{
  "version": "2026-09-11.1",
  "reviewedAt": "2026-09-11",
  "reviewer": "AI-assisted editorial research; no independent expert sign-off",
  "coverage": "Curated initial sample of 14 records across 6 organizations. This is not an exhaustive census or a systematic literature review. Missing organizations and missing claims are not evidence of no discoveries. Rankings describe only this sample.",
  "claims": [
    {
      "id": "alphaevolve-matrix", "title": "A smaller algorithm for complex matrix multiplication", "organization": "Google DeepMind", "model": "AlphaEvolve / Gemini", "domain": "Mathematics", "date": "2025-05-14", "scope": "discovery", "novelty": 2, "confidence": "Medium", "evidence": "Public algorithm",
      "dateBasis": "Dated public announcement. Internal discovery time is not independently established.",
      "claim": "A bilinear algorithm multiplies two 4 × 4 complex matrices with 48 scalar multiplications.",
      "assessment": "A new construction beyond the documented 49-multiplication baseline. The numerical gain is small, but the construction improves a long-standing bound in this specific setting. This does not establish a new matrix-multiplication exponent or a universal speedup.",
      "prior": {"title": "Strassen's recursive construction", "date": "1969", "summary": "Applying Strassen's 2 × 2 construction twice gives 49 multiplications for 4 × 4 matrices. AlphaTensor's 47-multiplication binary-field result is not the same setting.", "relation": "Direct task-matched baseline, as identified by the claimant", "url": "https://doi.org/10.1007/BF02165411"},
      "metric": {"label": "Scalar multiplications", "before": 49, "after": 48, "unit": "multiplications", "comparable": true, "note": "(49 − 48) / 49 = 2.04% fewer scalar multiplications. Counts exclude additions and hardware effects; this is not a measured runtime improvement."},
      "limitations": "Provisional assessment from the announcement and disclosed result. We have not independently run symbolic verification or exhaustively searched all bilinear constructions.",
      "sources": [{"title": "AlphaEvolve announcement and technical report", "url": "https://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/", "date": "2025-05-14", "role": "Claim"}, {"title": "Strassen — Gaussian elimination is not optimal", "url": "https://doi.org/10.1007/BF02165411", "date": "1969", "role": "Prior art"}, {"title": "AlphaTensor — distinct binary-field setting", "url": "https://www.nature.com/articles/s41586-022-05172-4", "date": "2022-10-05", "role": "Prior art"}],
      "queries": ["AlphaEvolve matrix multiplication 48 49 Strassen", "AlphaTensor 4x4 binary arithmetic versus complex matrices"]
    },
    {
      "id": "alphatensor-binary", "title": "A 47-multiplication construction over the binary field", "organization": "Google DeepMind", "model": "AlphaTensor", "domain": "Computer science", "date": "2022-10-05", "scope": "discovery", "novelty": 2, "confidence": "Medium", "evidence": "Peer-reviewed / algorithm",
      "dateBasis": "Online publication date of the Nature article.",
      "claim": "AlphaTensor finds a 4 × 4 matrix-multiplication algorithm requiring 47 scalar multiplications in arithmetic modulo two.",
      "assessment": "A new construction for the specified field. It advances the documented baseline within an established tensor-decomposition framework; its novelty should not be generalized to ordinary real or complex arithmetic.",
      "prior": {"title": "Strassen's recursive construction", "date": "1969", "summary": "Recursive use of the seven-product algorithm requires 49 scalar multiplications. AlphaTensor's improvement here relies on binary-field arithmetic.", "relation": "Direct task-matched baseline, as identified in the paper", "url": "https://doi.org/10.1007/BF02165411"},
      "metric": {"label": "Scalar multiplications over F₂", "before": 49, "after": 47, "unit": "multiplications", "comparable": true, "note": "(49 − 47) / 49 = 4.08% fewer multiplications in this field. This does not measure general-purpose matrix multiplication speed."},
      "limitations": "We have not rerun the released constructions. Subsequent algorithms do not change the novelty assessment at the 2022 cutoff.",
      "sources": [{"title": "Discovering faster matrix multiplication algorithms with reinforcement learning", "url": "https://www.nature.com/articles/s41586-022-05172-4", "date": "2022-10-05", "role": "Claim"}, {"title": "Strassen — Gaussian elimination is not optimal", "url": "https://doi.org/10.1007/BF02165411", "date": "1969", "role": "Prior art"}],
      "queries": ["Discovering faster matrix multiplication algorithms reinforcement learning 2022", "AlphaTensor 47 49 binary field Strassen"]
    },
    {
      "id": "funsearch-cap-sets", "title": "Larger cap-set constructions", "organization": "Google DeepMind", "model": "FunSearch / Codey", "domain": "Mathematics", "date": "2023-12-14", "scope": "discovery", "novelty": 2, "confidence": "Medium", "evidence": "Peer-reviewed / construction",
      "dateBasis": "Online publication date, preceding the January 2024 print issue.",
      "claim": "FunSearch produces improved cap-set constructions, including a set of size 512 in dimension eight.",
      "assessment": "The paper documents new constructions for a well-studied extremal problem. This is new mathematical output within an existing problem and search framework, rather than a new field or a solution of the general cap-set problem.",
      "prior": {"title": "Edel and Bierbrauer; generalized product caps", "date": "2004", "summary": "Earlier finite cap constructions and product-cap methods provide the comparison framework in the FunSearch paper. A full reconstruction of the record history remains outstanding.", "relation": "Relevant construction family; closest numerical predecessor not independently reconstructed", "url": "https://www.nature.com/articles/s41586-023-06924-6#ref-CR34"},
      "metric": {"label": "Reported cap-set size in dimension 8", "before": null, "after": 512, "unit": "elements", "comparable": false, "note": "The reported construction has 512 elements. No percentage delta is shown because this review has not independently established the exact previous record from the original literature."},
      "limitations": "Prior-art coverage is partial. We rely on the paper's comparison and have not reproduced its programs or its asymptotic construction.",
      "sources": [{"title": "Mathematical discoveries from program search with large language models", "url": "https://www.nature.com/articles/s41586-023-06924-6", "date": "2023-12-14", "role": "Claim"}, {"title": "Edel (2004), Extensions of generalized product caps — citation in the claim paper", "url": "https://www.nature.com/articles/s41586-023-06924-6#ref-CR34", "date": "2004", "role": "Prior art"}],
      "queries": ["FunSearch cap set 512 prior art", "Edel Bierbrauer Large caps in small spaces generalized product caps"]
    },
    {
      "id": "gnome-crystals", "title": "A large catalogue of predicted stable crystals", "organization": "Google DeepMind", "model": "GNoME", "domain": "Materials science", "date": "2023-11-29", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Peer-reviewed / computational",
      "dateBasis": "Nature online publication and public announcement. The paper's baseline databases use March 2021 snapshots, not the full public record at announcement.",
      "claim": "GNoME reports 381,000 new entries on an updated computational convex hull, within a larger set of predicted crystal structures.",
      "assessment": "A substantial catalogue expansion is documented. A blanket novelty rating would conceal differences between new compositions, new structures, and variants of known prototypes. Entry-level matching against announcement-date databases is required.",
      "prior": {"title": "Materials Project and OQMD catalogues", "date": "2021-03", "summary": "The study starts from March 2021 database snapshots. These are useful baselines, but leave a gap before the November 2023 public claim.", "relation": "Claimant's dataset baseline; announcement-date completeness unresolved", "url": "https://doi.org/10.1063/1.4812323"},
      "metric": {"label": "Reported new hull entries", "before": null, "after": 381000, "unit": "predicted entries", "comparable": false, "note": "A publisher-reported candidate count, not 381,000 independently validated discoveries or a distance from prior art. This catalogue counts as one claim record."},
      "limitations": "No structure-by-structure deduplication, patent search, or synthesis validation was performed here. Computational stability is distinct from experimental synthesizability.",
      "sources": [{"title": "Scaling deep learning for materials discovery", "url": "https://www.nature.com/articles/s41586-023-06735-9", "date": "2023-11-29", "role": "Claim"}, {"title": "The Materials Project: a materials genome approach", "url": "https://doi.org/10.1063/1.4812323", "date": "2013", "role": "Prior art"}, {"title": "Public materials discovery dataset", "url": "https://github.com/google-deepmind/materials_discovery", "date": "2023", "role": "Supporting artifact"}],
      "queries": ["GNoME materials discovery March 2021 snapshots novelty", "GNoME Materials Project OQMD novel prototypes"]
    },
    {
      "id": "alphafold-human-proteome", "title": "Predicted structures across the human proteome", "organization": "Google DeepMind", "model": "AlphaFold 2", "domain": "Biology", "date": "2021-07-22", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Peer-reviewed / predictions",
      "dateBasis": "Online publication date of the human-proteome study. This record concerns its prediction catalogue, not the earlier CASP14 method announcement.",
      "claim": "AlphaFold expands structural coverage across the human proteome, reporting confident predictions for 58% of residues.",
      "assessment": "Expanded predictive coverage can enable discoveries, but does not establish that every prediction is a new biological finding. Each structure needs its own comparison to experimental structures, templates, and earlier predictions.",
      "prior": {"title": "Experimental structures and earlier AlphaFold predictions", "date": "2020", "summary": "The study reports experimental coverage of 17% of human-protein residues. Earlier structure-prediction methods already existed, including the CASP13 AlphaFold system.", "relation": "Coverage and method context; no per-protein nearest-neighbor audit", "url": "https://www.nature.com/articles/s41586-019-1923-7"},
      "metric": {"label": "Confident predicted residue coverage", "before": null, "after": 58, "unit": "% of residues", "comparable": false, "note": "17% experimental coverage and 58% confident predicted coverage are different evidence categories. Their difference is not a validated discovery count or accuracy gain."},
      "limitations": "This record aggregates a catalogue and receives no blanket novelty score. Predictions and experimental observations are explicitly distinguished.",
      "sources": [{"title": "Highly accurate protein structure prediction for the human proteome", "url": "https://www.nature.com/articles/s41586-021-03828-1", "date": "2021-07-22", "role": "Claim"}, {"title": "Improved protein structure prediction using potentials from deep learning", "url": "https://www.nature.com/articles/s41586-019-1923-7", "date": "2020-01-15", "role": "Prior art"}],
      "queries": ["AlphaFold human proteome 2021 17 58 coverage", "AlphaFold CASP13 structure prediction 2020"]
    },
    {
      "id": "tokamak-control", "title": "Learning to control tokamak plasma shapes", "organization": "Google DeepMind", "model": "Deep reinforcement learning", "domain": "Physics", "date": "2022-02-16", "scope": "method", "novelty": null, "confidence": "Medium", "evidence": "Peer-reviewed / experiment",
      "dateBasis": "Nature online publication date; the author manuscript carries an earlier preparation date, not a verified public release date.",
      "claim": "A learned controller produces and maintains multiple plasma configurations on EPFL's TCV tokamak.",
      "assessment": "An experimentally demonstrated control-method advance. It is tracked as enabling research, not counted as a new physical law or a separate discovery for each plasma configuration.",
      "prior": {"title": "Conventional TCV magnetic feedback control", "date": "2021", "summary": "Model-based control and the study of shaped plasmas predate this work. The Nature paper describes the conventional controller architecture it replaces and cites the preceding control literature.", "relation": "Method baseline documented in the claim paper; original references not fully audited", "url": "https://www.nature.com/articles/s41586-021-04301-9#Sec1"},
      "metric": null,
      "limitations": "Research-method context record; excluded from discovery rankings. A separate controller-novelty review would require comparison with earlier control architectures.",
      "sources": [{"title": "Magnetic control of tokamak plasmas through deep reinforcement learning", "url": "https://www.nature.com/articles/s41586-021-04301-9", "date": "2022-02-16", "role": "Claim"}],
      "queries": ["Magnetic control tokamak plasmas deep reinforcement learning 2022 prior control"]
    },
    {
      "id": "microsoft-electrolyte", "title": "A mixed sodium–lithium solid electrolyte", "organization": "Microsoft", "model": "Azure Quantum Elements / materials AI", "domain": "Materials science", "date": "2024-01-08", "scope": "discovery", "novelty": 1, "confidence": "Low", "evidence": "Preprint / experiment",
      "dateBasis": "arXiv v1 submission date, one day before the Microsoft announcement. Partners include Pacific Northwest National Laboratory.",
      "claim": "AI-assisted screening leads to a mixed sodium–lithium solid electrolyte, followed by experimental validation with PNNL.",
      "assessment": "Provisionally an incremental compositional extension in an established family of solid electrolytes. Screening scale and development speed measure the discovery process, not the material's novelty or superiority.",
      "prior": {"title": "Asano et al., solid halide electrolytes", "date": "2018-09-14", "summary": "High-conductivity halide solid electrolytes for solid-state batteries were already demonstrated. This establishes family-level precedent, not an exact structural match to the new composition.", "relation": "Related material family; exact nearest composition remains to be audited", "url": "https://doi.org/10.1002/adma.201803075"},
      "metric": null,
      "limitations": "Low-confidence rating. Exact composition matching, patents, and a standardized performance comparison could change this classification. No percentage novelty or performance gain is assigned.",
      "sources": [{"title": "Accelerating computational materials discovery — arXiv v1", "url": "https://arxiv.org/abs/2401.04070v1", "date": "2024-01-08", "role": "Claim"}, {"title": "Microsoft / PNNL discovery announcement", "url": "https://azure.microsoft.com/en-us/blog/quantum/2024/01/09/unlocking-a-new-era-for-scientific-discovery-with-ai-how-microsofts-ai-screened-over-32-million-candidates-to-find-a-better-battery/", "date": "2024-01-09", "role": "Later context"}, {"title": "Solid halide electrolytes with high lithium-ion conductivity", "url": "https://doi.org/10.1002/adma.201803075", "date": "2018-09-14", "role": "Prior art"}],
      "queries": ["Microsoft PNNL lithium sodium solid electrolyte 2024 prior art", "Asano solid halide electrolytes high conductivity 2018"]
    },
    {
      "id": "rentosertib", "title": "An AI-designed antifibrotic drug candidate", "organization": "Insilico Medicine", "model": "PandaOmics / Chemistry42", "domain": "Medicine", "date": "2021-02-24", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Later peer-reviewed clinical data",
      "dateBasis": "Earliest announcement date identified, corroborated retrospectively by Insilico's November 2021 account. The target was not publicly named in the original announcement.",
      "claim": "Insilico reports an AI-selected antifibrotic target and AI-designed candidate, later disclosed as the TNIK inhibitor rentosertib.",
      "assessment": "TNIK inhibition was known before the claim. The possible novelty lies in the particular compound and its antifibrotic use. Neither target novelty nor chemical novelty can be resolved from the early undisclosed-target announcement alone.",
      "prior": {"title": "NCB-0846 and earlier TNIK inhibition", "date": "2016-08-26", "summary": "Masuda and colleagues published a small-molecule TNIK inhibitor in cancer research in 2016. A 2018 study also examined TNIK inhibition in an epithelial–mesenchymal transition context.", "relation": "Same target and related biology; full compound and indication priority unresolved", "url": "https://pubmed.ncbi.nlm.nih.gov/27562646/"},
      "metric": null,
      "limitations": "A dated patent and chemical-structure review is outstanding. Later clinical studies strengthen evidence for the candidate but cannot retroactively establish novelty at the 2021 cutoff. Clinical efficacy is not rated here.",
      "sources": [{"title": "Insilico's first-in-human account, documenting the February announcement", "url": "https://insilico.com/blog/fih", "date": "2021-11-30", "role": "Date provenance"}, {"title": "TNIK inhibition abrogates colorectal cancer stemness", "url": "https://pubmed.ncbi.nlm.nih.gov/27562646/", "date": "2016-08-26", "role": "Prior art"}, {"title": "NCB-0846 and epithelial–mesenchymal transition — research abstract", "url": "https://aacrjournals.org/cancerres/article/78/13_Supplement/2034/626440/Abstract-2034-Traf2-and-Nck-interacting-kinase", "date": "2018", "role": "Prior art"}, {"title": "TNIK inhibitor in preclinical and clinical models", "url": "https://www.nature.com/articles/s41587-024-02143-0", "date": "2024", "role": "Later evidence"}, {"title": "Rentosertib randomized phase 2a trial", "url": "https://www.nature.com/articles/s41591-025-03743-2", "date": "2025-06-03", "role": "Later evidence"}],
      "queries": ["Insilico February 24 2021 novel antifibrotic target", "TNIK inhibitor NCB-0846 2016 fibrosis prior art"]
    },
    {
      "id": "gentrl-ddr1", "title": "Generated inhibitors of the known DDR1 target", "organization": "Insilico Medicine", "model": "GENTRL", "domain": "Chemistry", "date": "2019-09-02", "scope": "discovery", "novelty": 1, "confidence": "Medium", "evidence": "Peer-reviewed / experiment",
      "dateBasis": "Online publication date of the Nature Biotechnology article.",
      "claim": "GENTRL generates candidate DDR1 inhibitors, with synthesized compounds tested experimentally.",
      "assessment": "An incremental medicinal-chemistry result on an established target. Existing selective DDR1 inhibitors and related scaffolds predate the work. Similarity is evidence of close prior art, not proof of copying or an identical molecule.",
      "prior": {"title": "Selective DDR1 inhibitors and prior kinase scaffolds", "date": "2013-04-10", "summary": "Gao and colleagues had reported selective, orally bioavailable DDR1 inhibitors in 2013. Bender's contemporaneous analysis found close chemical neighbors for a generated compound in ChEMBL.", "relation": "Same target; related chemical structures identified in an external analysis", "url": "https://pubmed.ncbi.nlm.nih.gov/23521020/"},
      "metric": null,
      "limitations": "We have not computed molecular fingerprints or run a patent search. Bender's 75% search threshold is not a measured universal novelty score and is not reported as one. Assay results from different studies are not directly compared.",
      "sources": [{"title": "Deep learning enables rapid identification of potent DDR1 kinase inhibitors", "url": "https://www.nature.com/articles/s41587-019-0224-x", "date": "2019-09-02", "role": "Claim"}, {"title": "Gao et al., selective and orally bioavailable DDR1 inhibitors", "url": "https://pubmed.ncbi.nlm.nih.gov/23521020/", "date": "2013-04-10", "role": "Prior art"}, {"title": "Andreas Bender's original chemical-similarity analysis", "url": "https://www.drugdiscovery.net/2019/09/03/so-did-ai-just-discover-its-first-drug-comment-on-deep-learning-enables-rapid-identification-of-potent-ddr1-kinase-inhibitors/", "date": "2019-09-03", "role": "Later analysis"}],
      "queries": ["GENTRL DDR1 novelty Bender 2019", "selective DDR1 inhibitors 2013 Gao"]
    },
    {
      "id": "sakana-regularization", "title": "A generated study of compositional regularization", "organization": "Sakana AI", "model": "The AI Scientist-v2", "domain": "Computer science", "date": "2025-03-12", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Workshop reviews / withdrawn",
      "dateBasis": "Public announcement date. The technical report was released later.",
      "claim": "An AI-generated manuscript reports negative results for compositional regularization and receives workshop review scores above the stated acceptance threshold.",
      "assessment": "Review scores assess a manuscript, not its distance from prior art. The specific negative result needs comparison with earlier regularization and compositional-generalization experiments before a novelty rating can be assigned.",
      "prior": {"title": "Earlier compositional-generalization and regularization research", "date": null, "summary": "The closest experiment has not been identified in this initial review. The earlier AI Scientist framework is process context and does not establish priority for this scientific result.", "relation": "Closest result unresolved", "url": "https://arxiv.org/abs/2408.06292v1"},
      "metric": null,
      "limitations": "The paper was withdrawn under the experiment protocol; the workshop did not perform a final meta-review. This record does not describe an accepted main-conference publication.",
      "sources": [{"title": "The AI Scientist's first peer-review experiment", "url": "https://sakana.ai/ai-scientist-first-publication/", "date": "2025-03-12", "role": "Claim"}, {"title": "The AI Scientist-v1 — earlier workflow", "url": "https://arxiv.org/abs/2408.06292v1", "date": "2024-08-12", "role": "Prior method context"}, {"title": "The AI Scientist-v2 technical report", "url": "https://arxiv.org/abs/2504.08066", "date": "2025-04-10", "role": "Later evidence"}],
      "queries": ["Sakana AI Scientist first peer reviewed publication 2025", "AI Scientist compositional regularization unexpected obstacles novelty"]
    },
    {
      "id": "claude-riemann", "title": "A claimed improvement in the zeta-zero proportion", "organization": "Anthropic", "model": "Unreleased research Claude", "domain": "Mathematics", "date": "2026-08-10", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Claimant report / proof artifacts",
      "dateBasis": "Dated announcement; the page notes an August 13 update to its paper. This review consulted the current page, not an archived first version.",
      "claim": "Anthropic reports raising a lower bound for the proportion of Riemann-zeta zeros on the critical line from 41.6% to 67.2%.",
      "assessment": "Potentially a new result combining existing mathematical tools. The cited numerical baseline, theorem assumptions, and full proof require specialist review before this leaderboard assigns novelty credit.",
      "prior": {"title": "Unconditional pair-correlation results", "date": "2023-06-08", "summary": "Baluyot, Goldston, Suriajaya, and Turnage-Butterbaugh developed an unconditional Montgomery theorem. The announcement also credits Aryan and Bombieri. These are stated ingredients, not a verified latest numerical record.", "relation": "Named mathematical ingredients; closest theorem and baseline audit pending", "url": "https://arxiv.org/abs/2306.04799v1"},
      "metric": {"label": "Publisher-reported lower-bound proportion", "before": 41.6, "after": 67.2, "unit": "%", "comparable": false, "note": "The announcement implies +25.6 percentage points. The baseline and theorem equivalence have not been independently established here; this comparison is not counted as a verified improvement."},
      "limitations": "We have not checked the Lean artifact, its assumptions, or the complete prior-art literature. This is not a claimed proof of the Riemann hypothesis itself.",
      "sources": [{"title": "Claude's progress on the Riemann hypothesis", "url": "https://www.anthropic.com/research/riemann-zeta", "date": "2026-08-10", "role": "Claim"}, {"title": "An unconditional Montgomery theorem for pair correlation", "url": "https://arxiv.org/abs/2306.04799v1", "date": "2023-06-08", "role": "Prior art"}],
      "queries": ["Anthropic Riemann zeta 2026 lower bound prior art", "unconditional Montgomery theorem pair correlation zeros 2023"]
    },
    {
      "id": "claude-fermat", "title": "A formalization of Fermat's Last Theorem", "organization": "Anthropic", "model": "Claude / Prove2Me", "domain": "Mathematics", "date": "2026-09-04", "scope": "method", "novelty": 0, "confidence": "High", "evidence": "Claimant report / formalization",
      "dateBasis": "Dated public announcement; the article describes earlier internal completion.",
      "claim": "Anthropic reports a complete Lean formalization of Fermat's Last Theorem, following existing mathematical proofs.",
      "assessment": "The theorem is a known result. The potentially new contribution is its formal verification artifact and the automation process. N0 applies only to discovery of the theorem, not to the value or novelty of formalization.",
      "prior": {"title": "Wiles; Darmon, Diamond, and Taylor", "date": "1995", "summary": "Wiles proved the theorem in work published in 1995. Anthropic explicitly says its formalization follows the later exposition by Darmon, Diamond, and Taylor.", "relation": "Same theorem and acknowledged proof lineage", "url": "https://www.math.mcgill.ca/darmon/pub/Articles/Expository/05.DDT/paper.pdf"},
      "metric": null,
      "limitations": "Method-context record, excluded from discovery rankings. This is acknowledged reuse, not a plagiarism allegation. The formalization's validity and method novelty have not been independently audited here.",
      "sources": [{"title": "Formalizing Fermat's Last Theorem", "url": "https://www.anthropic.com/research/formalizing-fermats-last-theorem", "date": "2026-09-04", "role": "Claim"}, {"title": "Darmon, Diamond and Taylor — Fermat's Last Theorem", "url": "https://www.math.mcgill.ca/darmon/pub/Articles/Expository/05.DDT/paper.pdf", "date": "1995", "role": "Prior art"}],
      "queries": ["Anthropic formalizing Fermat Last Theorem 2026", "Darmon Diamond Taylor Fermat Last Theorem proof"]
    },
    {
      "id": "openai-unit-distance", "title": "A claimed disproof of the unit-distance conjecture", "organization": "OpenAI", "model": "Unreleased reasoning model", "domain": "Mathematics", "date": "2026-05-20", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Claimant report / external checks reported",
      "dateBasis": "Date of the public announcement. This review has not reconstructed the earlier internal discovery timeline.",
      "claim": "OpenAI reports a construction giving a polynomial improvement over classical grid examples for the planar unit-distance problem.",
      "assessment": "A potentially major new result, but this initial review has not compared the proof with the closest constructions or independently established its quantitative improvement. It remains unassessed rather than receiving a score from the announcement alone.",
      "prior": {"title": "Classical grid constructions and incidence bounds", "date": "2025-07-21", "summary": "The existing literature distinguishes lower-bound constructions from upper bounds on the maximum number of unit distances. A 2025 paper provides pre-claim context; an upper bound is not an interchangeable baseline for a new lower bound.", "relation": "Pre-claim problem context; closest construction unresolved", "url": "https://arxiv.org/abs/2507.15679v1"},
      "metric": null,
      "limitations": "Independent specialists are reported as checking the result, but this leaderboard has not verified the proof or its exact prior-art distance. No numerical improvement is inferred from unmatched upper and lower bounds.",
      "sources": [{"title": "An OpenAI model has disproved a central conjecture in discrete geometry", "url": "https://openai.com/index/model-disproves-discrete-geometry-conjecture/", "date": "2026-05-20", "role": "Claim"}, {"title": "Erdős's unit distance problem and rigidity — v1", "url": "https://arxiv.org/abs/2507.15679v1", "date": "2025-07-21", "role": "Prior art"}],
      "queries": ["OpenAI unit distance conjecture May 2026", "Erdos unit distance problem prior constructions 2025 rigidity"]
    },
    {
      "id": "openai-navier-stokes", "title": "A proposed resolution of the Navier–Stokes problem", "organization": "OpenAI", "model": "Internal research model", "domain": "Physics", "date": "2026-09-08", "scope": "discovery", "novelty": null, "confidence": "Low", "evidence": "Claimant report / proof artifacts",
      "dateBasis": "Public announcement date. A September 10 update discusses concurrent work; that update is later context, not pre-claim evidence.",
      "claim": "OpenAI announces a proposed solution of the Navier–Stokes Millennium Prize problem and releases a paper and a formal proof artifact.",
      "assessment": "A potentially major result requiring independent specialist assessment. The closest proof strategy, theorem assumptions, and priority relative to concurrent work have not been resolved by this initial review.",
      "prior": {"title": "Tao's averaged Navier–Stokes blowup construction", "date": "2014-02-03", "summary": "Tao established finite-time blowup for a modified, averaged equation. That result is relevant background but is not a solution for the original Navier–Stokes equations.", "relation": "Related proof program, not an equivalent theorem or established closest precedent", "url": "https://arxiv.org/abs/1402.0290v1"},
      "metric": null,
      "limitations": "No proof verification or adjudication of priority was performed. Concurrent results for different equations must not be treated as identical. No claim of copying is made.",
      "sources": [{"title": "On the Navier–Stokes Millennium Prize Problem", "url": "https://openai.com/index/navier-stokes-solution/", "date": "2026-09-08", "role": "Claim"}, {"title": "Finite time blowup for an averaged three-dimensional Navier–Stokes equation", "url": "https://arxiv.org/abs/1402.0290v1", "date": "2014-02-03", "role": "Prior art"}],
      "queries": ["OpenAI Navier Stokes September 2026 prior work", "Tao averaged Navier Stokes 2014 blowup"]
    }
  ]
}