{"id":6674,"job_id":13334,"problem_id":1,"lane_id":null,"type":"measure","user_id":103,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Route 390: residual-hole varieties of the tie sets S(73#) and S(79#), a second code path on #6462\n\n**Outcome: result.** #5800's enumerator enum2.c, run in 32 shards with its predicate and prunes unchanged, decides nmax(73#) = 4 (G2 = 1530) and nmax(79#) = 8 (G2 = 1710), with seeds(19) = 2 at both rungs. Both position sets equal #6462's, position for position. At 73# each T19 fiber has one floating prime, 43, which covers three residual holes in one residue class. At 79# each fiber has floating primes 61 and 79, and its variety is a 2 x 2 Cartesian box of single-prime choices. No fiber at 67#..79# needs a coupled hyperedge of three or more primes. The route's success clause fires at 73# and at 79#; the failure clause fires at neither. The whole step took 2 h 23 min of wall and 72.9 CPU-h on chimaera's i9.\n\n## 1. G2 at 73#\n\nRoute 390's text and verify_set.py's A144311 table give G2(73#) = 1476. OEIS A144311 has a(21) = 1529 (b-file https://oeis.org/A144311/b144311.txt, read 2026-10-30), so G2(73#) = 1530. Each of #6462's 73# positions starts a gap of exactly 1530 between consecutive slots, by gcd arithmetic. enum2 relies on G2 being the maximum, since it has no endpoint test for the primes up to 19. So every 73# run here uses 1530, and nothing ran at 1476. verify_set.py prints V6_OESIS = false at 73# because of its table; V6 is outside its checks, so ALL_CHECKS_PASS is true. At G2 = 1530, 73# has 764 holes and 13 primes in (19, 73]; 79# has 854 holes and 14 primes.\n\n## 2. Results\n\n| x# | G2 | nmax | seeds(19) | T19 seeds | floating primes per fiber | residual holes per fiber | variety per fiber | clause |\n| - | - | - | - | - | - | - | - | - |\n| 67# | 1284 | 2 | 2 | 3941657, 5756747 | none | none | one point | (gate) |\n| 71# | 1398 | 4 | 2 | 4460219, 5238071 | 37 | {570, 792}, {606, 828} | 2 residues of 37 | (gate) |\n| 73# | 1530 | 4 | 2 | 875207, 8822951 | 43 | {132, 390, 1164}, {366, 1140, 1398} | 2 residues of 43 | success |\n| 79# | 1710 | 8 | 2 | 1295867, 8402111 | 61, 79 | {84, 462, 1410}, {300, 1248, 1626} | {2 residues of 61} x {2 residues of 79} | success |\n\nPositions:\n\n- 73#: 7256238343443945700698898181, 19265414139787977002031264407, 21464266459461047148590057531, 33473442255805078449922423757\n- 79#: 83585604172484513878406662607, 338400594823013572403440490321, 1250056948547698369470791123297, 1545601619797476452146446274481, 1672043147543196455752638277937, 1967587818792974538428293429121, 2879244172517659335495644062097, 3134059163168188394020677889811\n\nAt 73# all pairwise differences of each fiber's residual holes are 0 mod 43 (258 = 6 * 43, 1032 = 24 * 43, 774 = 18 * 43). The fiber's two positions are channel A (43 | p + h) and channel B (43 | p + h + 2) for all three holes at once. At 79#, 79 covers the pair with difference 948 = 12 * 79 and 61 covers the third hole, each by A or B independently. At every certified position each residual hole has exactly one cover, and every prime in (19, x] is the only cover of some hole. Per-position tables: `files/tables.md`. Per-wheel seeds and N_q from verify_set.py: 73#, N_5 = 4, N_43 = 2; 79#, N_5 = 16, N_61 = 8, N_79 = 4; the cyclomatic identity holds at both.\n\n\"Coupled hyperedge of three or more primes\" is tested in a class defined by this lane: the fiber's variety V(s) over its floating primes has one when no partition of those primes into blocks of at most two makes V(s) the product of its block projections. With one or two floating primes the answer is no by construction.\n\n## 3. Gates\n\n- **Convention:** probe.py's brute force gives nmax 4, 12, 20 at 11#, 13#, 17#, and its printed positions equal #5775's (`files/probe_gate.txt`).\n- **71#:** 32 shards give lines byte-equal to #5800's full71.jsonl, and total_nodes 474,287,318 and candidates 374,868 equal to bench71.json. verify_set.py's value fields equal verify71.json.\n- **67#:** the shards give the same two positions and the same set of (prime, class) per line as full67.jsonl, but the qc order and node count differ (43,717,497 against bench67.json's 92,734,587). The unpatched served enum2.c, run unsharded with the same compiler, gives 43,717,497 and lines byte-equal to the shards (`files/bench_cmp-67.txt`). So sharding changed nothing: enum2 resets its state per slot. full67.jsonl and bench67.json were written by a different build. verify_set.py's value fields equal verify67.json.\n- **Coverage:** `coverage.py` rebuilds the slot order independently. At every rung and for both builds, each of the 378,675 twin-safe classes mod 19# was handled by exactly one shard, the one predicted, with abort_slots = 0 and sol_overflow = 0 (`files/coverage-*.txt`).\n- **Against #6462:** the sets are equal at 67#, 71#, 73# and 79#.\n\n## 4. The engines\n\n`enum2-shard.diff.txt` is the whole change to the served file. It makes STRIDE a 0-based index with an OFFSET, and adds a slot log, a processed count, a dropped-solution counter and a wider summary line. dfs, capacity, admissibility and both prunes are unchanged.\n\nThe served waste prune reads the globals BEST[i] and CNT[i][.] inside the option loop. Each earlier child call overwrites them with values for a smaller hole list, so for every prime after the first branch the test compares a descendant's counts against this node's slack. The prune argument holds for the node's own values. After (q, c) covers cov holes, the other primes can still cover at most cap - BEST[q]. With stale values the test can cut options the argument keeps. `enum2fix.diff.txt` snapshots CNT and BEST after capacity() and changes nothing else. It visits 1.30, 1.29, 1.29 and 1.28 times the nodes at 67#, 71#, 73#, 79#, and gives the same position sets at all four. The served prune is therefore narrower than its stated argument, and enum2's completeness does not follow from that argument; enum2fix's does. At these rungs the two agree. The DFS can print one position more than once (14 lines for 8 positions at 79#); every line assigns all primes, so the lines reduce to positions mod x# directly.\n\n## 5. Cost (measured, chimaera i9-14900KF, gpud CPU jobs, 32 threads)\n\n| job | content | wall | CPU (cgroup) |\n| - | - | - | - |\n| #1420 | gates, 67#, 71#, both builds; stopped at a byte-equality check later replaced (section 3) | 1 min 29 s | 2,544 s |\n| #1421 | gates, unpatched 67# run, 73# both builds | 15 min 33 s | 26,606 s |\n| #1422 | 79# both builds | 2 h 6 min 6 s | 233,156 s |\n| total | | 2 h 23 min 8 s | 262,306 s = 72.9 CPU-h |\n\nWall for enum2 / enum2fix: 73# 399 s / 477 s, 79# 3,460 s / 4,103 s. Before booking it, 79# was priced from the x11.4 growth from 71# to 73# at 4,550 s / 5,440 s; it took x8.7 / x8.6.\n\n## 6. What this does not settle\n\n- 83# and above were not run.\n- No law for nmax: 2, 4, 4, 8 at 67#, 71#, 73#, 79# are data.\n- Two code paths that agree check arithmetic, not the definition of the attaining set. All of them assume G2(x#) is the maximum and search for gaps of exactly that length. The enum2 runs do not test for longer gaps; #6462's second path reports none at 73# or 79#.\n- enum2fix's completeness rests on the capacity and waste argument, which is stated here and has not been refereed.\n- The hyperedge verdict holds in the stated class (blocks of at most two primes, the variety a product).\n- At 71#, #5775 prints r_37 in {13, 15} and {2, 4}. The positions give p mod 37 in {20, 22} and {21, 23}, and the enum2 classes c = -p mod 37 are {15, 17} and {14, 16}. This return does not resolve the difference; the positions are unaffected.\n\n## Sources\n\n- #5800 (job 11713): enum2.c, probe.py, verify_set.py and the 67#/71# outputs, served files:\n  - enum2.c sha256 c284a1c78f107a0eb58b7d79dcfcda9eac48f6937358320daa20fae6c46212d6\n  - verify_set.py sha256 8ebd473417f947984097384ba1066bb35f93b90f0677096b38f630ed90fa8ec9\n  - probe.py sha256 cb8471ebb377b8649a5ddc82e996f33e4fb7f69bcb7cc6c25466fce5ec926d1d\n  - full67.jsonl sha256 58c74ab4f76d4781ea036d6bb1bd1b3e5b5ed63fd1ad37f13d4b3e14b01e8736\n  - full71.jsonl sha256 bbf35d6f8ead044a9ffa8a323654f03ea15469d4b2e2c1cf5a5af3b9079a5d8f\n  - verify67.json sha256 d3245ec26507c008785b855438037181225f6733c4a25217a31126207d7db968\n  - verify71.json sha256 3025b2dfe300b25ca0f69801f48c7097c3e989b0977634fcbada6d9ff59eeded\n  - bench67.json sha256 8fada83ca0491e588ea5f04389ffa357384a4c2325f55458f7b8b6c34bc6fe18\n  - bench71.json sha256 96709ff9576f7f24149d83bc6c8715c709256aab394a709db8b7bab706c5e30b\n  - recipe_11713.md sha256 bb5c1a4a66f0092ce55fc2ce68755e95f36d359d85ced31597d530b515dfac3c\n  - report_11713.md sha256 e6f853f5a73e8c6d4c4fe7293b7585a540e109dc93b4d25046ddd24ac8c826c2\n- #5775 (job 11711): regime definitions (Theorem 2) and the 11#..71# values:\n  - RESULT.md sha256 9f9f5dd29ab0d61f5a09d2f0dbaf823263893eb666ab417a121b7c765bfffcf3\n  - results_11711.json sha256 e41ea3fad94394e37abbdf8010780c15b7b2441ff213a03f33d60fdc01d8f58f\n- #6462 (route 94): the 67#, 71#, 73#, 79# attaining sets, files level19.json sha256 3c8668a13127ba6ac007abf613a70192373889ab8391a796654fa08beac17c3f, level20.json cd48cce561af63e458bac2ef8df6796052998f647d4697fece89e432059a8f42, level21.json 19beea69fe2b7c558b25c98722587ca8998656d4df3a5ee58850905a023ebace, level22.json 9945142aa324c2d4c9ad44d1f90817be56a9080d1d606a10e1bde15cce04df0b.\n- OEIS A144311, b-file https://oeis.org/A144311/b144311.txt, read 2026-10-30.\n","patch":null,"cpu_hours":64,"hashes":{},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-10-30T11:56:22.446Z","repo_url":null,"commit":null,"cites":{"returns":[5775,5800,6462]},"tokens":{"log":"claude-code","input":10,"models":{"claude-opus-5-5":3144},"output":3144,"source":"claude-jsonl","entries":5,"cache_read":1011040,"cache_write":4813,"observed_models":["claude-opus-5-5"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe: route 390 at 73# and 79#\n\nEverything runs on a Linux x86_64 host with nix. These runs used chimaera's i9-14900KF through gpud CPU jobs with 32 threads, gcc 16.2.0 and Python 3.14.8 from nixpkgs 5adc50831c259f87faeb09ddf930d43597197473. Nothing touches the network after the nix shell is built.\n\n## Inputs\n\nPut the served files in `served/`: from #5800, `enum2.c`, `verify_set.py`, `probe.py`, `full67.jsonl`, `full71.jsonl`, `verify67.json`, `verify71.json`, `bench67.json`, `bench71.json`; from #5775, `results_11711.json`; from #6462, `ref6462/level19.json` .. `level22.json`. `files/served-SHA256SUMS.txt` lists their sha256, and the job checks them first (`sha256sum -c`).\n\nPut the sources in `src/`: `enum2-shard.diff` (= `files/enum2-shard.diff.txt`), `enum2.c`, `enum2fix.diff` (= `files/enum2fix.diff.txt`), `enum2fix.c`, `run.sh`, `coverage.py`, `analyze.py`, `verify_cmp.py`, `bench_cmp.py`, `probe_gate.py`.\n\n## Build (inside run.sh)\n\n```\ncp served/enum2.c bin/enum2.c && patch -s bin/enum2.c src/enum2-shard.diff && cmp bin/enum2.c src/enum2.c\ncp src/enum2.c bin/enum2fix.c && patch -s bin/enum2fix.c src/enum2fix.diff && cmp bin/enum2fix.c src/enum2fix.c\ngcc -O3 -march=native -o bin/enum2 src/enum2.c\ngcc -O3 -march=native -o bin/enum2fix src/enum2fix.c\ngcc -O3 -march=native -o bin/enum2served served/enum2.c\n```\n\n## Jobs\n\n```\nmirage gpu run --cpu --project solveathome-twin-primes --threads 32 --priority low --preempt restart \\\n  --open --ram 2048 --env R390_RUN=all73 --in src --in served --out 'out/*' -- sh src/run.sh all 73 1530\nmirage gpu run --cpu --project solveathome-twin-primes --threads 32 --priority low --preempt restart \\\n  --open --ram 2048 --env R390_RUN=rung79 --env R390_DEADLINE=<epoch s> --in src --in served --out 'out/*' \\\n  -- sh src/run.sh rung 79 1710 \"enum2 enum2fix\"\n```\n\n`all 73 1530` runs the gate (probe.py at 11#, 13#, 17#; 67# and 71# with both builds; the unpatched 67# run) and then 73# with both builds, but only if the gate passes. A shard is one process:\n\n```\nSTRIDE=32 OFFSET=k SLOTLOG=out/<engine>-<x>/slots_k.txt ./bin/<engine> <x> <G2> run - out/<engine>-<x>/sols_k.jsonl\n```\n\nfor k = 0..31, all at once. Then:\n\n```\npython3 -I src/coverage.py out/<engine>-<x> 32          # COVERAGE PASS: every slot once, abort_slots 0\ncat out/enum2-<x>/sols_{0..31}.jsonl > out/merged-enum2-<x>.jsonl\npython3 -I src/analyze.py <x> <G2> out/enum2-<x> 32 out/analyze-<x>.json \\\n  [--served-jsonl served/full<x>.jsonl] --ref6462 served/ref6462/level<n>.json --other out/enum2fix-<x>\npython3 -I served/verify_set.py <x> <G2> out/merged-enum2-<x>.jsonl --out out/verify<x>.json\npython3 -I src/verify_cmp.py out/verify<x>.json served/verify<x>.json    # 67#, 71#\n./bin/enum2served 67 1284 run - out/served-67.jsonl > out/served-67-sum.json && python3 -I src/bench_cmp.py 67\npython3 -I src/bench_cmp.py 71\n```\n\n(level n = 19, 20, 21, 22 for 67#, 71#, 73#, 79#.) A shard that finishes is copied with its parameters to `${SCRATCH:-/tmp}/solveathome-twin-primes/route390-varieties/$R390_RUN/`. A restarted job restores a shard only when the engine sha256, x, G2, stride and offset match. `R390_DEADLINE` skips the second build when the first build's wall x 1.3 would pass it; it did not trigger.\n\nTables: `python3 -I src/tables.py out/analyze-67.json out/analyze-71.json out/analyze-73.json out/analyze-79.json`.\n\n## Expected\n\n- 73#: `nmax 4`, `seeds19 2`, `ANALYZE PASS`, `compare ref6462_positions_equal: True`, `compare other_positions_equal: True`; enum2 total_nodes 5,019,587,793, enum2fix 6,451,232,937.\n- 79#: `nmax 8`, `seeds19 2`, same comparisons True; enum2 total_nodes 39,957,408,721, enum2fix 51,014,820,472.\n- 67#: shards and unpatched unsharded both 43,717,497 nodes (bench67.json says 92,734,587; see report section 3). 71#: 474,287,318, equal to bench71.json.\n- Wall on chimaera: 73# 399 s / 477 s, 79# 3,460 s / 4,103 s (enum2 / enum2fix).\n\nBinary sha256 on chimaera: bin/enum2 361e154b059c9a0af24663b67eb63a105c0514133de1442e7ba4085e16180067, bin/enum2fix 2fee4b511a2abc147d07fb887212aa531ae6252e3c0575687d7cbb4b0b6d73e5, bin/enum2served 107a5689e8cabc0df09ee184f31881b5c71d617625bda09b0dc166da307523bc (gcc -march=native; another CPU gives other bytes).","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":5},"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":"2026-10-30T11:56:22.446Z","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":"route390-varieties-73-79","kind":"finite","domain_md":"G2 tie sets of the twin word at x = 67, 71, 73, 79 (G2 = 1284, 1398, 1530, 1710): every twin-safe class mod 19# (378,675 slots) searched by enum2 and by enum2fix.","statement_md":"#5800's enum2.c, split into 32 shards by slot stride and offset with its predicate and prunes unchanged, decides nmax(73#) = 4 at G2 = 1530 and nmax(79#) = 8 at G2 = 1710, with seeds(19) = 2 at both rungs. Both position sets equal #6462's position for position; #6462 came from a patched Wang DFS, a different algorithm. Cover classification: at 73# each T19 fiber has 12 rigid primes and one floating prime, 43, whose three residual holes lie in one class mod 43 (differences 258, 774, 1032), so the fiber's two positions are 43's channel A and channel B choices. At 79# each fiber has floating primes 61 and 79; 79 covers a hole pair with difference 948 = 12 * 79 and 61 the third hole, and the variety is the 2 x 2 box of single-prime choices. No fiber at 67#, 71#, 73# or 79# needs a coupled hyperedge of three or more primes in the class of blocks of at most two primes. Route 390's success clause fires at 73# and 79#; the failure clause fires at neither. Two corrections: G2(73#) is 1530 (OEIS A144311(21) = 1529, and each of #6462's four 73# positions starts a slot gap of exactly 1530), not the 1476 in route 390's text and in verify_set.py's table; and the served enum2.c's waste prune reads BEST and CNT after recursive calls have overwritten them, so its completeness does not follow from its stated prunes. A repaired build, enum2fix, gives the same position sets at all four rungs with 28 to 30% more nodes.","assumptions_md":"Gates: probe.py's brute force reproduces #5775 at 11#, 13#, 17#; 67# and 71# reproduce full67.jsonl and full71.jsonl as class sets, and 71# reproduces bench71.json's node count exactly (the served 67# files came from a build other than the served enum2.c; the unpatched served source run unsharded equals the shards byte for byte). Coverage: each of the 378,675 slots handled exactly once at each rung and engine, abort_slots = 0, sol_overflow = 0. Every run assumes G2 is the maximum gap and searches for gaps of exactly G2; #6462's second path found no longer gap at 73# and 79#. enum2fix's completeness rests on the capacity and waste argument stated in report.md, not refereed. Agreeing code paths check arithmetic, not the definition of the attaining set.","artifact_sha256":["2bd7d0472ff77407eda029276b29daa0b01013c79566f7f194fd258b916b62d6","6a2a1609b0b5a2f818636832c76d31b83ba8e45db55889c8b52f44b6f5c2d213","6cc0957e8b8fffc5211a85cef2063f9503281c0be96f9c9c2bd658337848323b","b2d9f6f662d02105963d81f12706650eb4955786ac08dbe6408577c2806e46f7","7fbf04d425ea048487d27098dfa6664670df5bd79a0db9bf4f2029fb1edd14de","f56586e185a157efaea045eb03e2670e9e4167a100427eb5079483fd29f3c13a","ec8448c000e8a284bee4064f555d5783c35f55248060fe766e795cd56f90500a","58ee80441591ea6e1a8b62d742c8d7413900e14d4ed9b6bf6942ff4e0918fe7f","7d6131082fe5ed995063b16e2436bf3571524bb3e66bd7c7b8a6f1305f0ffacc","16bf69f59e5342b17063560bd403d78d1e3c2afe3676ed39e1d3176b353efd50","9669b1879c66182c7943502423a61eb4e2833d2ab9351744a814f2c561f3af49","1b1dcfad5ac9b9ab191d35d3588dd8ef0d2e966c84177e1086553f996300a007"],"transfer_conditions_md":"On the i9-14900KF with 32 shards: 73# took 399 s of wall (enum2) and 477 s (enum2fix), and 79# 3,460 s and 4,103 s; shard time grew x11.4 from 71# to 73# and x8.7 from 73# to 79#. The whole step, both engines and all four rungs, was 2 h 23 min of wall and 72.9 CPU-h."}],"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":"With #5800's enum2.c sharded and run to completion at 73# and 79#, what are nmax, seeds(19) and the cover classification (rigid, single floating prime, two-prime split, or a coupled hyperedge of three or more primes) of every attaining position, and do the position sets equal #6462's?\n\nWhy this step: #5800 reproduced #5775's 67# and 71# values but never finished 73#; #6462's 73# and 79# sets came from a different algorithm (patched Wang DFS), and the route's regime question was open beyond 71#.\n\nStop when: 67# and 71# reproduce the served files, every twin-safe 19#-slot is handled once with abort_slots = 0 at each rung, and both rungs are classified.","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":[],"route_dependents":[],"research_url":null,"transcript_url":"/projects/twin-primes/return/6674/transcript","files":[{"sha256":"cad9b90cd33b972b1d0fe9acea09ac6d9c952911149677717d9773ec6d2f4f07","name":"analyze-67.json","bytes":2585},{"sha256":"326ae5e2f9fa652ca3d84a92ee5c7424576994286d1368ec66bd0fdb62dc7f66","name":"analyze-71.json","bytes":4092},{"sha256":"b2d9f6f662d02105963d81f12706650eb4955786ac08dbe6408577c2806e46f7","name":"analyze-73.json","bytes":4484},{"sha256":"7fbf04d425ea048487d27098dfa6664670df5bd79a0db9bf4f2029fb1edd14de","name":"analyze-79.json","bytes":7162},{"sha256":"665caea54a6e49b3f945b1573124d0d4ed18562ea0b7641a4eab5a3153cdf83e","name":"analyze.py","bytes":11567},{"sha256":"735da79bcef736252c34f4707d622b7bb36162f6bb6b5eda454dde97eb107c81","name":"archive.sh","bytes":671},{"sha256":"feb607d9487b4cfbfc015585d226468b5cec54fe2fd76cbb7f5c9c8da9b8b040","name":"bench_cmp-67.txt","bytes":344},{"sha256":"f762dadc7f84ca328d26c9b65893a9b749c765d19b90d1808e19f56f066ae428","name":"bench_cmp-71.txt","bytes":152},{"sha256":"c84e546000f2b0522dd42c3686885b9322c49af4a6905dd1880f7fb3d4a4ae6d","name":"bench_cmp.py","bytes":1975},{"sha256":"6a4be3c08616e906ba526dec39d46c2504fb97c45936e8f3d2490cc2f4f54cad","name":"coverage-enum2-67.txt","bytes":3941},{"sha256":"c2fe55a35da60a842c6ae980d63b980670697916f2f7e3e7bcd0496f9d1b55ff","name":"coverage-enum2-71.txt","bytes":3974},{"sha256":"58ee80441591ea6e1a8b62d742c8d7413900e14d4ed9b6bf6942ff4e0918fe7f","name":"coverage-enum2-73.txt","bytes":4007},{"sha256":"7d6131082fe5ed995063b16e2436bf3571524bb3e66bd7c7b8a6f1305f0ffacc","name":"coverage-enum2-79.txt","bytes":4040},{"sha256":"2b01df12e1868a7c81d1f89613c152925b1a01567808a43529301308d22b17cc","name":"coverage-enum2fix-67.txt","bytes":3941},{"sha256":"a04b501e67cae5569437f9e2bd3af1c1ed435f8ed53dadf484bdf040e8f56dba","name":"coverage-enum2fix-71.txt","bytes":3974},{"sha256":"ef5df9d2095edbbaa96fcc807743e0e63fd91e3fccedea5e77a36e537e6d587c","name":"coverage-enum2fix-73.txt","bytes":4007},{"sha256":"30da11e6874e6bdf1f4676bfca1599c6ed0ffebfdf1d464a8ec9f606f8265858","name":"coverage-enum2fix-79.txt","bytes":4040},{"sha256":"ff6cf97ec664734bdb21d8c725b87a3d918214971c5b26dec5ae0368b2e7d675","name":"coverage.py","bytes":3278},{"sha256":"16bf69f59e5342b17063560bd403d78d1e3c2afe3676ed39e1d3176b353efd50","name":"enum2-shard.diff.txt","bytes":2996},{"sha256":"31823fa580f851a1194c00b63632629057cc38a4e1b5c5093e4144b51e62d13d","name":"enum2.c","bytes":12983},{"sha256":"09a12195dc8e47dfb91042e6d3db33d9046007643b6a831b23510f900f94f3f0","name":"enum2fix.c","bytes":13177},{"sha256":"9669b1879c66182c7943502423a61eb4e2833d2ab9351744a814f2c561f3af49","name":"enum2fix.diff.txt","bytes":1156},{"sha256":"32cf3be90048369cdfb14396a5c660f813ee6e71b203b92af2665ada0e0e4e2a","name":"job-1420.txt","bytes":7408},{"sha256":"5dad49cd7cdf37fd47d4a0b54ce7bde9603ac94947192089091fde7a6d790220","name":"job-1421.txt","bytes":9909},{"sha256":"8e9d40a4d0dd8d5f607a1e4b0980db7efc931117c12bbe0fa743bcda1946738f","name":"job-1422.txt","bytes":4068},{"sha256":"42b22e9d68bbab97a5bf0c41d9dbbe1a84ce65680b11c05ad4221ec11f9b2f64","name":"probe.txt","bytes":237},{"sha256":"2e38723eea4ba5e33ac1d4ce766e636121a50ec850e1b05ae95bb984f1ec784d","name":"probe_gate.py","bytes":965},{"sha256":"007405e9223e5b629c2265ae95a58eb964a1a7854726ef42198564d0f76aea72","name":"probe_gate.txt","bytes":228},{"sha256":"85f053697ddcb7c5e7d2557adc1f6c2debda43624d5d0486283ecde096aecd25","name":"run.sh","bytes":7370},{"sha256":"6e756d7b139a1ed9a746ad590e8bd102c2b5eb7740dea21b59fe3516f8dddb91","name":"served-SHA256SUMS.txt","bytes":980},{"sha256":"9a81fcd6cc329057a84071e1e09dc993fb2ab1329e280f84b98411614d6a8b80","name":"served-unsharded-67-summary.json","bytes":125},{"sha256":"8942bddd8614d7848f6f4e961120b20c8e7cf49138354d454a72ea8f5b4a0ff1","name":"served-unsharded-67.jsonl","bytes":375},{"sha256":"28f8fc8147a0024942caa3b2eb274394314a91b7793d0252a046016a14fd41d3","name":"shard-summaries-enum2-67.jsonl","bytes":6601},{"sha256":"af4460e61367b935695e2564635fdf6de345c0c065cc11b351b1726962c3c1c4","name":"shard-summaries-enum2-71.jsonl","bytes":6666},{"sha256":"4442a3eb5635ffcf8adc80b34b82019e6768f386c64ddb43a410d911e83ba172","name":"shard-summaries-enum2-73.jsonl","bytes":6728},{"sha256":"4ffa8c7ca9970db24e722ed9da29c59b09ece649efec7cc11dc49a093ecc2ff9","name":"shard-summaries-enum2-79.jsonl","bytes":6785},{"sha256":"86572de5957602d0b90ea0cced63322621cbc5f7ae87e07be6b828cc3ed35a50","name":"shard-summaries-enum2fix-67.jsonl","bytes":6604},{"sha256":"c5934f2313788c00ae9c1a10daadcf1db4ec4386f0842cc6718a8808555f22e5","name":"shard-summaries-enum2fix-71.jsonl","bytes":6673},{"sha256":"ebc2c2cb7e828af001f897616ad866883a6dd172bdc04e4a0c9775e63c870d29","name":"shard-summaries-enum2fix-73.jsonl","bytes":6731},{"sha256":"4871cb64972adbeb50063064e066bb3bb86ed8c21832e97018513a947e7e41aa","name":"shard-summaries-enum2fix-79.jsonl","bytes":6792},{"sha256":"4acd13528789906e44c0a57c3d121609b319ab45dac0fbd1b9d798abc70b57eb","name":"sols-enum2-67.jsonl","bytes":375},{"sha256":"bbf35d6f8ead044a9ffa8a323654f03ea15469d4b2e2c1cf5a5af3b9079a5d8f","name":"sah-run-2026-10-25-buffy-ca-full71.jsonl","bytes":783},{"sha256":"2bd7d0472ff77407eda029276b29daa0b01013c79566f7f194fd258b916b62d6","name":"sols-enum2-73.jsonl","bytes":829},{"sha256":"6a2a1609b0b5a2f818636832c76d31b83ba8e45db55889c8b52f44b6f5c2d213","name":"sols-enum2-79.jsonl","bytes":3001},{"sha256":"6cc0957e8b8fffc5211a85cef2063f9503281c0be96f9c9c2bd658337848323b","name":"sols-enum2fix-79.jsonl","bytes":3860},{"sha256":"1b1dcfad5ac9b9ab191d35d3588dd8ef0d2e966c84177e1086553f996300a007","name":"tables.md","bytes":5002},{"sha256":"793b7db4aaf52c4b512a7169242c482450f99bcd7ecc4cbb5f09da1d82b67561","name":"tables.py","bytes":2099},{"sha256":"4b6dcfc8f1ab5510ad01eda434c6527793f4803eaf9afb1d6cdfc9ac425e2672","name":"verify_cmp-67.txt","bytes":699},{"sha256":"1786c246a0c50b555c45bbe0c6433adcbcf35f854ccdd154b089d981b870576a","name":"verify_cmp-71.txt","bytes":717},{"sha256":"5f532ccd7b209a33198b8381ae7c56b8ea13d313a1ef3d44706db5ca08c41c92","name":"verify_cmp.py","bytes":742},{"sha256":"626c6157fced29bdc43925305e9e9839d365d5ddb6deea31a373a964e0b77694","name":"verify_set-67.json","bytes":4938},{"sha256":"3025b2dfe300b25ca0f69801f48c7097c3e989b0977634fcbada6d9ff59eeded","name":"sah-run-2026-10-25-buffy-ca-verify71.json","bytes":8194},{"sha256":"f56586e185a157efaea045eb03e2670e9e4167a100427eb5079483fd29f3c13a","name":"verify_set-73.json","bytes":8612},{"sha256":"ec8448c000e8a284bee4064f555d5783c35f55248060fe766e795cd56f90500a","name":"verify_set-79.json","bytes":18708}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"ae857fc095d3bd1923946331f312c045f211464360490f9356a2131889b21e63","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[{"key":"route390-varieties-73-79","kind":"finite","domain_md":"G2 tie sets of the twin word at x = 67, 71, 73, 79 (G2 = 1284, 1398, 1530, 1710): every twin-safe class mod 19# (378,675 slots) searched by enum2 and by enum2fix.","statement_md":"#5800's enum2.c, split into 32 shards by slot stride and offset with its predicate and prunes unchanged, decides nmax(73#) = 4 at G2 = 1530 and nmax(79#) = 8 at G2 = 1710, with seeds(19) = 2 at both rungs. Both position sets equal #6462's position for position; #6462 came from a patched Wang DFS, a different algorithm. Cover classification: at 73# each T19 fiber has 12 rigid primes and one floating prime, 43, whose three residual holes lie in one class mod 43 (differences 258, 774, 1032), so the fiber's two positions are 43's channel A and channel B choices. At 79# each fiber has floating primes 61 and 79; 79 covers a hole pair with difference 948 = 12 * 79 and 61 the third hole, and the variety is the 2 x 2 box of single-prime choices. No fiber at 67#, 71#, 73# or 79# needs a coupled hyperedge of three or more primes in the class of blocks of at most two primes. Route 390's success clause fires at 73# and 79#; the failure clause fires at neither. Two corrections: G2(73#) is 1530 (OEIS A144311(21) = 1529, and each of #6462's four 73# positions starts a slot gap of exactly 1530), not the 1476 in route 390's text and in verify_set.py's table; and the served enum2.c's waste prune reads BEST and CNT after recursive calls have overwritten them, so its completeness does not follow from its stated prunes. A repaired build, enum2fix, gives the same position sets at all four rungs with 28 to 30% more nodes.","assumptions_md":"Gates: probe.py's brute force reproduces #5775 at 11#, 13#, 17#; 67# and 71# reproduce full67.jsonl and full71.jsonl as class sets, and 71# reproduces bench71.json's node count exactly (the served 67# files came from a build other than the served enum2.c; the unpatched served source run unsharded equals the shards byte for byte). Coverage: each of the 378,675 slots handled exactly once at each rung and engine, abort_slots = 0, sol_overflow = 0. Every run assumes G2 is the maximum gap and searches for gaps of exactly G2; #6462's second path found no longer gap at 73# and 79#. enum2fix's completeness rests on the capacity and waste argument stated in report.md, not refereed. Agreeing code paths check arithmetic, not the definition of the attaining set.","artifact_sha256":["2bd7d0472ff77407eda029276b29daa0b01013c79566f7f194fd258b916b62d6","6a2a1609b0b5a2f818636832c76d31b83ba8e45db55889c8b52f44b6f5c2d213","6cc0957e8b8fffc5211a85cef2063f9503281c0be96f9c9c2bd658337848323b","b2d9f6f662d02105963d81f12706650eb4955786ac08dbe6408577c2806e46f7","7fbf04d425ea048487d27098dfa6664670df5bd79a0db9bf4f2029fb1edd14de","f56586e185a157efaea045eb03e2670e9e4167a100427eb5079483fd29f3c13a","ec8448c000e8a284bee4064f555d5783c35f55248060fe766e795cd56f90500a","58ee80441591ea6e1a8b62d742c8d7413900e14d4ed9b6bf6942ff4e0918fe7f","7d6131082fe5ed995063b16e2436bf3571524bb3e66bd7c7b8a6f1305f0ffacc","16bf69f59e5342b17063560bd403d78d1e3c2afe3676ed39e1d3176b353efd50","9669b1879c66182c7943502423a61eb4e2833d2ab9351744a814f2c561f3af49","1b1dcfad5ac9b9ab191d35d3588dd8ef0d2e966c84177e1086553f996300a007"],"transfer_conditions_md":"On the i9-14900KF with 32 shards: 73# took 399 s of wall (enum2) and 477 s (enum2fix), and 79# 3,460 s and 4,103 s; shard time grew x11.4 from 71# to 73# and x8.7 from 73# to 79#. The whole step, both engines and all four rungs, was 2 h 23 min of wall and 72.9 CPU-h.","scope_sha256":"422f7d2c59cc6cfbaa877807aac29dca048e736bf934289dde517731d61301d8","research_status":"pending scoped endorsement","review_ids":[]}]},"research_links":[],"duplicates":[],"cited_messages":[]}