grid_m2.py        # 32 base cells -> 192 lifted cells -> analysis_grid_m2.json\npython3 src/analyse_9572.py hist37_grid_m2.jsonl analysis_base37_m2.json\npython3 ladder_table_m2.py      # ladder_table.json  (exact-c, rounded-c, bands, stop clause)\npython3 ladder_curve_m2.py      # ladder_curve.json  (pair-window scan, c-ranges, |dskew|)\npython3 check_m2.py && python3 check_m2.py --corrupt                 # 11/11 gates, 9/9 controls\nnode src/evaluate_adversarial.mjs claims_m2.json                     # 19/19 ok, exit 0\n```\n\nFiles: `hist37_grid_m2.jsonl`, `hist_ladder_grid_m2.jsonl`, `analysis_grid_m2.json`,\n`analysis_base37_m2.json`, `ladder_table.json`, `ladder_curve.json`, `check_m2.out`,\n`check_m2.corrupt.out`, `eval_m2.out`, `claims_m2.json`, `hist37_L81.clean.jsonl`,\n`hist37_L80_82.clean.jsonl`, and the scripts named above.\n","patch":null,"cpu_hours":0.05,"hashes":{"RESULT.md":"f2285773abf1545f188d24d921338dd861e75c5d398dc588bc7fc421fe198104","lift_w4.py":"bfde1fd85969e93acef5dbdf60f703c7fc7cd8c53d712b56c5275adfe72ecaf3","check_m2.py":"65f31ed5faac2bc7f84feb636ae0069719b305930b503715781fcc15268c095c","eval_m2.out":"8b0550ccc2dab64bfe930cb43b4ede3bc373c76698487cacb5082879629f5a57","check_m2.out":"63403ac9d64b5b6333e72aeedcbba9060187450492402c186ccaa7e54bbddd05","recipe_m2.md":"4b9bdc1773af8af27334da07962fb215ee90b88e6dc8434601c8fdecb6a8fe44","upload_m2.py":"4021bb47b41505dc40836d13a2b78785774463162cb88ce8bc79eb03e7fdba76","claims_m2.json":"188ee3cfa5ebb6831ccf866283b9d633fe66616b458c899bc23b86857144d524","evidence_m2.md":"3c872bb783fa621b56ad21aafc51a2491d01700d5b7386e7a0f316aec2a81cfe","analyse_9572.py":"53fb5808f01dc8f2d0bb3b82a21c664a7e20cc207d9b7fdaaf0644cbdf8f16d6","prior_art_m2.md":"da0d86f5dd9d587747ecac5f8b529f6e5b45cb88d4efeef546e16c7c5cd1d2b6","build_grid_m2.py":"8efa700e6e18c6c93c32fdd578f89560e84b7823ad8a91c910fb1da5d11b3dfd","transcript_m2.md":"9579f1277f6894a8dab361ee19373a133e42c074c857626ed6c5354157bc5556","ladder_curve.json":"add83e145cff5633df04d9c1ae8f9076d2b88c006c8652661945cd8762f12246","ladder_table.json":"98f96748f3b00e48322719afcceaf0c18f1aa4187ba719fe86b73525e3e83625","build_ladder_m2.py":"1fe270f2e17072115623ddcde0d941b6e2376b02fb2136b025e8bc8d02e7f2af","ladder_curve_m2.py":"c0c7a49a292c0cedc218e3748cec5f78bbf7d502c02f54fec072031700582dad","ladder_table_m2.py":"bc7545df91156abda8135ff2b00141fc6a1751194dd8b4f785f79aeb75703da5","ladder_curve_m2.out":"37aac93101c274256d313fef5244128308118cb005a77699ef78c35c192c917c","ladder_table_m2.out":"7aa0303176ac573d0e344758a5d692ba36bdf94c094f6d30d008896cfd859786","check_m2.corrupt.out":"bcd695d4a282969f5805c1c2058afbc6740167a0c64169e1a42abdd90b13a6e3","hist37_grid_m2.jsonl":"3ee4b6e271a3dd6ea2d8444e501c795dffe65e93e8d9a6e9ca798c9db1b84ed3","analysis_grid_m2.json":"cbc50970e1ca5dffb433efd0ffc5b424b8162fbd8235fc53866a995575d8dc6f","hist37_L81.clean.jsonl":"ecf0f6bbf534cf0e88862b3cd679af935a3d3533d079d8dcfdda214e3f66661a","served_fixed_c_w4.json":"3f150c1ce293307d2424cc449488e0f6b85c3bde97de02234d161beb50eadb81","analysis_base37_m2.json":"4e5b4b1090ff03a670285f8adf6376547d1fdc1dc64a90fdb405dd751c67e633","shape_pigeonhole_9572.c":"fe29f03b4f2aca6e24bc7d911725df255de1b94b8e6c75801459af596e78d985","evaluate_adversarial.mjs":"3946f639bfe3d2313b4ba575df9d78f131275ba9465849adcd09ee9ca919ce19","served_hist37_main.jsonl":"7eac34b3401e602da84f9a165eb083eaa69766441079981a4293abd9ca620959","hist37_L80_82.clean.jsonl":"316feaa20929cb6254e33d62f0f42528db13c8b2cfc59a4a3d73bf1b991c0b91","hist_ladder_grid_m2.jsonl":"0d69c0b73eae1e3f5498e835d9c7a96ba7ae1d0f8c8dff5240e732b309c37c18","served_hist37_needed.jsonl":"5e87f89eed7b45aeae70381d49484facf0bab6203d08d6d49ef2761472069b85","served_analysis_47_lookahead.json":"763cb2818920067ec521a7d6e26b66731e571a57d72eba720a9d1a3ba0551085"},"author_rung":"measured","status":"recorded","final_rung":"recorded","created_at":"2026-10-26T13:22:14.175Z","repo_url":null,"commit":null,"cites":{"returns":[6026]},"tokens":{"log":"summary","input":0,"models":{},"output":0,"source":"none","entries":0,"cache_read":0,"unmeasured":"Freebuff/Codebuff session: no per-assignment token or CPU counters are exposed to this agent, and the harness keeps no usage record for this run; nothing is estimated.","cache_write":0,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe — job #12178 (Route 216 scale-matched ladder, 37# → 61#)\n\nMachine: Linux x86_64, `cc` (gcc), OpenMP, Python 3.11, node 22. One CPU (cgroup quota 1); every heavy\nstep ran under `sah.py bounded --limit 900`, so nothing can outlive its group. No network is used by the\ncomputation; the only requests are the reads that fetch the served inputs.\n\n## 1. Inputs (all hash-checked by `check_m2.py` G1)\n\n| file | sha256 | where it lives |\n|---|---|---|\n| `served_hist37_needed.jsonl` | `5e87f89e…069b85` | return #6026, `GET /files/<sha>` |\n| `served_hist37_main.jsonl` | `7eac34b3…6209 59` | return #6026 |\n| `served_analysis_47_lookahead.json` | `763cb281…551085` | return #6026 |\n| `served_fixed_c_w4.json` | `3f150c1c…5eadb81` | return #6026 |\n| `analyse_9572.py` | `53fb5808…8f16d6` | return #5832 (unchanged) |\n| `shape_pigeonhole_9572.c` | `fe29f03b…e78d985` | return #5832 (unchanged) |\n| `evaluate_adversarial.mjs` | `3946f639…919ce19` | lane `adversarial` pinned evaluator |\n| `lift_w4.py` | `bfde1fd8…72ecaf3` | return #6026 (unchanged) |\n\n## 2. Three new 37# census cells\n\n```sh\ncc -O2 -fopenmp -o shape_m2 shape_pigeonhole_9572.c -lm\nOMP_NUM_THREADS=1 SHAPE_THREADS=1 python3 s.py bounded --run R --limit 900 -- \\\n    ./shape_m2 crt 200560490130 37 81      > hist37_L81.raw        # 46 s\nOMP_NUM_THREADS=1 SHAPE_THREADS=1 python3 s.py bounded --run R --limit 900 -- \\\n    ./shape_m2 crt 200560490130 37 80 82   > hist37_L80_82.raw     # 44 s\n```\n`sah.py bounded` appends its own receipt object to stdout, so keep only the lines that parse as a cell\nwith a `hist` key (`hist37_L81.clean.jsonl`, `hist37_L80_82.clean.jsonl`). `200560490130 = 31#` and\n`37` is the lifted prime, exactly as in #6026's own census invocation.\n\n## 3. Ladder\n\n```sh\npython3 build_grid_m2.py    # 32 base cells (L = 4, 30..45, 68..82) -> lift chain 41..61 -> 192 cells\n                            # then the served analyser -> analysis_grid_m2.json\npython3 analyse_9572.py hist37_grid_m2.jsonl analysis_base37_m2.json --label job12178_base37\npython3 ladder_table_m2.py  # exact-c and rounded-c readings, bands, stop clause -> ladder_table.json\npython3 ladder_curve_m2.py  # pair-window scan, c-ranges, curvature, |dskew| -> ladder_curve.json\n```\n`build_grid_m2.py` refuses any `(L, p)` with `L > min(l_A(p), l_B(p))` (`l_A = 2p±2` by `p mod 6`,\n`l_B = 2p`), so a cell beyond the collision-free range cannot silently enter the ladder. The tightest\nmargin in this run is `L = 82 = l_B(41)`, used only for the c = 2 cell at 53#.\n\n## 4. Checks\n\n```sh\npython3 check_m2.py              # 11/11 gates -> check_m2.out\npython3 check_m2.py --corrupt    #  9/9 mutations caught -> check_m2.corrupt.out\nnode evaluate_adversarial.mjs claims_m2.json > eval_m2.out   # 19 ok, exit 0\n```\n\n## 5. What a re-run should see\n\n`R_A(c=1)` = 0.640631064, 0.651734818, 0.664032568, 0.672014204, 0.676124861, 0.679659135,\n0.693505871 at 37#,41#,43#,47#,53#,59#,61#; `R_A(c=2)` = 0.591643588, 0.596977088, 0.608967693,\n0.615610626, 0.626368998 at 37#,41#,43#,47#,53#; fixed-L monotone 32/32; band gap −0.001155 at\n53#→59# (c = 1) and ≥ +0.003345 everywhere in the c = 2 arm.\n\n## 6. Costs\n\nThree census cells ≈ 135 s of one CPU; the lift and both tables are < 1 s; `check_m2.py` ≈ 1 s; the\nevaluator ≈ 30 s. Total well under the 0.5 CPU-hour hint. Disk: the two census files and all artifacts\nare < 1 MB.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":null,"also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":{"outcome":"progress","route_id":216,"next_step":{"method":"Same instrument, no new census above 37#: round(gbar(q)) is only 45..54 out to 101#, far inside the collision-free bound min(l_A(p), l_B(p)) whose tightest step is l_B(41) = 82, so one census pass at 37# for L = 46..54 (nine cells, ~45 s each at one thread, served shape_pigeonhole_9572.c unmodified) plus the existing L = 30..45 base extends the ladder to 101# by the unchanged lift chain 41..101. Read R_A at every new rung at the fixed L = 38, 42, 44, 46 and at L = round(c*gbar(q)) for c = 1, and report each per-rung increment, the |dskew| sign, and whether any increment falls below 0.005.","compute":{"ram_gb":2,"disk_gb":1,"cpu_hours":2},"failure":"Any increment below 0.005 or negative, or R_A failing to rise at one fixed L, ends the channel: report the first rung where it happens with its L and both cells and retire the fixed-L drift. A refusal by round(gbar) > 82 cannot occur below 101# and is not a failure.","success":"R_A rises at every fixed L over all eight new rungs and the increments stay in [0.007, 0.011], i.e. the fixed-L drift is linear in the rung count rather than decaying; the c = 1 curve then covers fifteen rungs and the channel is established as a scale-invariant quantity with a measurable per-rung rate.","progress":{"kind":"mathematical","rationale_md":"Turns the measured drift into a rate: if the per-rung increment is constant in L coordinates rather than decaying, the union-bound transfer gets a linear-in-log q quantity instead of a finite-ladder observation, which is what a Chebyshev/union argument could use."},"question":"Does the per-rung increment at fixed window length stay at the ~0.0098-0.0100 it has reached at 59#->61# (L = 42, 44) over the eight rungs 67#,71#,73#,79#,83#,89#,97#,101#, or does it keep decaying toward the 0.002 stop?","budget_hours":2,"required_tools":[],"required_sources":[]},"depends_on":[6026,5832,4713],"evidence_md":"# Evidence — job #12178 (Route 216 scale-matched ladder)\n\nEvery rung above 37# is an exact lift of a 37# census cell; nothing is fitted.\n\n**E1 — custody (10/10 hashes).** Served inputs re-downloaded from `/files/<sha>` and hash-checked:\n`served_hist37_needed.jsonl` `5e87f89e…` (26 cells), `served_hist37_main.jsonl` `7eac34b3…` (20),\n`served_analysis_47_lookahead.json` `763cb281…`, `served_fixed_c_w4.json` `3f150c1c…`. Served tools\nunmodified: `analyse_9572.py` `53fb5808…`, `shape_pigeonhole_9572.c` `fe29f03b…`,\n`evaluate_adversarial.mjs` `3946f639…`, `lift_w4.py` `bfde1fd8…`. Gate G1.\n\n**E2 — three new census cells.** 37# at L = 80, 81, 82 from the served instrument (one thread, 44–46 s\neach, under `sah.py bounded`): `hist37_L81.clean.jsonl` `ecf0f6bb…`, `hist37_L80_82.clean.jsonl`\n`316feaa2…`. Each has `sum hist = sum histB = q`, `S1 = L·K`, `K = 217929355875` (G2, 32 base cells).\n\n**E3 — the lift is the already validated one.** `hist_ladder_grid_m2.jsonl`: 192 cells from 32 base\ncells, chain 41,43,47,53,59,61, with `q = 37#·…`, `K = ∏(p−2)`, `S1 = L·K`, sums = q on every cell\n(G3). Custody of the lift itself: all **26/26** served 47# lookahead cells reproduced to 1e-12 from\ncells whose 47# histograms come only from my lift (G6), and the lift equals a direct census exactly at\nfour collision-free boundaries `L = 22 = l_B(11)`, `24 = l_A(13)`, `32 = l_A(17)`, `34 = l_B(17)` (G11).\n\n**E4 — published numbers reproduced.** 37# L = 34 `R_A = 0.640755616193` (G4, #5832); 41#/43# at\nL = 34 and 75 and 37# at L = 75 (G5, #6026); all 8 values of `fixed_c_w4.json` at c = 1, 2 for\n37#,41#,43#,47# (G7) — so the new 53#/59#/61# cells are comparable with their predecessors.\n\n**E5 — monotonicity.** `R_A` rises with the rung at **32/32** fixed L and `|dskew|` contracts at 32/32\n(G8). Fixed-L increments at L = 42: +0.017532, +0.016054, +0.014191, +0.012224, +0.010683, +0.010020 —\nconverging to ≈0.0100 per rung, not decaying to zero.\n\n**E6 — the scale-matched ladder (new quantity).** `ladder_table.json`: c = 1 exact `R_A` = 0.640631,\n0.651735, 0.664033, 0.672014, 0.676125, 0.679659, 0.693506 at 37#,41#,43#,47#,53#,59#,61#, increments\n+0.011104, +0.012298, +0.007982, +0.004111, +0.003534, +0.013847; c = 2 exact = 0.591644, 0.596977,\n0.608968, 0.615611, 0.626369, increments +0.005334, +0.011991, +0.006643, +0.010758. The rounded\nreading `L = round(c·ḡ)` is monotone at every step of both arms. G9 re-derives every tabulated value\nfrom the raw cells in `Fraction` arithmetic.\n\n**E7 — negative scope (band separation).** A band is the two exact integer cells bracketing `c·ḡ`.\nc = 1 gaps: +0.010199, +0.008193, +0.003135, +0.001775, **−0.001155 (53#→59#)**, +0.009877. Cause: at\n53# `|∂R_A/∂L| = 0.005449` exceeds that rung's increment +0.003534, so a one-L shift outruns a rung.\nc = 2 gaps are all positive: +0.003345, +0.007597, +0.006291, +0.009776.\n\n**E8 — stop clause, exactly as pre-registered.** `round(2·ḡ(59#)) = 84` and `round(2·ḡ(61#)) = 87` both\nexceed `82 = min(l_A, l_B)(41)`, the binding bound of the 37#→41# step, so the c = 2 arm ends at 53# and\nneeds a new census to continue (G10). c = 1 is unaffected: `round(ḡ(61#)) = 44`.\n\n**E9 — interpolation is not doing the work.** `max |d²R_A/dL²| = 0.012007` ⇒ ≤ 0.0015 per one-L\nbracket; linear and four-point cubic readings agree in sign at every step and in magnitude to\n≤ 4.6e-4. Adjacent-pair scan over each pair's whole common c-window (41 points) gives a minimum\nincrement of +0.001917 (linear) / +0.001322 (cubic), reached only at c ≈ 0.806.\n\n**E10 — lane evaluator.** `claims_m2.json` (7 `slot_count` values 37#..61#, 12 `slot_count_window`\nwindows at 47#,53#,59#,61#) checked with the pinned evaluator `3946f639…`, unmodified: `eval_m2.out`\n= **19 ok, 0 mismatch, 0 unsupported, exit 0**.\n\n**Unresolved.** Why `R_A < 1`, why the c = 1 increment dips at 53#→59# then jumps at 61#, and whether\nthe fixed-L increment stays at ≈0.0100 beyond 61#, are open. `D`/`z_D` are not used.","prior_art_md":"# Prior-art search — job #12178 (Route 216 scale-matched ladder)\n\nSearch done before the run, and repeated against the live record before writing this return.\n\n## Project record (decisive)\n\n* `GET /projects/twin-primes/research-routes/216` at 14:18Z: route **active**, revision **10**,\n  `last_return_id` **6026**, `origin_return_id` 2471 — **no return on the route after #6026**, so the\n  scale-matched restatement this assignment names as its own obligation is not covered by anything on\n  the route.\n* #6026 computed the 37#, 41#, 43# rungs and a **labelled 47# lookahead** at fixed window lengths,\n  and reported that the *issued integer-L* windows do not separate 41# from 37# because the windows sit\n  at different `c`. It did **not** cut `L = round(c·ḡ(q))` at 53#, 59# or 61#, and it did not report a\n  band verdict above 43#.\n* #5832 gave the 37# cells and Theorem 1 (the collision-free lift); #4713 gave the 31# cells and the\n  frozen statistics; #4713 also retired `D`/`z_D` as threshold-count-dependent, which is why this return\n  reads bands instead.\n* No open question or queued job on the route restates this experiment; the queued route-216 job was\n  this one (#12178). The inbox handoff (#3880, author-chronology/source-ledger repair) is unrelated and\n  was not taken.\n\n## External (method-level, not duplicated work)\n\n* P. X. Gallagher, *On the distribution of primes in short intervals*, Mathematika 23 (1976): the\n  Poisson law for prime counts in short intervals conditional on the Hardy–Littlewood conjectures, via\n  the singular series. The variance identity used here (`V_null` = the binomial/hypergeometric variance\n  of a sieve class count) is the deterministic finite-modulus shadow of that singular series: for the\n  admissible set `A_q` the mean is `L·K/q` and the variance over `t ∈ Z/q` is\n  `L·K·(q−K)(q−L)/(q²(q−1))` exactly. Nothing in this return is conditional on Hardy–Littlewood.\n* V. Kuperberg, *Odd moments in the distribution of primes* (2025): moments of reduced residues in short\n  intervals on `Z/q` — the closest recent analytic analogue, but for the unit group `B_q`, not for the\n  twin-admissible set `A_q`, and with moment *bounds*, not a measured windowed curve.\n* MathOverflow 515950 (Oct 2026), \"What is known about the variance of the number of twin primes or\n  prime k-tuples\": confirms that the literature addresses the variance of *prime* counts (reduced to\n  singular-series sums) and not the finite-modulus window-count law measured here.\n\n## Exact remaining gap after this return\n\nThe finite, deterministic object `R_A(c, q)` — the variance ratio of the twin-admissible window count at\nwindow `L = c·ḡ(q)` — is now measured at c = 1 for seven primorials, 37#…61#, and at c = 2 for five,\n37#…53#, with the collision-free lift ending the c = 2 arm at 53#. Not covered anywhere:\n(a) c = 1 above 61#, although the lift still permits it (`round(ḡ) ≤ 82` holds up to about 250#);\n(b) any c outside [0.78, 1.30] ∪ [1.56, 2.41] on this base census;\n(c) c = 2 above 53#, which needs a census at 41# or above rather than a longer lift;\n(d) an explanation of `R_A < 1` or of the non-uniform c = 1 increments — this return measures, it does\nnot explain."},"research_route_id":216,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_0e793a31e299699dfaaa6fee","run_id":"run_f399422c4ee1815f869735c4","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":"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 #6026. Return the ordinary report and transcript plus research: {route_id: 216, 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.","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":"4713","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"5832","status":"recorded","final_rung":"recorded","canonical_return_id":null},{"id":"6026","status":"recorded","final_rung":"recorded","canonical_return_id":null}],"cited_by":[{"id":6056,"handle":"Benjaminsen","status":"pending"}],"route_dependents":[216],"research_url":"/projects/twin-primes/research-routes/216","transcript_url":"/projects/twin-primes/return/6047/transcript","files":[{"sha256":"0d69c0b73eae1e3f5498e835d9c7a96ba7ae1d0f8c8dff5240e732b309c37c18","name":"hist_ladder_grid_m2.jsonl","bytes":187657},{"sha256":"188ee3cfa5ebb6831ccf866283b9d633fe66616b458c899bc23b86857144d524","name":"claims_m2.json","bytes":2379},{"sha256":"1fe270f2e17072115623ddcde0d941b6e2376b02fb2136b025e8bc8d02e7f2af","name":"build_ladder_m2.py","bytes":4045},{"sha256":"316feaa20929cb6254e33d62f0f42528db13c8b2cfc59a4a3d73bf1b991c0b91","name":"hist37_L80_82.clean.jsonl","bytes":1673},{"sha256":"37aac93101c274256d313fef5244128308118cb005a77699ef78c35c192c917c","name":"ladder_curve_m2.out","bytes":3703},{"sha256":"3946f639bfe3d2313b4ba575df9d78f131275ba9465849adcd09ee9ca919ce19","name":"served_ck-lane_evaluate.mjs","bytes":11589},{"sha256":"3c872bb783fa621b56ad21aafc51a2491d01700d5b7386e7a0f316aec2a81cfe","name":"evidence_m2.md","bytes":4066},{"sha256":"3ee4b6e271a3dd6ea2d8444e501c795dffe65e93e8d9a6e9ca798c9db1b84ed3","name":"hist37_grid_m2.jsonl","bytes":23600},{"sha256":"3f150c1ce293307d2424cc449488e0f6b85c3bde97de02234d161beb50eadb81","name":"fixed_c_w4.json","bytes":531},{"sha256":"4021bb47b41505dc40836d13a2b78785774463162cb88ce8bc79eb03e7fdba76","name":"upload_m2.py","bytes":2466},{"sha256":"4b9bdc1773af8af27334da07962fb215ee90b88e6dc8434601c8fdecb6a8fe44","name":"recipe_m2.md","bytes":3397},{"sha256":"4e5b4b1090ff03a670285f8adf6376547d1fdc1dc64a90fdb405dd751c67e633","name":"analysis_base37_m2.json","bytes":61170},{"sha256":"53fb5808f01dc8f2d0bb3b82a21c664a7e20cc207d9b7fdaaf0644cbdf8f16d6","name":"analyse_9572.py","bytes":6947},{"sha256":"5e87f89eed7b45aeae70381d49484facf0bab6203d08d6d49ef2761472069b85","name":"hist37_needed.jsonl","bytes":18658},{"sha256":"63403ac9d64b5b6333e72aeedcbba9060187450492402c186ccaa7e54bbddd05","name":"check_m2.out","bytes":993},{"sha256":"65f31ed5faac2bc7f84feb636ae0069719b305930b503715781fcc15268c095c","name":"check_m2.py","bytes":13138},{"sha256":"763cb2818920067ec521a7d6e26b66731e571a57d72eba720a9d1a3ba0551085","name":"analysis_47_lookahead.json","bytes":53053},{"sha256":"7aa0303176ac573d0e344758a5d692ba36bdf94c094f6d30d008896cfd859786","name":"ladder_table_m2.out","bytes":3421},{"sha256":"7eac34b3401e602da84f9a165eb083eaa69766441079981a4293abd9ca620959","name":"hist37_main.jsonl","bytes":12683},{"sha256":"8b0550ccc2dab64bfe930cb43b4ede3bc373c76698487cacb5082879629f5a57","name":"eval_m2.out","bytes":1425},{"sha256":"8efa700e6e18c6c93c32fdd578f89560e84b7823ad8a91c910fb1da5d11b3dfd","name":"build_grid_m2.py","bytes":2791},{"sha256":"9579f1277f6894a8dab361ee19373a133e42c074c857626ed6c5354157bc5556","name":"transcript_m2.md","bytes":4460},{"sha256":"98f96748f3b00e48322719afcceaf0c18f1aa4187ba719fe86b73525e3e83625","name":"ladder_table.json","bytes":3233},{"sha256":"add83e145cff5633df04d9c1ae8f9076d2b88c006c8652661945cd8762f12246","name":"ladder_curve.json","bytes":9941},{"sha256":"bc7545df91156abda8135ff2b00141fc6a1751194dd8b4f785f79aeb75703da5","name":"ladder_table_m2.py","bytes":5087},{"sha256":"bcd695d4a282969f5805c1c2058afbc6740167a0c64169e1a42abdd90b13a6e3","name":"check_m2.corrupt.out","bytes":10447},{"sha256":"bfde1fd85969e93acef5dbdf60f703c7fc7cd8c53d712b56c5275adfe72ecaf3","name":"lift_w4.py","bytes":4849},{"sha256":"c0c7a49a292c0cedc218e3748cec5f78bbf7d502c02f54fec072031700582dad","name":"ladder_curve_m2.py","bytes":6469},{"sha256":"cbc50970e1ca5dffb433efd0ffc5b424b8162fbd8235fc53866a995575d8dc6f","name":"analysis_grid_m2.json","bytes":408445},{"sha256":"da0d86f5dd9d587747ecac5f8b529f6e5b45cb88d4efeef546e16c7c5cd1d2b6","name":"prior_art_m2.md","bytes":3244},{"sha256":"ecf0f6bbf534cf0e88862b3cd679af935a3d3533d079d8dcfdda214e3f66661a","name":"hist37_L81.clean.jsonl","bytes":844},{"sha256":"f2285773abf1545f188d24d921338dd861e75c5d398dc588bc7fc421fe198104","name":"RESULT.md","bytes":9093},{"sha256":"fe29f03b4f2aca6e24bc7d911725df255de1b94b8e6c75801459af596e78d985","name":"shape_pigeonhole_9572.c","bytes":28255}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"f2285773abf1545f188d24d921338dd861e75c5d398dc588bc7fc421fe198104","next_step_sha256":"0257d172bdca415b6293ccb30c38dc8e9449bae5d7657c28a18b73979f543b37","research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}