{"id":6468,"job_id":12924,"problem_id":1,"lane_id":null,"type":"measure","user_id":103,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Route 93: the Mertens-factor model of the staircase ladder, predicted then tested at x = 31\n\n**Outcome: result.** The model predicted K*_model(31) = 169, and that prediction was committed before any exact x = 31 ladder number existed. The exact staircase ladder at x = 31 gives K*(31) = 168 as decided by this engine. The model closes one modulus late, the same one-modulus miss as at x = 29 (70 against 69). The miss is 1 (ratio 1.006), inside the route's factor-2 band [85, 338] and inside the 10 percent bracket [153, 185], so the route's falsifier does not fire at x = 31. K*_model(37) = 420 stands as an open prediction, since exact x = 37 was not run. A match at one more level is a data point, not a proof of the mechanism.\n\n## 1. Objects\n\nThese follow the definitions of the staircase note (return #1151, sections 1 and 7). W = x# and the comb is N_x = { r in [0, W) : r = 11 or 17 mod 30, r mod p not in {0, p-2} for 7 ≤ p ≤ x }, with N = |N_x| = 2 prod_{7≤p≤x}(p-2). The scour primes are the primes q with x < q ≤ isqrt(W), marched in ascending order. Fresh strikes, cap2 = cap_0 (cofactor counts with the comb's side conditions plus s(q)), cap_K (adding the first K scour primes below q as freshness moduli) and K* = least K with sum_q cap_K(q) < N are as in the note. The model of route 93 (return #1163) is\n\n  model_K = sum_q cap2(q) prod_{i < K, Q_i < q}(1 - 1/(Q_i - 1)),   K*_model = least K with model_K < N,\n\nwith Q_0 < Q_1 < ... the scour primes. Here model_K is computed in exact rationals.\n\n## 2. Method\n\nTwo independent counts:\n\n- **`cap2cof.c`, cofactor domain.** For each scour prime q and each side, the code sieves the cofactors m in [q, floor((W∓1)/q)] in the two classes mod 30 that put v = qm in the side's classes. It strikes m = 0 mod p for 7 ≤ p ≤ min(q-1, isqrt(t)), and at every wheel prime p ≤ x it also strikes the side class m = ∓2/q mod p. The survivors are exactly the m with P^-(m) ≥ q that satisfy the side conditions.\n- **`ladder.c`, slot domain.** For each comb slot r the code records a(r), the index of the least scour prime dividing r, and b(r), the same for r+2. The march gives r to min(a, b) as a fresh strike, and r survives iff neither exists. An A-side cofactor of q is a slot with a(r) = idx(q) and r ≠ q. Its depth, the first violated freshness modulus, is b(r) when b(r) < a(r); the B side is the mirror. The whole cap_K ladder at every K therefore comes from one pass: sum_q cap_K(q) = sum_q cap2(q) - sum_{d<K} G[d], where G[d] counts the contributions of depth d.\n\nBoth are C with OpenMP. The slot engine is segmented over the wheel progressions r = 11, 17 mod 30, in segments of 2^17 positions, and split into 4 parts that are summed afterwards. Everything ran as gpud jobs on chimaera's i9-14900KF at 24 threads.\n\n## 3. Calibration\n\n130 items were compared with the served outputs of #1163 and #1151 and none failed. The table shows the main ones; the commands are `sh src/run.sh calib` followed by `python3 -I calib.py out served1163 served1151`.\n\n| x | N | sum cap2, ours = #1163 | K*_model, ours = #1163 | K*, ours = #1151 | floor, ours = #1151 | removed / survivors, ours = #1151 |\n|---|---|---|---|---|---|---|\n| 11 | 90 | 56 | 0 | 0 | 34 | 45 / 45 |\n| 13 | 990 | 880 | 0 | 0 | 110 | 683 / 307 |\n| 17 | 14,850 | 16,135 | 2 | 2 | 82 | 11,751 / 3,099 |\n| 19 | 252,450 | 308,401 | 10 | 10 | 1,877 | 214,070 / 38,380 |\n| 23 | 5,301,450 | 7,034,588 | 27 | 27 | 4,841 | 4,703,975 / 597,475 |\n| 29 | 143,139,150 | 202,133,083 | 70 | 69 | 31,327 | 130,831,312 / 12,307,838 |\n\nThe comparison also covers:\n- #1163's per-prime cap2 tables at x = 19, 23, 29 (435, 1,739 and 7,863 primes), all equal;\n- #1163's model ladders, all rungs equal after rounding to nearest (#1163 printed rounded values);\n- the exact cap_K ladders at every rung printed in #1151's check1426.out (K ≤ 12), run08.out (K ≤ 12), run11.out (26 rungs to K = 31) and run18.out (32 rungs to K = 73), for example 143,107,823 at K = 69, x = 29;\n- the self-strike counts 2, 6, 16, 52, 175, 673;\n- the per-prime fresh / cap2 / cap_K rows of run08 (K = 2, 120 primes at x = 17), run11 (K = 27) and run18 (K = 69).\n\nThe two engines agree on every per-prime cap2 at x = 11..29.\n\n## 4. Prediction (committed first)\n\n| x | W | N | scour primes | sum cap2 | sum cap2/N | K*_model | model at K*_model - 1 | model at K*_model |\n|---|---|---|---|---|---|---|---|---|\n| 31 | 200,560,490,130 | 4,151,035,350 | 37,534 (37..447,829) | 6,127,829,466 | 1.476217 | 169 | 4,151,801,969 | 4,150,743,941 |\n| 37 | 7,420,738,134,810 | 145,286,237,250 | 198,274 (41..2,724,079) | 222,251,743,832 | 1.529751 | 420 | 145,292,123,244 | 145,281,530,435 |\n\nThe model values are floors of the exact rationals. These numbers and the exact x = 31 command were committed in the department's git history as commit `940cb10` before the exact x = 31 run. The exact outputs first appear in the later commit `2ecbcf7`.\n\n## 5. Test: the exact ladder at x = 31\n\n| quantity | value |\n|---|---|\n| comb slots counted | 4,151,035,350 = N = 2 prod_{7≤p≤31}(p-2) |\n| fresh removals; survivors | 3,867,586,163; 283,449,187 (removed + survivors = N) |\n| fresh self-strikes; sum s(q) | 2,804; 18,770 |\n| sum cap2 | 6,127,829,466, equal to the cofactor count, prime by prime at all 37,534 primes |\n| cap_K, K = 1, 2, 10, 50, 100, 150 | 5,963,848,039; 5,825,152,538; 5,176,754,331; 4,457,379,796; 4,258,643,567; 4,171,677,087 |\n| cap_K, K = 166, 167 | 4,153,030,127; 4,151,955,707 (≥ N) |\n| cap_K, K = 168, 169, 170 | 4,150,884,942 (< N); 4,149,823,971; 4,148,783,480 |\n| **K*(31)** | **168**, certified floor N - cap_168 = 150,408 twin pairs in Houses 11/17 |\n| gains near K* | Δ(166) = 1,074,420, Δ(167) = 1,070,765, Δ(168) = 1,060,971 |\n| cap_infinity anchor | removed - sum self + sum s = 3,867,602,129 (0.9317 N) |\n\nThe following identities hold:\n- removed + survivors = N;\n- slots = 2 prod(p-2);\n- self(q) ≤ s(q) at every q;\n- sum cap2 - sum_d G[d] = removed - sum self + sum s (the note's cap_infinity anchor);\n- the ladder is non-increasing;\n- fresh(q) ≤ cap_168(q) ≤ cap2(q) and fresh(q) ≤ cap_169(q) ≤ cap2(q) at every q;\n- the per-prime cap_168 and cap_169 sum to the ladder values.\n\n**Verdict in the route's terms.** K*_model(31) = 169 against K*(31) = 168. That is inside a factor 2 and inside 10 percent; the miss is one modulus (+1, 0.6 percent).\n\n## 6. Model against exact\n\n| x | K*_model | K* | model - exact at K* | step at K* | ratio |\n|---|---|---|---|---|---|\n| 23 | 27 | 27 | 122 (0.0023 %) | 15,844 (0.299 % of N) | 0.01 |\n| 29 | 70 | 69 | 32,492 (0.0227 %) | 123,279 (0.086 % of N) | 0.26 |\n| 31 | 169 | 168 | 917,027 (0.0221 %) | 1,070,765 (0.026 % of N) | 0.86 |\n\nAt x = 31 the model is within 0.023 % of the exact ladder at every K ≤ 172. It is below the exact ladder at K = 1..19 and 22..76, and above it at K = 20, 21 and from K = 77 on. At K = 168 the model's excess by band is:\n- +8,183 (0.0003 %) for the 110 primes q ≤ W^{1/4};\n- +394,293 (0.067 %) for the 648 primes in (W^{1/4}, W^{1/3}];\n- +514,551 (0.073 %) for the 36,776 primes above W^{1/3}.\n\nSo the excess sits where the cofactor interval W/q is short against the product of the freshness moduli. The relative excess at K* is nearly the same at 29 and 31, while the relative step falls by a factor 3.3. If both trends held at x = 37 (model step at K*_model 0.0073 % of N), the excess would span about three steps. That is an extrapolation from two levels and does not revise the pre-registered 420.\n\n## 7. Fit of sum cap2/N (a fit)\n\n| x | 11 | 13 | 17 | 19 | 23 | 29 | 31 | 37 |\n|---|---|---|---|---|---|---|---|---|\n| sum cap2/N | 0.6222 | 0.8889 | 1.0865 | 1.2216 | 1.3269 | 1.4121 | 1.4762 | 1.5298 |\n\nLeast squares on 1/ln x over all eight levels gives 3.3120 - 6.3095/ln x, rms residual 0.0357, fitted limit 3.31. The last four levels (23..37) give 2.8966 - 4.9333/ln x, rms 0.0128. The seven levels without 37 predict 1.5815 at 37 against the measured 1.5298. The levels span 1/ln x in [0.277, 0.417], the residuals change sign along x, and the intercept is an extrapolation. Both intercepts exceed 1, so the route's \"fitted limit below 1\" falsifier does not fire on this family.\n\n## 8. Compute\n\nJobs #1373 (calibration, cap2 at 31 and 37) and #1374 (exact ladder at 31, 4 parts) took 113 s of wall time and 2,538 CPU-s together (cgroup counters). An earlier calibration pass, job #1372, added 9.8 CPU-s. The total is 0.71 CPU-h. Peak RSS was 113 MB (cap2 at 37) and 58 MB (exact ladder at 31). The exact x = 29 ladder takes 0.29 s wall and 6.9 CPU-s on this engine. The route had priced the step at 12 CPU-h, a price set by the JS engines.\n\n## 9. What this does not settle\n\n- The factorisation cap_K(q) = cap2(q) prod(1 - 1/(q'-1))(1 + O(E)) with a fundamental-lemma error term is not proved; section 6 locates the error to be controlled above W^{1/4}.\n- Exact K*(37) is not computed; K*_model(37) = 420 is open.\n- K* → ∞ is not proved; sum cap2/N rises through x = 37, and the fitted limits are extrapolations.\n- The one-modulus miss at x = 29 and x = 31 is two data points about the model's crossing, not a law.\n\n## Sources\n\n| Return | File | sha256 | Use |\n|---|---|---|---|\n| #1151 | `staircase-note.md` | f595ed1250e6d97c0a2c55ff1b5948ce1d435e6acb47053e0500bb4732adb1f4 | definitions (sections 1, 4, 7), Theorem 8 table |\n| #1151 | `check1426.out` | 24da302eb56e5fc18d73300dbef91e3ce279df717c4eb7d93dfb63010f55e875 | ladders x = 11..19, K ≤ 12 |\n| #1151 | `run08.out` | cac7510dac47afa4985eea72fdc16cde725f9b74a2bb038ab255460f73305455 | certified ladders x = 11..19, per-prime rows |\n| #1151 | `run11.out` | 6a1a59b30ebc686ece1a4027a7590595ffaa6f606a043b4c8f6d1e633bab6de3 | certified ladder x = 23 |\n| #1151 | `run18.out` | 8a518a846724be94fc5382b0e3b7ce351bb01e3332978790c9451401b080f68a | certified ladder x = 29, marches x = 11..29 |\n| #1151 | `check1426.py` | 6128a53536e7f74796e1435d0a7744d9e144d4243eb28afda37a9546de60f8f5 | read; not run |\n| #1151 | `report.md` | 2302a816ffb010c9b4f9c414992c54cf02cec2d03eb2bf0202e2263a9b87fd9f | the certified levels and their producers |\n| #1163 | `cap2-19.json`, `cap2-23.json`, `cap2-29.json` | b91c35bcb592e939aab17baf7f703ce0e63fa77b134d265073cc28ad2c081e65, b8554886230d449700e289998bdd40c74f6479a1ebe3d9b91218415ceb69dddd, 3b52d680d2e935b228d7c12c94ad5145d85f1fc0ff719f2dd7da09f239680ccc | per-prime cap2 and model ladders |\n| #1163 | `cap2-23.out`, `cap2-29.out`, `kmodel2468.out` | 8a7f00b3db1451eb38eb751cfe83f5cb36eee32ed695e31d141961c5a216ac49, 9734795fecec10b8b12f2eb46d97ffe6e513c62a7cb6daf91174f3577ada4d27, 37e3708a1ba5f7b07773180ed83b39c183908902c82ec88d0d41a50e835608b7 | K*_model and ladders x = 11..29 |\n| #1163 | `report.md` | acbccc0c82868bb33b49c1dda7b1c47150022da017da4608e623993905f5b1f4 | the model and its definition |\n| #1162 | `report.md` (no files served) | 393584a562da99c8f2e6d769128ba8c252e05e3ba59991017b27807b99da3956 | the route and its falsifiers |\n| #1148 | `report.md` | 4dd78c348252640b59ab9ae65b76eaa9c41cbb4ae6efe8ba7faaf9d7d66d5c8f | cited by #1163; read, not used here |\n\n#1163's scripts `kmodel2468.py` and `cap2vec2468.py` were not served; this return reproduces their outputs with its own code. Files: see `recipe.md` and `files/`.\n\n## Attached files\n\n`PREDICTION.md` is the pre-registration exactly as committed in `940cb10`, before the exact x = 31 run (sha256 8144dd30b3dcaafc070678cf1a5b93f6935bdd29851d75374bc7dc76281e8e88). One sentence in it is wrong: #1162's heuristic c2 of about 2.2 lies below both fitted intercepts (2.90 on the last four levels, 3.31 on all eight), not between them, so both fits overshoot that constant. Each `.out`, `.log` and `.tsv` file is attached with `.txt` appended, since the server takes text extensions only; the bytes and the sha256 values in `recipe.md` are unchanged.\n","patch":null,"cpu_hours":0.8,"hashes":{},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-10-28T21:55:38.085Z","repo_url":null,"commit":null,"cites":{"returns":[1148,1151,1162,1163]},"tokens":{"log":"claude-code","input":4,"models":{"claude-opus-5-5":2733},"output":2733,"source":"claude-jsonl","entries":2,"cache_read":355282,"cache_write":3505,"observed_models":["claude-opus-5-5"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe: route 93, K*_model(31), K*_model(37) and the exact K*(31)\n\n## Scripts (this return)\n\n| File | sha256 | Role |\n|---|---|---|\n| `cap2cof.c` | 2fef18b66956da8cfb077a27ec976cce11336bd98511d863ca4b11fb5b099a74 | per-prime #A_0, #B_0 in the cofactor domain (cap2 = A_0 + B_0 + s(q)) |\n| `ladder.c` | b881efa56f9abe4f0a6f83163c8797621b80b0387963423bf7a2938473095371 | exact march and cap_K ladder in the slot domain |\n| `merge.py` | 5b5194e97869534689b79838edd4543b1179b7acf85facac3c63647d6373904e | sums parts; cap2, exact-rational model ladder, K*_model; ladder, K*, floor, identities |\n| `calib.py` | 9f8320bb2211706e8fb6e48bcba48e83e8fc742526831c37552c2dc8fd0bfb31 | compares outputs with the served files of #1163 and #1151 |\n| `fit.py` | ed5701f7347557cd6b77b6fb5ac3e16adaa04151094c46c4e4f32f7abbd03f93 | least squares of sum cap2/N on 1/ln x |\n| `run.sh` | 230ba9ab60c84f2c10e2456cf709f0aa8c99dfed19308bddffb8fb3ac8b0b8b6 | gpud job driver: nix gcc build, GNU time, cgroup counters, parts skipped when complete |\n\nToolchain: gcc 16.2.0, GNU time and Python 3.14.8 from `github:NixOS/nixpkgs/5adc50831c259f87faeb09ddf930d43597197473` (`run.sh` enters `nix shell` itself). Build: `gcc -O3 -march=native -fopenmp -Wall`. Host: x86-64 Linux, Intel i9-14900KF, 24 threads. `merge.py`, `calib.py` and `fit.py` are Python stdlib only. `run.sh` records the sha256 of every source and binary in its log. Job #1373 ran `run.sh` with sha256 7b8a6792ceb8ff02c7bc7aac5ab122b793e6ba08f67c6daa843c573a9b5ab8d9; the served `run.sh` differs from it only in reading the cap2 table for the `ladder` step from `$R93_CAP2` (job #1374 ran the served version). The calib and cap2 steps are byte-identical in both.\n\n## Served inputs (read by `calib.py`; no served script was run)\n\n| Return | File | sha256 |\n|---|---|---|\n| #1151 | `check1426.out` | 24da302eb56e5fc18d73300dbef91e3ce279df717c4eb7d93dfb63010f55e875 |\n| #1151 | `run08.out` | cac7510dac47afa4985eea72fdc16cde725f9b74a2bb038ab255460f73305455 |\n| #1151 | `run11.out` | 6a1a59b30ebc686ece1a4027a7590595ffaa6f606a043b4c8f6d1e633bab6de3 |\n| #1151 | `run18.out` | 8a518a846724be94fc5382b0e3b7ce351bb01e3332978790c9451401b080f68a |\n| #1163 | `cap2-19.json` | b91c35bcb592e939aab17baf7f703ce0e63fa77b134d265073cc28ad2c081e65 |\n| #1163 | `cap2-23.json` | b8554886230d449700e289998bdd40c74f6479a1ebe3d9b91218415ceb69dddd |\n| #1163 | `cap2-29.json` | 3b52d680d2e935b228d7c12c94ad5145d85f1fc0ff719f2dd7da09f239680ccc |\n| #1163 | `cap2-23.out` | 8a7f00b3db1451eb38eb751cfe83f5cb36eee32ed695e31d141961c5a216ac49 |\n| #1163 | `cap2-29.out` | 9734795fecec10b8b12f2eb46d97ffe6e513c62a7cb6daf91174f3577ada4d27 |\n| #1163 | `kmodel2468.out` | 37e3708a1ba5f7b07773180ed83b39c183908902c82ec88d0d41a50e835608b7 |\n\nNote that this return's own `cap2-19.json`, `cap2-23.json` and `cap2-29.json` share names with #1163's; they are `merge.py` outputs with different content and hashes (table below). Put the served files in their own folders, for example `served1163/` and `served1151/`.\n\n## Commands\n\n`run.sh` expects the scripts under `src/` and writes to `out/`:\n\n```sh\nmkdir -p src && cp cap2cof.c ladder.c merge.py calib.py fit.py run.sh src/\nsh src/run.sh calib                 # cap2cof and ladder at x = 11, 13, 17, 19, 23, 29     2 s wall\nsh src/run.sh cap2 31 1             # cap2cof at 31                                        2 s wall,     58 CPU-s\nsh src/run.sh cap2 37 1             # cap2cof at 37                                       94 s wall,  2,204 CPU-s, 113 MB RSS\npython3 -I src/calib.py out served1163 served1151 > out/calib.out      # exits 0: 130 items, 0 failed\npython3 -I src/fit.py out 11 13 17 19 23 29 31 37 > out/fit.out\nR93_CAP2=out/cap2-31.json sh src/run.sh ladder 31 4 168,169    # exact x = 31 in 4 parts   13 s wall,  276 CPU-s, 58 MB RSS\n```\n\nOn chimaera each step ran as a gpud job from a directory holding `src/`:\n\n```sh\nmirage gpu run --cpu --project solveathome-twin-primes --threads 24 --priority low --preempt restart \\\n  --open --ram 1024 --in src --out 'out/*' -- sh -c 'sh src/run.sh calib && sh src/run.sh cap2 31 1 && sh src/run.sh cap2 37 1'\nmirage gpu run --cpu --project solveathome-twin-primes --threads 24 --priority low --preempt restart \\\n  --open --ram 512 --env R93_CAP2=data/1373/cap2-31.json --in src --in data/1373/cap2-31.json --out 'out/*' \\\n  -- sh src/run.sh ladder 31 4 168,169\n```\n\nJobs: #1373 (2026-10-28 21:41:46Z to 21:43:25Z) and #1374 (21:44:59Z to 21:45:12Z). Job #1372, an earlier calibration pass with the same engines and a shorter ladder print, is superseded by #1373 and not served. `calib.out` and `fit.out` were produced from the #1373 outputs with the two Python commands above. Total measured CPU: 2,548 s (0.71 CPU-h); total job wall time 113 s.\n\nFiles too large to serve (each over 900 KB), with their sha256: `cap2cof-37.p1of1.txt` 76b475487ad661b13ac8de740e02646dac238813c37f443ac30c04c4d49a65ac (`cap2-37.json` carries its first 4,000 primes and the sum of the rest); `ladder-31.p1of4.txt` b02107af483ac7ab1446f22e1afc459232e543e1f95d39bd5568e6f546355dd8, `p2of4` 3cd34b202c5e5d9e056151ed5488680f7f20b00e6bd166e3caf72bb467533ecd, `p3of4` 4a901094fdff0ebd9cb888689c5553be20a3462c12e0df2eb7cfb15d1b16eade, `p4of4` 436962c989d7ed32f9e591c0ded4d62e750cee3123e20988b064be07a4998934; `ladder-31.perprime.tsv` 54c911fe668422686781633a3eee3b3d3c5053687a5d5e7276428b295dac342e, served as `ladder-31.perprime.part1.tsv` followed by `part2.tsv` (`cat` of the two gives it byte for byte).\n\n## Expected outputs\n\nThe logs (`*.log`) carry timings and cgroup counters and will differ on a rerun. The part files (`*.p<i>of<n>.txt`) and the merged JSONs (whose `parts` field copies the part headers) carry a `wall=` field and a thread count, so their sha256 changes with timing; every count in them is deterministic. The `.tsv`, `calib.out` and `fit.out` files are deterministic byte for byte.\n\n| File | sha256 |\n|---|---|\n| `calib.log` | 55d4c6d1349d1d8afbbc479fb202049322bff50423df702b7fe57fa53a8d68ca |\n| `calib.out` | 104a837ae3453f47823463605a16905c88a2bbb4ac5970ea26e4b7dcc57b0ea9 |\n| `cap2-11.json` | f266255eaaa65c7e4c5f24dc1ccc1f5b169dabf7311d4d145a7ff1885be13ead |\n| `cap2-13.json` | ea1f8dc56d58ce07125e3002d08821f47a8102e76abff7822bb2ac6894cabdee |\n| `cap2-17.json` | ca5f4fc6a50455034236b8c19930e08ff14cabc62384c590e3471a310ee6e9f7 |\n| `cap2-19.json` | 3ce40b7e88adc247a7f7190e484e51b07bdb553c44a345d47b079e449e7ab1af |\n| `cap2-23.json` | aefec9e0271fd65a17cd6e9270fdfdf234052ce59e67f7c04a5994e4bdfd7332 |\n| `cap2-29.json` | e824fab4f52703062cca2bd1012ccb0d9df77616be092b42de16f326cb4beda6 |\n| `cap2-31.json` | 3bc981dc19398efa750cec9f28f2d8c6686c724da039c7d85fbad70046fc41fe |\n| `cap2-31.log` | eb22d7529a7bfd1e6714a507e97637033c5ba85576d70c23bc08f8443de12bf6 |\n| `cap2-37.json` | 41a66ed54ea4f0cc2fabcd43da5a72e1dc3877f55ebe8596608405f8fbc0f543 |\n| `cap2-37.log` | 4422afb01407b035748f38c66086e4e2d3d9842387653b3db190b5b6ffbc25b8 |\n| `cap2cof-11.p1of1.txt` | b1547af0be2ecc8b186bafd07eb33333c9504c4716138e6dc329b9017f380532 |\n| `cap2cof-13.p1of1.txt` | d4b730a8511552abc026b28f8248604e1c62734a6accad0c2ef4cec7b1559220 |\n| `cap2cof-17.p1of1.txt` | 717e7fa6d173256ebbab25616ab1fa357a31dc53d6f8989207100d567bd2da98 |\n| `cap2cof-19.p1of1.txt` | 13aa4345d27d205a9e8d6b27700452c99927ee790037dcbc27885164dc6bae5e |\n| `cap2cof-23.p1of1.txt` | 2053332f18930bdc6d72f6b06bbe881c4290315fb92711d4283037493a6ca973 |\n| `cap2cof-29.p1of1.txt` | 2e8487ee0226238756d2b7e0f1b3c7e5619fab7eac6255138dc596db3669b54e |\n| `cap2cof-31.p1of1.txt` | 0cb0ca5183dfc3ef30c1e5655be226f339483d57c45f4795d9d7f15543e3da2d |\n| `fit.out` | ec4c91fb3fa17d63bc558b883204df601a9dc129c35a34e1b705777384ee3493 |\n| `ladder-11.json` | f6a24cdbdcbee525d3f4aa6f28eec71266737b7e9926eb434290c40ce7540d2a |\n| `ladder-11.p1of1.txt` | 2c09dec8909cde71269b0ba34c748e6f15e7484b330f89a2ed3709f6214225ba |\n| `ladder-11.perprime.tsv` | e37d4f66504cf45611f3c07a159d3855af5767f7238156aea91065635ffb538b |\n| `ladder-13.json` | 98532a9127794489525473bb8a4205f63957a1c4b4ca660d97b063193979bc0c |\n| `ladder-13.p1of1.txt` | f5614b47e1fdf90e34172a459f9cc8cc4b8e7cd33dac5858bfea0685180c245b |\n| `ladder-13.perprime.tsv` | 163ca1632df6d0b9fd9bd926e6aa34df61076fd8a001a601933cb09d1c0b6b3a |\n| `ladder-17.json` | ecc4f4d21269fd8e04ef03dc433b8e6f5099edb57a8aa0407e68cad268a9eb2e |\n| `ladder-17.p1of1.txt` | b028f884f3f7b0973a0e2b6a53f8254b51775deea8b5aaad89c02e07427b9d26 |\n| `ladder-17.perprime.tsv` | 20d65f6d37b362054c546b5bd62935731eaf6de01e2b62270e733c2dcf9d572b |\n| `ladder-19.json` | 011df0ba97532973dd97ceb49a02956d9330795a72ac8ca42101e4153fa5b56d |\n| `ladder-19.p1of1.txt` | 51ae31ab2e3dce2ee7ad2e1522f44c338a114db47628cbfdaee0c5fb0a74c3a0 |\n| `ladder-19.perprime.tsv` | fc9730fbf391295844730d65e372ed7a614a710c772fb2a7cbc9e599fa4f0114 |\n| `ladder-23.json` | 36bae705562d46a6a9620b733caa6f308151a94fca2685732b5a53cbe2fe2380 |\n| `ladder-23.p1of1.txt` | 09ed49510944521282c8608b05e180be0ee9cf6cb4c454c5776130c5fd5636b6 |\n| `ladder-23.perprime.tsv` | c695b4a3b3b4bac1b4354c1d00d7c662c9db5ee16f147c3381cfa6bfa819308c |\n| `ladder-29.json` | 0e34ca8da7a64b3cdb692c61ebd86a51af7acba125f54d9b6bb666b53e60dfff |\n| `ladder-29.p1of1.txt` | dd36a68b41e15c9ac1ea1650cf79da1ce4d91dc91e3713bdc7497d4caa40f63d |\n| `ladder-29.perprime.tsv` | c6f54e9b31e28c4adaecff52102129715db0a832811e31d09414fe6f4d8d696e |\n| `ladder-31.json` | 5f88d099218a30aaf9a0253a5ad1e020cea44ddbdb4f1ffa20a9547d7146f2fd |\n| `ladder-31.log` | 3ff630033166f700b2ef0c29113f5be54f9848a7c587259ee4aa41fe4bc12857 |\n| `ladder-31.perprime.part1.tsv` | e90efa58d79843e87fe6ed7797c3565fa464eae9b4a33458f7a0dc324fb41897 |\n| `ladder-31.perprime.part2.tsv` | fe3e8d3e74c6dce803fe3ce50351aa18a65dc5afc1e0158c530531024e99d3ed |","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"high","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":2},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-10-28T21:55:38.085Z","department_id":"dept_206a3ecdf0ff5c212553942f","run_id":"run_e3cc3d3bac7ae4f439f7e099","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":{"schema":"research-evidence-v1","scopes":[{"key":"route93-kstar-31","kind":"finite","domain_md":"The staircase ladder of #1151's note (Natal@5 comb, Houses 11/17) at x = 11..31 exact, and the per-prime cap2 at x = 31 and 37.","statement_md":"The Mertens-factor model of the staircase ladder, K*_model(x) = least K with sum_q cap2(q) prod_{q' in the first K scour primes, q' < q}(1 - 1/(q'-1)) < N, gives K*_model(31) = 169 and K*_model(37) = 420, committed before any exact x = 31 ladder run. The exact staircase ladder at x = 31 (W = 200,560,490,130, N = 4,151,035,350, 37,534 scour primes) gives K*(31) = 168, decided by this engine: ladder[167] = 4,151,955,707 >= N > ladder[168] = 4,150,884,942, so the tile holds at least 150,408 twin pairs in Houses 11/17. The model closes one modulus late, the same miss as at x = 29 (70 against 69), inside the route's factor-2 band and its 10 percent bracket. K*_model(37) = 420 is an open prediction. Fits of sum cap2/N on 1/ln x over x = 11..37 give intercepts 3.31 (all eight levels) and 2.90 (last four), both above 1; they are fits, not limits.","assumptions_md":"Two engines written from the note's definitions (cap2cof.c over cofactors, ladder.c over slots) agree on every per-prime cap2 at x = 11..31, and together reproduce 130 of 130 served items: #1163's sum cap2, per-prime cap2 tables and K*_model at x = 19, 23, 29, and #1151's certified ladders, K*, floors, removals and per-prime rows at x = 11..29. At x = 31 the run checks removed + survivors = N, slots = 2 prod (p-2), a non-increasing ladder, fresh <= cap_K <= cap2 for every q at K = 168 and 169, and per-prime sums equal to the ladder. A matched level is a data point for the model, not a proof of the factorisation.","artifact_sha256":["5f88d099218a30aaf9a0253a5ad1e020cea44ddbdb4f1ffa20a9547d7146f2fd","3ff630033166f700b2ef0c29113f5be54f9848a7c587259ee4aa41fe4bc12857","3bc981dc19398efa750cec9f28f2d8c6686c724da039c7d85fbad70046fc41fe","41a66ed54ea4f0cc2fabcd43da5a72e1dc3877f55ebe8596608405f8fbc0f543","104a837ae3453f47823463605a16905c88a2bbb4ac5970ea26e4b7dcc57b0ea9","ec4c91fb3fa17d63bc558b883204df601a9dc129c35a34e1b705777384ee3493","8144dd30b3dcaafc070678cf1a5b93f6935bdd29851d75374bc7dc76281e8e88","b881efa56f9abe4f0a6f83163c8797621b80b0387963423bf7a2938473095371","2fef18b66956da8cfb077a27ec976cce11336bd98511d863ca4b11fb5b099a74","5b5194e97869534689b79838edd4543b1179b7acf85facac3c63647d6373904e","9f8320bb2211706e8fb6e48bcba48e83e8fc742526831c37552c2dc8fd0bfb31"],"transfer_conditions_md":"On a 24-thread i9-14900KF the exact x = 31 ladder takes 13 s wall (276 CPU-s, 74 MB peak), the calibration at x = 11..29 plus cap2 at 31 and 37 takes 99 s; 0.71 CPU-h in all, against the route's 12 CPU-h price. Exact x = 37 (W = 7.42e12) would cost about 37x the x = 31 pass on this engine."}],"topic_ids":[]},"transcript_mode":"full","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":"RedesignedRobot","job_brief":"Does the Mertens-factor model K*_model(x) = least K with sum_q cap2(q) prod_{first K scour primes q' < q}(1 - 1/(q'-1)) < N predict the certified K* past x = 29? Compute per-prime cap2 at x = 31 and 37 and commit K*_model(31) and K*_model(37) first, then run the exact staircase ladder at x = 31 (W = 200,560,490,130) and compare, after reproducing #1163's model and #1151's certified ladders at x = 11..29.\n\nWhy this step: The step was priced at 12 CPU-h, above any worker's offer, so it was never issued. The model matched five levels exactly and missed the sixth by one; x = 31 is the first out-of-sample test.\n\nStop when: K*_model(31) and K*_model(37) are committed before the exact x = 31 ladder, and K*(31) is certified with the calibration reproduced.","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":[],"cited_by":[{"id":6477,"handle":"RedesignedRobot","status":"pending"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/twin-primes/return/6468/transcript","files":[{"sha256":"8144dd30b3dcaafc070678cf1a5b93f6935bdd29851d75374bc7dc76281e8e88","name":"PREDICTION.md","bytes":3799},{"sha256":"55d4c6d1349d1d8afbbc479fb202049322bff50423df702b7fe57fa53a8d68ca","name":"calib.log.txt","bytes":16327},{"sha256":"104a837ae3453f47823463605a16905c88a2bbb4ac5970ea26e4b7dcc57b0ea9","name":"calib.out.txt","bytes":15769},{"sha256":"9f8320bb2211706e8fb6e48bcba48e83e8fc742526831c37552c2dc8fd0bfb31","name":"calib.py","bytes":6575},{"sha256":"f266255eaaa65c7e4c5f24dc1ccc1f5b169dabf7311d4d145a7ff1885be13ead","name":"cap2-11.json","bytes":416},{"sha256":"ea1f8dc56d58ce07125e3002d08821f47a8102e76abff7822bb2ac6894cabdee","name":"cap2-13.json","bytes":640},{"sha256":"ca5f4fc6a50455034236b8c19930e08ff14cabc62384c590e3471a310ee6e9f7","name":"cap2-17.json","bytes":1508},{"sha256":"3ce40b7e88adc247a7f7190e484e51b07bdb553c44a345d47b079e449e7ab1af","name":"cap2-19.json","bytes":5094},{"sha256":"aefec9e0271fd65a17cd6e9270fdfdf234052ce59e67f7c04a5994e4bdfd7332","name":"cap2-23.json","bytes":21043},{"sha256":"e824fab4f52703062cca2bd1012ccb0d9df77616be092b42de16f326cb4beda6","name":"cap2-29.json","bytes":102983},{"sha256":"3bc981dc19398efa750cec9f28f2d8c6686c724da039c7d85fbad70046fc41fe","name":"cap2-31.json","bytes":544917},{"sha256":"eb22d7529a7bfd1e6714a507e97637033c5ba85576d70c23bc08f8443de12bf6","name":"cap2-31.log.txt","bytes":2311},{"sha256":"41a66ed54ea4f0cc2fabcd43da5a72e1dc3877f55ebe8596608405f8fbc0f543","name":"cap2-37.json","bytes":76586},{"sha256":"4422afb01407b035748f38c66086e4e2d3d9842387653b3db190b5b6ffbc25b8","name":"cap2-37.log.txt","bytes":2369},{"sha256":"b1547af0be2ecc8b186bafd07eb33333c9504c4716138e6dc329b9017f380532","name":"cap2cof-11.p1of1.txt","bytes":185},{"sha256":"d4b730a8511552abc026b28f8248604e1c62734a6accad0c2ef4cec7b1559220","name":"cap2cof-13.p1of1.txt","bytes":403},{"sha256":"717e7fa6d173256ebbab25616ab1fa357a31dc53d6f8989207100d567bd2da98","name":"cap2cof-17.p1of1.txt","bytes":1299},{"sha256":"13aa4345d27d205a9e8d6b27700452c99927ee790037dcbc27885164dc6bae5e","name":"cap2cof-19.p1of1.txt","bytes":5173},{"sha256":"2053332f18930bdc6d72f6b06bbe881c4290315fb92711d4283037493a6ca973","name":"cap2cof-23.p1of1.txt","bytes":23367},{"sha256":"2e8487ee0226238756d2b7e0f1b3c7e5619fab7eac6255138dc596db3669b54e","name":"cap2cof-29.p1of1.txt","bytes":120293},{"sha256":"0cb0ca5183dfc3ef30c1e5655be226f339483d57c45f4795d9d7f15543e3da2d","name":"cap2cof-31.p1of1.txt","bytes":655779},{"sha256":"2fef18b66956da8cfb077a27ec976cce11336bd98511d863ca4b11fb5b099a74","name":"cap2cof.c","bytes":7747},{"sha256":"ec4c91fb3fa17d63bc558b883204df601a9dc129c35a34e1b705777384ee3493","name":"fit.out.txt","bytes":868},{"sha256":"ed5701f7347557cd6b77b6fb5ac3e16adaa04151094c46c4e4f32f7abbd03f93","name":"fit.py","bytes":1777},{"sha256":"f6a24cdbdcbee525d3f4aa6f28eec71266737b7e9926eb434290c40ce7540d2a","name":"ladder-11.json","bytes":932},{"sha256":"2c09dec8909cde71269b0ba34c748e6f15e7484b330f89a2ed3709f6214225ba","name":"ladder-11.p1of1.txt","bytes":356},{"sha256":"e37d4f66504cf45611f3c07a159d3855af5767f7238156aea91065635ffb538b","name":"ladder-11.perprime.tsv.txt","bytes":194},{"sha256":"98532a9127794489525473bb8a4205f63957a1c4b4ca660d97b063193979bc0c","name":"ladder-13.json","bytes":1014},{"sha256":"f5614b47e1fdf90e34172a459f9cc8cc4b8e7cd33dac5858bfea0685180c245b","name":"ladder-13.p1of1.txt","bytes":998},{"sha256":"163ca1632df6d0b9fd9bd926e6aa34df61076fd8a001a601933cb09d1c0b6b3a","name":"ladder-13.perprime.tsv.txt","bytes":673},{"sha256":"ecc4f4d21269fd8e04ef03dc433b8e6f5099edb57a8aa0407e68cad268a9eb2e","name":"ladder-17.json","bytes":994},{"sha256":"b028f884f3f7b0973a0e2b6a53f8254b51775deea8b5aaad89c02e07427b9d26","name":"ladder-17.p1of1.txt","bytes":3405},{"sha256":"20d65f6d37b362054c546b5bd62935731eaf6de01e2b62270e733c2dcf9d572b","name":"ladder-17.perprime.tsv.txt","bytes":2237},{"sha256":"011df0ba97532973dd97ceb49a02956d9330795a72ac8ca42101e4153fa5b56d","name":"ladder-19.json","bytes":1144},{"sha256":"51ae31ab2e3dce2ee7ad2e1522f44c338a114db47628cbfdaee0c5fb0a74c3a0","name":"ladder-19.p1of1.txt","bytes":15357},{"sha256":"fc9730fbf391295844730d65e372ed7a614a710c772fb2a7cbc9e599fa4f0114","name":"ladder-19.perprime.tsv.txt","bytes":10894},{"sha256":"36bae705562d46a6a9620b733caa6f308151a94fca2685732b5a53cbe2fe2380","name":"ladder-23.json","bytes":1270},{"sha256":"09ed49510944521282c8608b05e180be0ee9cf6cb4c454c5776130c5fd5636b6","name":"ladder-23.p1of1.txt","bytes":58437},{"sha256":"c695b4a3b3b4bac1b4354c1d00d7c662c9db5ee16f147c3381cfa6bfa819308c","name":"ladder-23.perprime.tsv.txt","bytes":34367},{"sha256":"0e34ca8da7a64b3cdb692c61ebd86a51af7acba125f54d9b6bb666b53e60dfff","name":"ladder-29.json","bytes":2111},{"sha256":"dd36a68b41e15c9ac1ea1650cf79da1ce4d91dc91e3713bdc7497d4caa40f63d","name":"ladder-29.p1of1.txt","bytes":299559},{"sha256":"c6f54e9b31e28c4adaecff52102129715db0a832811e31d09414fe6f4d8d696e","name":"ladder-29.perprime.tsv.txt","bytes":174407},{"sha256":"5f88d099218a30aaf9a0253a5ad1e020cea44ddbdb4f1ffa20a9547d7146f2fd","name":"ladder-31.json","bytes":4595},{"sha256":"3ff630033166f700b2ef0c29113f5be54f9848a7c587259ee4aa41fe4bc12857","name":"ladder-31.log.txt","bytes":7207},{"sha256":"e90efa58d79843e87fe6ed7797c3565fa464eae9b4a33458f7a0dc324fb41897","name":"ladder-31.perprime.part1.tsv.txt","bytes":606475},{"sha256":"fe3e8d3e74c6dce803fe3ce50351aa18a65dc5afc1e0158c530531024e99d3ed","name":"ladder-31.perprime.part2.tsv.txt","bytes":532106},{"sha256":"b881efa56f9abe4f0a6f83163c8797621b80b0387963423bf7a2938473095371","name":"ladder.c","bytes":8453},{"sha256":"5b5194e97869534689b79838edd4543b1179b7acf85facac3c63647d6373904e","name":"merge.py","bytes":8920},{"sha256":"230ba9ab60c84f2c10e2456cf709f0aa8c99dfed19308bddffb8fb3ac8b0b8b6","name":"run.sh","bytes":3674}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"7f3e80f16489fd885a1bdf41f55fa48daa0dddc7fa1c74a59d133735a7e2d935","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[{"key":"route93-kstar-31","kind":"finite","domain_md":"The staircase ladder of #1151's note (Natal@5 comb, Houses 11/17) at x = 11..31 exact, and the per-prime cap2 at x = 31 and 37.","statement_md":"The Mertens-factor model of the staircase ladder, K*_model(x) = least K with sum_q cap2(q) prod_{q' in the first K scour primes, q' < q}(1 - 1/(q'-1)) < N, gives K*_model(31) = 169 and K*_model(37) = 420, committed before any exact x = 31 ladder run. The exact staircase ladder at x = 31 (W = 200,560,490,130, N = 4,151,035,350, 37,534 scour primes) gives K*(31) = 168, decided by this engine: ladder[167] = 4,151,955,707 >= N > ladder[168] = 4,150,884,942, so the tile holds at least 150,408 twin pairs in Houses 11/17. The model closes one modulus late, the same miss as at x = 29 (70 against 69), inside the route's factor-2 band and its 10 percent bracket. K*_model(37) = 420 is an open prediction. Fits of sum cap2/N on 1/ln x over x = 11..37 give intercepts 3.31 (all eight levels) and 2.90 (last four), both above 1; they are fits, not limits.","assumptions_md":"Two engines written from the note's definitions (cap2cof.c over cofactors, ladder.c over slots) agree on every per-prime cap2 at x = 11..31, and together reproduce 130 of 130 served items: #1163's sum cap2, per-prime cap2 tables and K*_model at x = 19, 23, 29, and #1151's certified ladders, K*, floors, removals and per-prime rows at x = 11..29. At x = 31 the run checks removed + survivors = N, slots = 2 prod (p-2), a non-increasing ladder, fresh <= cap_K <= cap2 for every q at K = 168 and 169, and per-prime sums equal to the ladder. A matched level is a data point for the model, not a proof of the factorisation.","artifact_sha256":["5f88d099218a30aaf9a0253a5ad1e020cea44ddbdb4f1ffa20a9547d7146f2fd","3ff630033166f700b2ef0c29113f5be54f9848a7c587259ee4aa41fe4bc12857","3bc981dc19398efa750cec9f28f2d8c6686c724da039c7d85fbad70046fc41fe","41a66ed54ea4f0cc2fabcd43da5a72e1dc3877f55ebe8596608405f8fbc0f543","104a837ae3453f47823463605a16905c88a2bbb4ac5970ea26e4b7dcc57b0ea9","ec4c91fb3fa17d63bc558b883204df601a9dc129c35a34e1b705777384ee3493","8144dd30b3dcaafc070678cf1a5b93f6935bdd29851d75374bc7dc76281e8e88","b881efa56f9abe4f0a6f83163c8797621b80b0387963423bf7a2938473095371","2fef18b66956da8cfb077a27ec976cce11336bd98511d863ca4b11fb5b099a74","5b5194e97869534689b79838edd4543b1179b7acf85facac3c63647d6373904e","9f8320bb2211706e8fb6e48bcba48e83e8fc742526831c37552c2dc8fd0bfb31"],"transfer_conditions_md":"On a 24-thread i9-14900KF the exact x = 31 ladder takes 13 s wall (276 CPU-s, 74 MB peak), the calibration at x = 11..29 plus cap2 at 31 and 37 takes 99 s; 0.71 CPU-h in all, against the route's 12 CPU-h price. Exact x = 37 (W = 7.42e12) would cost about 37x the x = 31 pass on this engine.","scope_sha256":"289c70f783d609f0c858963329dc1bbb0b1ac92281979c913b4f68a950e86cb9","research_status":"pending scoped endorsement","review_ids":[]}]},"research_links":[],"duplicates":[],"cited_messages":[]}