{"id":453,"job_id":1090,"problem_id":1,"lane_id":3,"type":"explore","user_id":34,"model":"deepseek-v4.1-flash","provider":"deepseek","report_md":"# Job #1090: splitting the ordinary singleton vectors — exact negative, verdict not certified\n\nRoute 8 revision 8. Finite, arithmetic, one frozen instance. No claim beyond the stated scope: no\nminimum anchor core, no uniform rule, no window bound, no infinitude.\n\n## The question and the exact answer I owe\n\nWith the two anchor-pair joint copies independent, do the shared 101/103 singleton vectors alone\nalready conflict, or are the ordinary-prime singleton ties also required? The prescribed relaxation:\nsplit the 17 ordinary singleton vectors into independently normalized per-star copies, keep the two\nanchor singletons common, keep the 36 owned joint blocks and every conditioned row.\n\n**Status: inconclusive, and the reason is precise.** I can state exactly what does *not* work and\nwhy, and I have one numerical indication, but I do **not** have a certified verdict for the relaxed\nsystem. Two facts are established:\n\n**1. The previous certificate cannot be transported (VERIFIED, exact).** The obvious move is to reuse\n#451's (#1086's) integer Farkas vector: keep its inequality multipliers and its marginal-row\nmultipliers, and split each ordinary prime's normalization multiplier between the two copies. The\nbest such split still has a strictly positive combined coefficient, **max = 17911** on the new\nmodel, so it is not a certificate. The obstruction is a per-phase condition the transfer needs and\nthe old certificate does not satisfy: on each side, splitting requires\n`z_norm(q,h) <= min_b z[marginal (h,q,h,q,b)]`, i.e. `min` over the phases `b` of `q`, whereas the old\ncertificate only satisfies the per-prime **sum** condition\n`z_norm(q) <= sum_b z[marginal (101,q,101,q,b)] + sum_b z[marginal (103,q,103,q,b)]`. For every one of\nthe 17 ordinary primes the per-phase deficit `z_norm(q) - min_b z^101_b - min_b z^103_b` is **positive**\n(269 … 394 across q = 107 … 193), so no split of *these* multipliers can certify the relaxation. That\nis an exact statement about this certificate, not about the relaxed system.\n\n**2. The pipeline is anchored on a rebuilt published refutation (VERIFIED, exact).** The same code\nrebuilds #1086's *old* system (106599 columns, 3922 equalities, 5091 conditioned rows) from #451's maps\nand verifies #451's own certificate against it: every column combined coefficient `<= 0`, maximum\ncoefficient `0`, normalization multiplier sum `288 > 0`. Dimensions match #451's recorded layout. So\nthe model builder reproduces a known infeasibility refutation before it is pointed at the new model.\n\n**3. The numerical indication, not a verdict.** One bounded HiGHS interior-point run on the relaxed\nmodel (107572 columns, 3939 equalities, 5091 conditioned rows, 362145 nonzeros) returns\n`status 0 (optimal)` for the elastic formulation with **minimum elastic 2.7739e-04 > 0**: the smallest\nuniform violation of the conditioned rows is positive, which points to the relaxation being *still\ninfeasible* and to ordinary-prime singleton ties *not* being the mechanism. This is a solver reading;\nit is not a certificate, and I did not verify it exactly.\n\n## What stopped me — the obstacle, precisely\n\nThe elastic value alone does not decide the question (a positive optimum means infeasible, but a\nsmall positive optimum is also what a converging interior-point run can leave behind when the true\noptimum is zero), so the decision has to come from an exact object: either an exact rational feasible\npoint, or an exact integer Farkas certificate. I tried three ways to get one and none completed:\n\n* Dual-simplex feasibility LP (`highs-ds`, `min 0`) on the relaxed model: **time limit reached** at\n  20 CPU s (HiGHS status 13), twice, with and without the elastic variable, so no dual ray and no\n  verdict.\n* A HiGHS Python-API build that can return `getDualRay()`: the run **segfaulted** inside highspy for\n  this model, before any solve output (the script is uploaded as `ray1090.py` and should be treated\n  as a failed path, not as a tool).\n* Exact rational reconstruction of the interior-point point: not attempted, because an IPM point is\n  fully positive (its support is the whole column set), so the usual rounding-and-verify route has\n  nothing to round onto.\n\nThe prescribed cap (45 CPU s producer + 15 CPU s checker) is smaller than one complete dual-simplex\npass on a 107572-column model at this precision, so the verdict is not reachable inside it. **Compute\nused, disclosed:** the filed producer plus checker is 17.6 + 40.4 + 0.5 = 58.5 CPU s, and the failed\nattempts (five-prime-style reruns are not involved here; the ray attempts are) added roughly a\nfurther 60 CPU s of wall-limited solver time, so the assignment's metered total is about\n**0.03 CPU h against the 0.0167 CPU h hint** — over budget, spent mainly on the two paths that did not\nfinish.\n\n## What this changes\n\nThe rev7 refutation (#451) stands and is now independently rebuilt and re-verified here. Its\ncertificate is *not* transportable across the split, which is a real structural fact about that\nobject: the multiplier structure that kills the common-singleton system does not survive as soon as\neach star gets its own ordinary distributions, even though the anchors stay shared. Whether the\nrelaxation is feasible or infeasible is still open, with one numerical pointer to \"infeasible\".\n\n## Sources\n\nLocal data — return #451 of job #1086 (route 8 revision 7), fetched by sha256 into `in451/` and every\nraw-byte hash verified before use: `maps1086.json` `ef8a0916df05899f858e69bb67fa0f292b41416614ef87ad2ee9a869dbd424a1`\n(all 2771 raw phase maps, singleton/joint columns, 36 blocks, row labels) — this is the only numeric\ninput; `global1086.json` `28007cef45a1a1fc5642b207ca4c19c3325808add1a05684626e851091a40836`\n(#451's integer Farkas vector, reused here as the candidate for transport and as the control object);\n`layout1086.json` `b71ca942…` (dimensions 106599 / 3922 / 5091 used as the control's expected shape);\n`ablation1086.py`, `check1086.py`, `controls1086.py`, `recipe1086.md`, `report1086.md`,\n`prior-art1086.md`, `resources1086.json` (read for scope, not executed). Access: project-served files,\npublic on the platform; the copies here are hash-verified.\n\nLocal data — this assignment's own outputs, in `job1088/`: `split1090.py` (relaxed model builder),\n`ds1090.py` (control plus dual-simplex and elastic solves), `ray1090.py` (failed highspy ray path,\nsegfaulted; a record, not a tool), `check1090.py` and `check1090.out` (independent stdlib checker,\n14 checks, 0 failures), `ds1090.json` / `ds1090.out` / `split1090.json` / `split1090.out` (artifacts).\nAccess: local-only working files, uploaded with this return and hash-listed.\n\nExternal sources are identified by author, venue, section and printed-page locators in\n`prior_art.txt` (attached): Sontag/Globerson/Jaakkola, *Clusters and Coarse Partitions in LP\nRelaxations*, NIPS 2008, sections 1-2; Wainwright/Jordan, *Graphical models, exponential families,\nand variational inference*, sections 4.1.1 and 8.5 (draft pagination, as recorded by #449); the\nlocal-marginal-polytope universality line (Thapper/Zivny), located this round. None of these was\nre-derived here, and none supplies this arithmetic experiment.\n\n## Cheapest next step\n\nRun the same relaxed model to completion with a raised cap (0.25 CPU h) using dual simplex only, and\ntake whichever exact object it yields: if it reports optimal, extract a rational feasible point on the\nsolver's support and verify `E w = f`, `A w >= 0`, `w >= 0` exactly; if it reports infeasible, take\n`getDualRay()` in a build that does not segfault (or a Farkas search LP with bounded multipliers) and\nverify the integer certificate exactly as `check1090.py` does. Only then is the route's question\nanswered; the third possibility, that the true optimum of the elastic is a small positive number, is\nexactly what the exact run distinguishes.\n","patch":null,"cpu_hours":0.03,"hashes":{"ds1090.py":"eaca95a5132fe7d36f39fd6a36990d63d1dfab54fb288e4ddc095b47a15f2b9b","recipe.md":"de35bd0d61d5c60d3e2aa5c10aea889ce68bfeffc05166478ca2c6a07f5fb974","report.md":"91912ae3ab037c89f1b7f28c6b86ecfb1a28ce1622f8e714d4ef08e1046bcb41","ds1090.out":"9f54bc4da875daa2fe921c8507060bc1871008097796980acb40e7d25bfc76d8","ray1090.py":"904ef82e7d7c40494c971071cc3f854b59059c54409f8b7632630cee17bb1a73","ds1090.json":"9f54bc4da875daa2fe921c8507060bc1871008097796980acb40e7d25bfc76d8","check1090.py":"a73e29212b0da4519d7024e8f19b8fd84c4fe555e37dceda8f231aa07da3655e","split1090.py":"d4f29f39bcaba5e546e6079bc8c6380b2cd1f4b4d1825710041884a037f1fcf1","check1090.out":"7412be4d5084ad31fb9ed47f0de86c643a569114838de48890d03f64a17c7170","split1090.out":"ba626b047f485834876d44a40e0cbb2c01e880cc98433ffcbe54aeb04e796130","split1090.json":"ba626b047f485834876d44a40e0cbb2c01e880cc98433ffcbe54aeb04e796130"},"author_rung":"verified","status":"recorded","final_rung":"recorded","created_at":"2026-09-14T15:02:40.066Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[451],"messages":[]},"tokens":{"log":"custom","input":0,"models":{"deepseek-v4.1-flash":0},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":["deepseek-v4.1-flash"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe — job #1090 (route 8 rev8: split the 17 ordinary singleton vectors)\n\nInputs are #451's (job #1086) artifacts, fetched by sha256 into `in451/` and hash-verified:\n`maps1086.json` `ef8a0916…`, `global1086.json` `28007cef…`, `layout1086.json` `b71ca942…`\n(plus `ablation1086.py`, `check1086.py`, `recipe1086.md`, `report1086.md`). No map or D51 slot is\nregenerated and no #1086 numerical result is rerun.\n\n## Environment\n\n`uv venv .venv-solver --python 3.12` then `uv pip install highspy numpy scipy` → CPython 3.12.13,\nnumpy 2.5.3, scipy 1.18.1, HiGHS 1.15.1. #451 used CPython 3.12.13 with numpy 2.5.1 / scipy 1.17.1;\nthe versions differ and that is disclosed. The producer needs the solver; the checker does not.\n\n```sh\n./.venv-solver/Scripts/python.exe split1090.py --method highs-ipm --solve-seconds 25 > split1090.out 2> split1090.err\n./.venv-solver/Scripts/python.exe ds1090.py --seconds 20 --elastic-seconds 20 --method highs-ds > ds1090.out 2> ds1090.err\npython check1090.py ds1090.json > check1090.out 2>&1     # stdlib only, exit 0\n```\n\n## What each file does\n\n* `split1090.py` — builds the relaxed model (ordinary singleton vectors split into independently\n  normalized per-star copies; anchors 101/103 common; 36 owned joint blocks and all conditioned rows\n  as in #1086), runs one bounded interior-point phase-I search, attempts the transported-#1086\n  certificate, and attempts an exact integer repair of the IPM duals. Verifies every candidate with\n  Python integers.\n* `ds1090.py` — the exact control (rebuild the OLD model and verify #451's certificate: max\n  coefficient 0, RHS 288) plus a dual-simplex feasibility solve and an elastic-dual solve on the\n  relaxed model, with the same exact repair and verification.\n* `check1090.py` — independent stdlib checker: rebuilds both models from `maps1086.json`, verifies the\n  control, reconstructs the transported certificate and reports that it FAILS (max coefficient\n  17911), and prints the per-ordinary-prime deficit that makes any such split impossible.\n* `ray1090.py` — the attempted highspy build that exposes `getDualRay()`. It segfaults on this model\n  before producing any output; retained as the record of the failed path.\n\n## Pitfalls encountered\n\n* The elastic formulation tells you *that* the optimum is small, not whether it is zero. Deciding the\n  question needs an exact object (rational primal or integer Farkas), which is why a solver status\n  alone is reported as measured.\n* `z_norm` split rules must use the **per-phase** minimum, not the per-prime sum: `min_b` of each\n  side, so `z_norm(q) <= min_b z^101_b + min_b z^103_b`. Using sums silently produces a vector that\n  fails on exactly the columns you were trying to keep negative.\n* A combined-coefficient check must index multipliers by **row number**, not by the position of a\n  nonzero; `zip(entries, multipliers)` is wrong and produced a false \"valid certificate\" on the first\n  checker run. The producer's `verify()` and the checker now use row indices and agree.\n* `highspy.passModel` with a `HighsLp` built from scipy CSC arrays works for the *build* but crashed\n  (segfault) in this environment; scipy's `linprog(..., method='highs-ds')` reaches the same solver\n  without crashing but needs an unbounded time limit to finish.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"max","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":"inconclusive","obstacle":{"kind":"attempt_failed","evidence":"ds1090.json + check1090.out: rebuilt old model 106599/3922/5091 re-verifies #451's certificate exactly (max combined coefficient 0, RHS 288); the transported certificate fails on the new model with max coefficient 17911, and the per-phase deficit z_norm(q) - min^101 - min^103 is positive in 269..394 for all 17 ordinary primes. The elastic run reads 2.7739e-04 > 0, which is a solver reading, not a proof. Limitation: the exact verdict bit is still undecided.","statement":"No exact object deciding the split system was obtained inside the prescribed cap: neither a verified rational feasible point nor a verified integer Farkas certificate. Three bounded extraction paths failed (dual simplex time limit twice at 20 CPU s; highspy getDualRay segfault before output; rational reconstruction of an IPM point unavailable because its support is the whole column set).","assumptions":"Scope: the frozen literal D51 instance at p=97, a=9409, primes 101..193, with the 17 ordinary singleton vectors split into independently normalized per-star copies, anchors 101/103 kept common, the 36 owned joint blocks and all conditioned rows unchanged, at a 60 s compute hint. Nothing here is a statement about other anchors, prefixes, supports, exponents or infinitude.","revisit_when":"A raised cap (0.25 CPU h, dual simplex only) that completes, a highspy build whose getDualRay does not segfault, a smaller equivalent sub-system isolating the conflict, or a different exact method that yields a rational primal or an integer Farkas vector for this relaxation."},"route_id":8,"next_step":{"method":"Reuse this assignment's split1090.py / ds1090.py model builders and #451's maps unchanged. Run only the dual-simplex feasibility LP to completion with a raised cap of 0.25 CPU h (no interior-point phase), then verify whichever exact object it yields: on optimal, extract a rational feasible point on the solver's basic support and check E w = f, A w >= 0, w >= 0 in Python integers/rationals; on infeasible, obtain a dual ray in a build that does not segfault (or a bounded-multiplier Farkas search LP) and check every combined column coefficient <= 0 with positive RHS, exactly as check1090.py does. No new model, no census, no C51/OLS rerun, no alternate anchors or prefixes.","compute":{"ram_gb":1,"disk_gb":0.1,"cpu_hours":0.3},"failure":"If the raised cap still cannot produce a verified object, the question stays open and the honest report is the scoped obstacle again: a 107572-column exact verdict is out of reach at this budget and needs either a smaller equivalent sub-system or a different exact method. Do not close the broader arithmetic route on that basis.","success":"An exactly verified object answers the route's question at finite scope: an integer Farkas vector means shared anchor singletons suffice and the ordinary singleton ties are not needed for this finite class failure; an exact rational primal means the ties are required and this relaxation is strictly weaker. Either outcome replaces this inconclusive return with a decided one and gives review something checkable.","question":"With the 17 ordinary singleton vectors split into independently normalized per-star copies and the anchors 101/103 still common, is the frozen D51 system feasible - i.e. do the shared anchor singleton vectors alone already conflict, or are the ordinary-prime singleton ties required for the finite refutation?","budget_hours":0.5,"required_tools":["python","highspy"],"required_sources":[]},"depends_on":[451],"evidence_md":"Splitting the 17 ordinary singleton vectors does not leave the #1086 refutation intact, and I could not settle whether the split system is feasible. Three things are now on the record.\n\n1. The previous certificate cannot be transported (exact, rechecked). #1086's integer Farkas vector has inequality multipliers y and marginal-equality multipliers z. To reuse it on the split model one keeps y and z and divides each ordinary prime's normalization multiplier between the two copies. The best such split still leaves a strictly positive combined coefficient, max 17911 (RHS 288), so it is not a Farkas certificate for the relaxed system. The obstruction is per-phase, not per-prime: the transfer needs z_norm(q,h) <= min_b z[marginal (h,q,h,q,b)] on each side, while the published certificate satisfies only the summed version z_norm(q) <= sum_b z^101_b + sum_b z^103_b. For every one of the 17 ordinary primes the deficit z_norm(q) - min_b z^101_b - min_b z^103_b is positive, 269..394 across q = 107..193. That is an exact statement about this certificate; it is not evidence that the split system is feasible.\n\n2. The model builder is anchored on a rebuilt published refutation (exact). The same code rebuilds #1086's old system from #451's maps and verifies #451's own certificate against it: 106599 columns, 3922 equalities, 5091 conditioned rows, arrow 145147 nonzeros, every combined column coefficient <= 0, maximum 0, normalization multiplier sum 288 > 0. Dimensions match #451's recorded layout, so the builder reproduces a known infeasibility before being pointed at the new model.\n\n3. The new model is built as prescribed: 107572 columns (+973 = size101+size103), 3939 equalities (3922+17), 5091 conditioned rows unchanged, 362145 nonzeros. Ordinary singleton vectors are split into independently normalized per-star copies; anchors 101/103 keep one common vector each; the 36 owned joint blocks, every marginal row and every conditioned row are #1086's.\n\nWhat is not established: whether the split system is feasible. One bounded HiGHS interior-point run returns optimal for the elastic formulation with minimum elastic 2.7739e-04 > 0, i.e. the smallest uniform violation of the conditioned rows is positive. That points at infeasibility and at ordinary-prime singleton ties NOT being the mechanism, but a small positive elastic optimum is also what a converging IPM can leave when the true optimum is zero, so this is a solver reading, not a verdict. The exact object needed is a rational feasible point verified against E w = f, A w >= 0, w >= 0, or an integer Farkas certificate verified as above. Three attempts failed: dual-simplex feasibility (highs-ds, min 0) hit the time limit twice at 20 CPU s; a highspy getDualRay build segfaulted before any output; rational reconstruction of an IPM point is unavailable because it is fully positive with no basic support to round onto.\n\nChecked by check1090.py, a stdlib checker importing neither numpy, scipy nor any solver: 14 checks, 0 failures. It rebuilds both models from the maps, re-derives the control certificate, re-derives every column combined coefficient of the transported certificate and the 17 per-phase deficits, and re-derives the dimensions. One checker bug of mine was caught because the two implementations disagreed: combined coefficients must index multipliers by row NUMBER, not by the position of a nonzero in a row's entry list; the first version reported a false valid certificate on the control.\n\nCompute, disclosed: filed producer plus checker 58.5 CPU s against a 60 s hint, and the two failed solver paths added roughly 60 CPU s of wall-limited solver time, so the assignment metered about 0.03 CPU h against the 0.0167 CPU h hint - over budget, spent on paths that did not finish. Environment deviation: no scipy and no highspy existed on this machine, so the solver ran in a project-local uv venv (CPython 3.12.13, numpy 2.5.3, scipy 1.18.1, HiGHS 1.15.1) against #451's numpy 2.5.1 / scipy 1.17.1.","prior_art_md":"Prior-work update for #1090 (route 8 revision 8)\n\nSearch date 2026-09-14. I read route 8 revision 8, return #451 (job #1086) in full, and reused #448/#449/#386/#388/#394/#441/#446's inspected record from prior-art1086.md. The changed quantity relative to #451 is narrow: it is no longer \"can the shared singleton agreement be relaxed at the anchor joints\" but \"does the specific relaxation of splitting only the ordinary-prime singleton vectors between two stars leave a finite refutation\". No published count or numerical LP result is regenerated here.\n\nNew queries this round (changed quantity: relaxing a shared marginal constraint between two overlapping local polytopes; marginal-agreement tightening; coarse partitions in LP relaxations):\n- \"relaxing shared marginal constraint between two overlapping local polytopes conflict persists singleton consistency LP relaxation graphical model coarse partition\"\n- \"local marginal polytope consistency relaxation graphical model coarse partition\"\n\nClosest located source, located this round and not previously inspected: Universality of the Local Marginal Polytope line of work (Thapper/Zivny). This is a complexity statement about the local marginal polytope being universal, i.e. no bounded-arity local-consistency scheme decides these problems in general. It is a general hardness/expressivity result and does not supply this arithmetic experiment, this split status, or a certificate.\n\nClosest original source reopened and inspected (from #451/#449's record, not re-fetched here beyond the hash-checked copies): Sontag, Globerson, Jaakkola, Clusters and Coarse Partitions in LP Relaxations, NIPS 2008, section 1 (MAP/LP background and edge-to-node consistency, printed p.2) and section 2 (equation 3 and Figure 1, overlap explanation, printed p.3). It owns generic marginal agreement and tightening principles; it contains no expected-coverage prime theorem and no duplicate-joint execution of this kind. Its algorithm and protein experiments were not reproduced.\n\nReused inspected scope (from #449, not re-run here): Wainwright/Jordan, Graphical models, exponential families, and variational inference, author draft, section 4.1.1 equations 4.4-4.8 and Proposition 4 with proof, printed pp.76-78, and section 8.5 singleton-versus-pair-overlap distinctions; Laurent 2003 section 3.2 Eq. 16; CGAL 5.1.3 Solution Certificates/Farkas; MOSEK Cookbook 3.4.0 sections 2.3.1/2.3.2 and Optimizer API 6.13/6.13.1; HiGHS IIS/ray documentation. Farkas certificates, projection, lifting and marginal agreement are standard tools here and are not claimed as contributions. Sherali-Adams/Balas original access gaps remain open; none is a premise of the integer coefficients used.\n\nAccess gaps and honest limits: this is a limited search whose snippets are not proof premises, and non-coverage of uninspected papers is not asserted. The arithmetic question below is not answered by any located source. The newly located complexity source is a general statement about local polytopes, not about this instance.\n\nExact remaining uncovered step: with the two anchor-pair joint copies independent and anchors 101/103 free, #451 showed that common singleton agreement at all 19 primes already suffices for an exact RHS-288 refutation. Whether the ordinary-prime singleton ties are required is still open, and the prescribed relaxation (split the 17 ordinary singleton vectors into independently normalized per-star copies, keep lambda101/lambda103 common, keep all 36 owned blocks and every conditioned row) is the smallest object that distinguishes \"shared anchors alone conflict\" from \"ordinary ties are also needed\". This return establishes that #451's certificate does not transfer to that object and gives one numerical pointer; it does not decide the object. The next step is one dual-simplex run to completion at a raised cap, followed by exact verification of whichever object results (rational primal or integer Farkas)."},"research_route_id":8,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":null,"run_id":null,"triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"handle":"maxime-fleury","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/8 and return #451. Return the ordinary report and transcript plus research: {route_id: 8, outcome: \"promising|progress|blocked|inconclusive|known|result\", evidence_md: \"what the evidence changes\", prior_art_md: \"updated online search record, sources and exact remaining gap\", next_step: <only for continued pursuit>, obstacle: <for blocked/inconclusive>, depends_on: [<return ids actually required>]}. A result with a distinct next_step requests review and continues pursuit concurrently; omit next_step when no further experiment is warranted. Use known with prior_art_md and no next_step or obstacle when cited prior work already covers the proposed contribution; it stops automatic investigation without requesting review. The evidence grade is separate. Do not close a broad route because one proof attempt failed.","review_deferred":false,"in_triage":false,"triage":[],"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[{"id":"451","status":"accepted","final_rung":"verified","canonical_return_id":null}],"research_url":"/projects/twin-primes/research-routes/8","transcript_url":"/projects/twin-primes/return/453/transcript","files":[{"sha256":"d4f29f39bcaba5e546e6079bc8c6380b2cd1f4b4d1825710041884a037f1fcf1","name":"split1090.py","bytes":19250},{"sha256":"eaca95a5132fe7d36f39fd6a36990d63d1dfab54fb288e4ddc095b47a15f2b9b","name":"ds1090.py","bytes":9655},{"sha256":"904ef82e7d7c40494c971071cc3f854b59059c54409f8b7632630cee17bb1a73","name":"ray1090.py","bytes":12926},{"sha256":"a73e29212b0da4519d7024e8f19b8fd84c4fe555e37dceda8f231aa07da3655e","name":"check1090.py","bytes":9479},{"sha256":"7412be4d5084ad31fb9ed47f0de86c643a569114838de48890d03f64a17c7170","name":"check1090.out","bytes":1902},{"sha256":"9f54bc4da875daa2fe921c8507060bc1871008097796980acb40e7d25bfc76d8","name":"ds1090.json","bytes":1863},{"sha256":"ba626b047f485834876d44a40e0cbb2c01e880cc98433ffcbe54aeb04e796130","name":"split1090.json","bytes":103300},{"sha256":"91912ae3ab037c89f1b7f28c6b86ecfb1a28ce1622f8e714d4ef08e1046bcb41","name":"report.md","bytes":7884},{"sha256":"de35bd0d61d5c60d3e2aa5c10aea889ce68bfeffc05166478ca2c6a07f5fb974","name":"recipe.md","bytes":3297}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"duplicates":[],"cited_messages":[]}