{ "registry": "SIMARC claims registry", "version": "v1", "updated": "2026-08-21", "note": "Every claim carries the evidence standard that decided it. A claim may not be restated with vocabulary above its standard. Kill tests, where declared, are the specific computation whose outcome would falsify the claim.", "verification_invitation": { "independent_verifications_to_date": 0, "protocol": "https://simarc.org/agents.html", "record_schema": { "claim_id": "string", "outcome": "CONFIRMED | REFUTED | INCONCLUSIVE", "input_hashes": "object", "output_hashes": "object", "environment": "object", "notes": "string", "verifier": { "name": "string", "kind": "agent | human", "contact": "string" } } }, "claims": [ { "id": "P01-S26-BOUNDARY", "program": "P-01", "codename": "TWO SLITS", "statement": "Every support of at most 25 cells admits four mutually orthogonal partial Latin structures; at 26 cells exactly one orbit refuses.", "standard": "PROOF", "status": "CONFIRMED", "certificate": "145 MB DRAT proofs, replayed twice under a hash-pinned checker", "independent_checks": "Orbit counts 934 and 1,711 agree between direct canonical augmentation and an independent Burnside count; all witnesses re-verified.", "kill_test": "Replay the DRAT proofs under an independent checker; or exhibit a 26-cell support in the refusing orbit that admits four mutually orthogonal partial Latin structures.", "source_artifact": "P-01 program tree: certificates/", "verifications": [] }, { "id": "P01-ORDER6-INTEGRAL", "program": "P-01", "codename": "TWO SLITS", "statement": "The order-six obstruction is purely integral: over a complete census of all 9,408 reduced Latin squares of order six, every relaxation-based certificate (clique, Lovasz theta, fractional chromatic) equals 6 while the true chromatic number exceeds 6 in every case.", "standard": "PROOF", "status": "CONFIRMED", "certificate": "complete census, 9,408 / 9,408", "note": "The 108 relaxation-blind squares are pinned exactly by two classical hand-computable invariants.", "kill_test": "Exhibit any relaxation-based certificate that separates the obstruction, or any order-six reduced Latin square outside the census.", "source_artifact": "P-01 program tree", "verifications": [] }, { "id": "P01-A6-65-N28", "program": "P-01", "codename": "TWO SLITS", "statement": "A6(6,5) at N = 28 is decided completely: 324 of 324 cubes closed.", "standard": "EXACT", "status": "CONFIRMED", "certificate": "per-cube certificates on disk", "note": "The last seven cubes were closed by an orthogonality-capacity argument after both leading SAT engines consumed 43,200 s each without a verdict.", "kill_test": "Re-solve any cube in the family and obtain a different verdict.", "source_artifact": "P-01 program tree", "verifications": [] }, { "id": "P01-PARITY-THEOREM", "program": "P-01", "codename": "TWO SLITS", "statement": "For every pair of orthogonal 5x6 Latin rectangles, the completion-discrepancy permutation is odd.", "standard": "EXACT", "status": "OPEN", "note": "Verified by complete census; no human proof. The statement provably implies the Euler-Tarry theorem, so no proof of it can be easier than a proof that no pair of orthogonal Latin squares of order six exists. Twenty proof routes were proposed and kill-tested against a frozen dataset; all twenty failed, and the convergence identified the structural reason: because the theorem contains Euler-Tarry, the sign is constant on every object that exists, so any invariant agreeing with it does so vacuously.", "kill_test": "Exhibit a pair of orthogonal 5x6 Latin rectangles whose completion-discrepancy permutation is even.", "source_artifact": "P-01 program tree: 22 isotopy representatives, frozen kill-test dataset", "verifications": [] }, { "id": "P01-SEAM-LATTICE", "program": "P-01", "codename": "TWO SLITS", "statement": "No cube-aligned (l=4 cubic harmonic) power is detected in the quasar sky.", "standard": "STATISTICAL", "status": "NULL", "seam": "a grid in space", "evidence": "756,000 quasars", "kill_test": "Detect l=4 cubic-harmonic power above the stated bound in an independent quasar catalog.", "source_artifact": "P-01 program tree: lattice_cosmology/", "verifications": [] }, { "id": "P01-SEAM-CLOCK", "program": "P-01", "codename": "TWO SLITS", "statement": "No energy-dependent photon delay of the form v_g = c*cos(ka/2) is detected.", "standard": "STATISTICAL", "status": "NULL", "seam": "a grid in time", "evidence": "GRB 090510", "kill_test": "Detect energy-dependent arrival delay consistent with lattice dispersion in an independent gamma-ray burst.", "source_artifact": "P-01 program tree: clock_seam/", "verifications": [] }, { "id": "P01-SEAM-BOUNDARY", "program": "P-01", "codename": "TWO SLITS", "statement": "The CMB's missing large-angle power is a robust ~0.6%-level anomaly with no working physical explanation. Cut-sky S(1/2) exceeds 3 sigma, reproduced in four independent sky-cleanings and in both WMAP and Planck. Every finite-box explanation tested is now exhausted.", "standard": "STATISTICAL", "status": "ANOMALY", "seam": "memory - a finite box", "explanations_killed": [ "periodic box - killed by circle searches", "walled boxes - killed by the absence of a parity signature", "corner-observer geometry - the one geometry that does write an odd sky (~5x LambdaCDM) fails its own anti-mirror kill-shot on real data" ], "caveats": "The 'axis' is fragile; T-E polarization is inconclusive.", "kill_test": "Produce a finite-box or other physical model that reproduces the cut-sky S(1/2) deficit AND survives the circle, parity, and anti-mirror tests; or show the deficit is an artifact of the sky cut or cleaning.", "source_artifact": "P-01 program tree: boundary_seam/", "verifications": [] }, { "id": "P01-SEAM-RNG", "program": "P-01", "codename": "TWO SLITS", "statement": "No PRNG periodicity or structure is detected in quantum random bit streams.", "standard": "STATISTICAL", "status": "NULL", "seam": "a random number generator", "kill_test": "Detect period, correlation, or algorithmic structure in a quantum bit stream above the stated bound.", "source_artifact": "P-01 program tree: randomness_seam/", "verifications": [] }, { "id": "P01-SEAM-PRECISION", "program": "P-01", "codename": "TWO SLITS", "statement": "Reality's arithmetic exceeds float64: no round-off or irreversibility signature is detected.", "standard": "STATISTICAL", "status": "NULL", "seam": "arithmetic", "kill_test": "Detect a round-off or irreversibility signature at or above the stated precision bound.", "source_artifact": "P-01 program tree: precision_seam/", "verifications": [] }, { "id": "P01-SEAM-MEMORY", "program": "P-01", "codename": "TWO SLITS", "statement": "No breakdown of the exponential 2^N entanglement scaling is detected.", "standard": "STATISTICAL", "status": "NO_SIGNAL", "seam": "memory - entanglement", "note": "Live frontier. Quantum mechanics holds at every scale tested.", "kill_test": "Detect a departure from 2^N Hilbert-space scaling at any accessible N.", "source_artifact": "P-01 program tree: compute_budget/", "verifications": [] }, { "id": "P02-CHSH-CEILING", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "No quantum strategy in any Hilbert space of any dimension exceeds a CHSH value of 2*sqrt(2).", "standard": "PROOF", "status": "CONFIRMED", "certificate": "Sum-of-squares identity 2*sqrt(2)*I - S = (1/(2*sqrt(2)))[(A0+A1-sqrt(2)B0)^2 + (A0-A1-sqrt(2)B1)^2], verified symbolically with residual exactly zero.", "note": "A 4x4 symbolic check suffices because every irrep of Z2*Z2 has dimension at most 2. See-saw numbers show the bound is attained; the identity shows it cannot be beaten.", "kill_test": "Exhibit any strategy, in any dimension, exceeding 2*sqrt(2); or find a nonzero residual in the sum-of-squares identity.", "source_artifact": "P-02 program tree: lib/rigor.py, sims/02_substrates.py", "verifications": [] }, { "id": "P02-CHSH-FIELD-BLIND", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "CHSH cannot distinguish real quantum mechanics from complex quantum mechanics: the ceilings are identical, because the optimal strategy never needs the imaginary generator.", "standard": "PROOF", "status": "CONFIRMED", "note": "Reproduces from scratch the structural reason Renou et al. (Nature 600, 625) needed a network of independent sources rather than a single Bell test.", "kill_test": "Exhibit a single-source Bell inequality whose quantum ceiling differs between real and complex QM.", "source_artifact": "P-02 program tree: sims/02_substrates.py", "verifications": [] }, { "id": "P02-TOMOGRAPHY-GAP", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "The local-tomography gap of real quantum mechanics at local dimension d is exactly [d(d-1)/2]^2 - the tensor square of what the real field is missing. The gap is 0 for classical and for complex QM at all d.", "standard": "PROOF", "status": "CONFIRMED", "certificate": "Proved symbolically for every d, with exact integer rank computation kept as an independent check rather than as the evidence.", "checked_values": { "d=2": 1, "d=3": 9, "d=4": 36, "d=5": 100 }, "kill_test": "Compute the gap at any d and obtain a value other than [d(d-1)/2]^2.", "source_artifact": "P-02 program tree: lib/substrates.py, lib/rigor.py", "verifications": [] }, { "id": "P02-QUATERNIONIC-COMPOSITION", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "Quaternionic quantum mechanics fails at composition itself: two quaternionic systems do not form a quaternionic system.", "standard": "PROOF", "status": "CONFIRMED", "note": "Not a wrong prediction but a structural failure - a substrate that cannot build a joint system cannot run a universe containing two of anything. This was nearly reported as the number -8 before a guard caught that two ranks were being subtracted across different vector spaces.", "kill_test": "Exhibit a composition rule under which two quaternionic systems form a quaternionic system.", "source_artifact": "P-02 program tree: lib/substrates.py", "verifications": [] }, { "id": "P02-BOXWORLD-CEILING", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "The PR-box (box world) CHSH ceiling is exactly 4 with no-signalling verified; the classical ceiling is exactly 2 by complete enumeration of all 16 deterministic vertices.", "standard": "EXACT", "status": "CONFIRMED", "certificate": "complete enumeration, exact arithmetic", "kill_test": "Exhibit a no-signalling box exceeding 4, or a classical strategy exceeding 2.", "source_artifact": "P-02 program tree: sims/02_substrates.py", "verifications": [] }, { "id": "P02-MODALITY-RANKING", "program": "P-02", "codename": "QUANTUM COMPUTER", "statement": "The trapped ion leads every currency simultaneously; 'the ion leads' survives 98% of realistically perturbed parameter tables, but the full four-device ranking survives only 34%.", "standard": "STATISTICAL", "status": "CONFIRMED", "development_only": true, "note": "The pre-registered hypothesis that the winner would change with circuit shape was REFUTED by the sim itself and was not edited afterwards to match. Circuit shape moves the margin, not the winner: long-range routing costs the ion 1%, the transmon 24%, and the silicon spin qubit 51%. All device parameters are verified=false, and the preflight refuses this run if declared for publication.", "kill_test": "Re-run with an independently sourced, verified device parameter table and obtain a different winner.", "source_artifact": "P-02 program tree: sims/01_modalities.py, lib/devices.py", "verifications": [] }, { "id": "P03-PILOT-BASELINE", "program": "P-03", "codename": "CANCER", "statement": "In 528 TCGA-KIRC patients, 7,315 of 20,257 expressed genes (36% of the transcriptome) reach FDR < 0.05 against survival; 94% of top hits replicate in held-out patients; 12 survive adjustment for stage and age.", "standard": "OBSERVATIONAL", "evidence_rung": "OBSERVATIONAL", "status": "CONFIRMED", "licenses": "a hypothesis, and nothing more", "kill_test": "Re-run the screen on the same cohort with an independent pipeline and obtain a materially different baseline fraction.", "source_artifact": "P-03 program tree: src/ (2026-08-20 pilot)", "verifications": [] }, { "id": "P03-RISK-SCORE-USELESS", "program": "P-03", "codename": "CANCER", "statement": "The 200-gene risk score derived from that screen reaches C = 0.594, statistically indistinguishable from patient age alone, and significantly worse than stage and age together (C = 0.745, p = 0.007).", "standard": "OBSERVATIONAL", "evidence_rung": "OBSERVATIONAL", "status": "CONFIRMED", "note": "The negative result is the point: a true result and a useful result are almost unrelated in this field.", "kill_test": "Produce a transcriptome-derived score on this cohort that significantly beats stage + age out of sample.", "source_artifact": "P-03 program tree", "verifications": [] }, { "id": "P03-MITOTIC-INDEX", "program": "P-03", "codename": "CANCER", "statement": "A three-gene mitotic index carries HR = 1.60 per SD, held-out and stage-adjusted - real, replicable, and changes no decision anyone makes, because the subgroup where it would matter had 16% power.", "standard": "OBSERVATIONAL", "evidence_rung": "OBSERVATIONAL", "status": "CONFIRMED", "clinical_relevance": "none - the decision gate returns 'biology-only'", "kill_test": "Fail to replicate HR >= 1.60 per SD in an independent stage-adjusted renal cohort with adequate events (Schoenfeld: 143 deaths).", "source_artifact": "P-03 program tree", "verifications": [] }, { "id": "P03-T1-REFUTED", "program": "P-03", "codename": "CANCER", "statement": "Robust undrugged selective dependencies are NOT under-studied: they are roughly twice as well studied as random genes.", "standard": "OBSERVATIONAL", "evidence_rung": "OBSERVATIONAL", "status": "REFUTED_OWN_PREMISE", "note": "T1 was the first target run through the full gate chain. Its premise was refuted by its own baseline gate before it consumed compute.", "kill_test": "Re-run the prior-art tiering against an independent literature registry and obtain the opposite ordering.", "source_artifact": "P-03 program tree: sims/02_undrugged_dependencies.py, sims/05_t1_verdict.py", "verifications": [] }, { "id": "P04-GRAVEYARD", "program": "P-04", "codename": "DIMENTAI", "statement": "Of 8,224 interventional dementia trials, 731 are terminated, withdrawn or suspended, and only 59 of those (8.1%) carry any evidence about anything. The largest single class is failure to recruit (164); COVID alone accounts for 67; 78 state no reason at all. Restricted to trials naming a dementia syndrome, 32 carry a primary result about a mechanism and 17 more only appear to, because they stopped when a sibling trial failed.", "standard": "OBSERVATIONAL", "evidence_rung": "CLINICAL", "status": "CONFIRMED", "certificate": "classifier passes 44 controls in both directions", "note": "The obvious trial query returns 5,242; the field-vocabulary query returns 8,224.", "kill_test": "Re-pull ClinicalTrials.gov with the published query and classify the dead trials independently; a materially different evidence-bearing fraction refutes the classification.", "source_artifact": "P-04 program tree: the trial termination classifier", "verifications": [] }, { "id": "P04-GENETIC-RUNG", "program": "P-04", "codename": "DIMENTAI", "statement": "In 21,101,114 variants (Bellenguez 2022, n=487,511), APOE reaches p = 5.75e-659, an arbitrary control window is null, and 0.027% of all variants reach genome-wide significance - against 36% of the transcriptome clearing FDR in the CANCER pilot.", "standard": "OBSERVATIONAL", "evidence_rung": "GENETIC", "status": "CONFIRMED", "licenses": "a causal direction, never a target", "certificate": "md5-verified against the EBI published checksum on every run; positive and negative controls run before use", "note": "That ratio is the argument for where this program's causal rung lives. Genetic causality is not pharmacological tractability: APP and PSEN1 cause Alzheimer's, and the therapeutic programme that licensed took three decades to produce 0.45 CDR-SB points.", "kill_test": "Re-download the GWAS, verify the md5, and recompute the genome-wide significant fraction.", "source_artifact": "P-04 program tree: lib/fetch_gwas.py", "verifications": [] }, { "id": "P04-CLARITY-MCID", "program": "P-04", "codename": "DIMENTAI", "statement": "CLARITY-AD passes the statistical power gate and fails the MCID gate: it was powered to detect an effect patients would notice, and it found 45% of one.", "standard": "OBSERVATIONAL", "evidence_rung": "CLINICAL", "status": "CONFIRMED", "note": "Both facts are true; only the first is usually reported. This is why the seal gate runs two checks rather than one.", "kill_test": "Recompute the trial's detectable effect size and the field's MCID and show the former is smaller.", "source_artifact": "P-04 program tree: the seal gate", "verifications": [] } ] }