What AI has
actually contributed

A curated registry of scientific and mathematical results AI contributed to, graded on how each result was verified and how much the AI actually did, so you can tell a machine-checked proof from a press release.

Evidence vs. autonomy

supports the claim ↑counts against it ↓RetrievalSearch scaffoldAI-assistedCollaborativeAI-ledAutonomousRefutedDisputedAlready knownClaimedAuthor verifiedIndependent / peerFormalMore AI-driven →Better evidence →9419125

Circle area = findings in that combination. All 70 entries carry both grades.

Latest activity

  1. AddedSOP_2 and SOP_3 theories shown to coincideGraded Author verified and Collaborative.
  2. AddedComplete minimizer picture for Gamow's liquid drop modelGraded Author verified and AI-led.
  3. AddedA 112-vertex counterexample to the Petersen coloring conjectureGraded Formally verified and AI-assisted.
  4. AddedSendov's conjecture proved for every degreeGraded Formally verified and Collaborative.
  5. AddedCounterexamples to Schiffer's conjecture and the Pompeiu problemGraded Formally verified and AI-assisted.
  6. AddedCycle double cover conjecture proved for all bridgeless multigraphsGraded Formally verified and AI-led.
  7. RegradedCandidate treatment for dry age-related macular degenerationUpgraded from claimed to peer reviewed: the work was published in Nature (655, 497-505) on 19 May 2026, which also supplies the primary source the entry previously lacked.
48 of 70 entries have never been independently checked.Open the review queue
70Entries on record
56Well verified
28AI-led or autonomous
5Negative or contested

Evidence chain

Of 70 entries. 62 link no counterargument: a gap, not a consensus.

Findings per year

By topic area

Every figure here is derived from the registry on each build, so none of it can drift from the entries below. See every chart.

Every link is labelled by what it is: the original work, the announcement, press coverage, independent commentary, or a challenge to the claim. Most of the distance between a result and a headline is in the last four. How we classify →

Independent
2026-08-13
Author verified Collaborative
Model
ChatGPT 5.6
Field
mathematics
Posed
2004 · open 22 yrs

SOP_2 and SOP_3 theories shown to coincide

The classes of SOP_2 and SOP_3 first-order theories are the same, answering a 2004 question of Džamonja and Shelah and collapsing the bottom of the SOP hierarchy to a single class.

Džamonja and Shelah introduced the tree configurations SOP_1 and SOP_2 in 2004 and asked whether either implication in the chain from SOP_3 to SOP_2 to SOP_1 reverses. Mutchnik answered the second, proving SOP_1 equals SOP_2, and the surviving half organised much of the subsequent NSOP_1 literature, with partial results by Conant, by Kaplan, Ramsey and Simon, and by Mutchnik. This five-page paper proves that SOP_2 implies SOP_3, the converse having been known from the start, which collapses the bottom of the hierarchy to a single class and carries consequences for Keisler-order maximality. The SOP_n hierarchy for n at least three is untouched, as is everything above it.

WithArtem Chernikov

Original work
arXiv
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The question is one of the two 2004 Džamonja–Shelah questions that organised two decades of classification theory, and the paper's own introduction traces it as open through Mutchnik's celebrated resolution of the other half and the partial results that followed. It is repeatedly highlighted in the NSOP_1 literature as the surviving open case. No prior proof appears. The result is new.

Caveats

A preprint one day old at entry, not peer-reviewed and not formalized. The acknowledgements record comments from Itay Kaplan and Scott Mutchnik on a preliminary version, which is expert attention on a draft rather than independent verification, so the entry stays at author verified. Autonomy graded collaborative: the entire disclosure is one sentence, that the proof was found using ChatGPT 5.6 and simplified and streamlined by the author, which is ambiguous between the model supplying the key insight and the author using it as a tool. The weaker defensible reading applies.

Independent
2026-08-12
Author verified AI-led
Model
ChatGPT 5.6 Pro
Field
mathematics

Complete minimizer picture for Gamow's liquid drop model

Balls uniquely minimize the Gamow liquid drop energy at every volume up to a sharp threshold of about 3.51, and above that threshold no minimizer exists, closing a gap that partial results had narrowed from both sides without meeting.

Gamow's liquid drop functional adds De Giorgi perimeter to a Coulomb repulsion term, and the central question in its literature is the fixed-volume minimization picture: for which volumes a minimizer exists, and whether it is the ball. This proves that balls uniquely minimize, up to translation and null sets, at every volume up to a threshold of 5 times 2 minus 2 to the two thirds, divided by 2 to the two thirds minus 1, roughly 3.51, and that above that threshold no minimizer exists at all. A corollary settles the minimal binding energy question of Frank and Lieb, identifying the infimum of energy per unit volume and the ball of volume 5/2 as the unique minimizer. The mechanism is a capacitary estimate sharpening an Agostiniani–Mazzieri monotonicity formula through Gauss–Bonnet, an improvement the authors note applies only to this particular weight and only in three dimensions.

WithOtis Chodosh, Matilde Gianocca

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The sharp threshold is the tracked open problem of the liquid drop literature, followed across calculus of variations, mathematical physics and geometric analysis, with a 2017 Notices of the AMS survey and partial results carrying Lieb, Otto, Figalli and Maggi. The prior state of the art is stated in the paper and checks out: minimality of the ball was known only up to volume 1 (Chodosh–Ruohoniemi, CPAM 2025) and nonexistence only from volume 7.5 upward (Schulz, arXiv:2608.09000, posted two days earlier), leaving the interval from 1 to 7.5 open. Frank and Nam had proved existence up to the threshold in 2021, and the new proof uses that result rather than superseding it. Uniqueness across the whole range and nonexistence immediately above the threshold are new.

Caveats

A preprint, not peer-reviewed, with no formalization. Verification is author-side: the authors state they checked and reworked the proof, which is not independent replication. Autonomy graded ai-led on the paper's own disclosure, which says the results were obtained by ChatGPT 5.6 Pro over a series of chats without significant assistance from the authors and that the fundamental strategy remains close to the model's original output; the authors checked the proof, reworked it and wrote the manuscript, and state the article contains no AI-written text.

Anthropic
2026-08-10
Formally verified AI-led
Model
Claude (unreleased research version)
Field
mathematics

Proportion of zeta zeros on the critical line raised to 67.25%

An unconditional proof that at least 67.25% of the nontrivial zeros of the Riemann zeta function are simple and lie on the critical line, up from a previous record of about 41.6%, with a machine-checked Lean formalization.

The Riemann hypothesis asserts every nontrivial zero lies on the critical line. Short of proving it, the standard measure of progress is the proportion of zeros known unconditionally to lie there: Selberg established a positive proportion, Levinson reached a third in 1974, Conrey two fifths in 1989, and the record had crept to roughly 41.6%. This work proves at least 3/2 - (1/sqrt 2)cot(1/sqrt 2) = 67.25...% of zeros are simple and on the line, and that at least 0.83625 of zeros are distinct. The decisive step combined the unconditional pair-correlation work of Baluyot, Goldston, Suriajaya and Turnage-Butterbaugh with a 2000 paper of Bombieri, treating the whole function space at once with the quadratic form allowed to be non-diagonal rather than splitting it. The process is itself the record worth noting: the model was asked to attempt the Riemann hypothesis outright, with the mathematical choices left to it, and the bound improvement fell out of failing at that. It generated and discarded roughly 650 ideas in a first session; in a second it coordinated about 60 subagents which ran some 2,400 shell commands, wrote hundreds of scripts, checked numerically against known zeros and refereed one another. Two subagents developed the key ideas, thirteen fed them, thirty tried and failed, thirteen validated, two drafted the paper, across roughly 31 million output tokens in two Claude Code sessions. The model also proposed writing the result up, checked 54 arXiv papers for prior art, and recommended a human number theorist validate it.

WithJarred Sumner, Levent Alpöge, Ralph Furman, Eric Easley

Announced
Anthropic
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The proportion ladder is a documented literature running from Hardy (1914) and Selberg (1942) through Levinson's one third (1974) and Conrey's two fifths (1989), with subsequent refinements reaching about 41.6%; the paper states the prior record and cites it. The model itself searched 54 arXiv papers for prior art before write-up. No earlier unconditional bound above 42% appears in the literature. The result is a new bound, not a retrieval of a known one.

Caveats

Not peer-reviewed. The Riemann hypothesis itself is untouched, and Anthropic states it does not expect these techniques to lead to a proof of it. The Lean formalization is sorry-free and anchors its statements on Mathlib's own riemannZeta and analyticOrderAt with the corresponding counting functions rather than an assumed form of the result, which is real statement anchoring; but the assembly of those primitives into the proportion claim is the paper's own, and that assembly is exactly what nobody independent has audited. The same organisation produced both the proof and its formalization. Two Anthropic mathematicians studied and validated the work, which is author-side review, not independent replication. Autonomy graded ai-led rather than autonomous: humans posed the problem, supplied the environment and validated the output.

Independent
2026-08-08
Formally verified AI-assisted
Model
Unnamed OpenAI model
Field
mathematics
Posed
1985 · open 41 yrs

A 112-vertex counterexample to the Petersen coloring conjecture

An explicit simple bridgeless cubic graph on 112 vertices admits no Petersen coloring, and hence no normal 5-edge-coloring, refuting a conjecture of Jaeger from 1985 with machine-checked UNSAT certificates.

Jaeger conjectured in 1985 that every bridgeless cubic graph admits a Petersen coloring: a map from its edges into the edges of the Petersen graph sending the three edges at each vertex to three edges meeting at a common vertex of the Petersen graph. Equivalently, every bridgeless cubic graph has a normal 5-edge-coloring. The conjecture implies both the Berge–Fulkerson conjecture and the 5-cycle-double-cover conjecture. This paper gives an explicit simple connected bridgeless cubic graph on 112 vertices, of girth five and edge and vertex connectivity three, with no such coloring; the graph is pinned by a SHA-256 digest in the theorem statement itself. Non-existence is established by direct SAT formulations, with CaDiCaL 3.0.1 returning UNSAT and drat-trim verifying the resulting DRAT proofs. A second, non-isomorphic counterexample of the same order is supplied. The implication runs one way, so Berge–Fulkerson and the 5-cycle-double-cover conjecture stay open; combined with a theorem of Ma, Mattiolo, Steffen and Wolf, one counterexample yields infinitely many.

WithBryce Putman

Original work
arXivDOI
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Petersen coloring conjecture is tracked as open in the Open Problem Garden, whose entry calls it an extraordinary conjecture, and it carries a substantial literature on normal edge-colorings and sublinear approximations built on the assumption that it holds. No prior counterexample appears anywhere in that literature. The construction is a new concrete graph, not a retrieval of an existing example.

Caveats

A preprint, not peer-reviewed. The paper's computational provenance section states only that OpenAI language-model systems were used extensively in the discovery, computational search, verification and preparation of the work, and names no product, version or division of labour, so the model field records an unnamed OpenAI system and autonomy is graded ai-assisted, the weakest reading the disclosure permits. The paper does not claim 112 is the minimum order for a counterexample.

Independent checks

vibemathed.com curator reproduction: rebuilt the 112-vertex graph from the appendix edge table with a SHA-256 matching Theorem 1.1, rederived every property claimed there, and re-proved non-existence with an independently written CNF encoder solved by CaDiCaL rather than by replaying the shipped certificates; six control graphs including the Petersen graph itself came back satisfiable through the same encoder · link ↗

Independent
2026-08-05
Formally verified Collaborative
Model
GPT-5.6 Pro
Field
mathematics
Posed
1959 · open 67 yrs

Sendov's conjecture proved for every degree

For every complex polynomial of degree at least two whose zeros all lie in the closed unit disk, each zero has a critical point of the polynomial within distance one, closing a question open since 1959.

Sendov's conjecture states that if every zero of a complex polynomial of degree n at least two lies in the closed unit disk, then every zero has a critical point within distance one. Degrees up to eight were settled piecemeal between 1969 and 1999, and Tao proved the conjecture for all sufficiently large degree in 2020 without specifying the threshold, which left the middle range open. Lech Mazur, working with GPT-5.6 Pro, produced an argument covering every degree along with a Lean development of roughly 90,000 lines. Terence Tao then digested the proof, describing it as remarkably elementary with no complex analysis used beyond the fundamental theorem of algebra, and formalized the entire argument himself in about 15,000 lines. He reports that the same argument resolves the Phelps–Rodriguez conjecture in full generality.

WithLech Mazur

Challenge
none linked
Novelty check, caveats & sources
Novelty check

Sendov's conjecture is a named 1959 problem with its own Wikipedia article in four languages and a partial-results literature spanning 67 years. Tao's 2020 paper on the sufficiently-high-degree case (arXiv:2012.04125) states the general case as open, which fixes the prior state of the art precisely. No earlier proof covering all degrees appears in the literature, and the Phelps–Rodriguez corollary is new with it. The result is a new proof, not a retrieval.

Caveats

Not peer-reviewed. Mazur's own Lean package cannot be recompiled as distributed: the published bundle ships no lakefile or manifest and excludes Mathlib, so the formal grade rests on Tao's independent formalization rather than on the original artifact. Autonomy graded collaborative rather than ai-led: the disclosure credits GPT-5.6 Pro with substantial contribution to the discovery and derivation of the proof, including exploration, proof development, exact computational testing and adversarial auditing, but it also has Mazur directing the research workflow and selecting and reconciling the model's outputs, which is mathematical judgment rather than cleanup. The weaker defensible reading applies.

Independent checks

Terence Tao: digested the argument and formalized the whole of it in Lean independently, at about 15,000 lines against the original's roughly 90,000, and states that it resolves both the Sendov and Phelps–Rodriguez conjectures in full generality · link ↗

Independent
2026-08-05
Formally verified AI-assisted
Model
GPT-5.6, Claude Opus 4.8, Claude Fable 5
Field
mathematics
Posed
1929 · open 97 yrs

Counterexamples to Schiffer's conjecture and the Pompeiu problem

Infinitely many planar domains that are not balls admit a Neumann eigenfunction of the Laplacian that is constant on the boundary, disproving Schiffer's conjecture and, through a classical equivalence, the 1929 Pompeiu problem.

Pompeiu asked in 1929 whether a bounded domain over which some nonzero function integrates to zero under every rigid motion must be a ball. Schiffer's 1957 reformulation asks the same thing through Neumann eigenfunctions of the Laplacian that are constant on the boundary, and appears as Problem 80 on Yau's 1982 list; Williams proved the two formulations equivalent for simply connected domains in 1976. Cao-Labora and de Dios Pont construct infinitely many N-fold symmetric planar domains, with N large, that are not balls and carry such an eigenfunction. The strategy is to relax the problem so that N may be any real number, which corresponds to the Schiffer problem only at integers, apply bifurcation theory there, and show the size of the local bifurcation branch can be taken independently of N. Their Corollary 1.2 applies Williams' equivalence to the same domains, so a single construction settles both problems.

WithGonzalo Cao-Labora, Jaume de Dios Pont

Original work
arXivGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Schiffer's conjecture is Problem 80 on Yau's list, with a partial-results literature running since the 1970s, and the Pompeiu problem has stood since 1929. Both are recorded as open in the standard references and in DeepMind's formal-conjectures repository, which carries a Lean statement of the Pompeiu problem as an unsolved challenge that this work's formalization closes. No prior counterexample to either appears. The construction is new.

Caveats

A preprint, not peer-reviewed. The Lean 4 verification was written by GPT-5.6 from an early draft of the paper and lives in an author's own repository; it had no independent audit at entry time, and the same group produced both the proof and its formalization. Autonomy graded ai-assisted: the paper is explicit that the novel construction strategy is the authors' own, and that the models were used to verify the asymptotic estimates numerically, to produce first drafts of the proofs of the Bessel function estimates, and to help with exposition.

Independent
2026-08-04
Formally verified AI-assisted
Model
GPT-5.6 Pro
Field
mathematics
Posed
1978 · open 48 yrs

Asymptotic degree-diameter problem resolved for fixed diameter

The maximum order of a graph with maximum degree d and diameter k satisfies n_k(d)/d^k tending to 1 as d grows, for every fixed k, resolving the asymptotic form of a question open since 1978, with a Lean 4 formalization.

The Moore bound caps the order of a graph of maximum degree d and diameter k at roughly d^k. Whether that cap is asymptotically attainable for fixed diameter was open: Bannai and Ito had settled which exact Moore graphs exist, and constructions had approached the bound only for small diameters. This proves the limit of n_k(d)/d^k is 1 for every fixed k, with a companion lower bound on the edge variant and a tight asymptotic for the bipartite version. The paper carries an explicit tool and computational resource disclosure: GPT-5.6 Pro was used in exploratory brainstorming directed by the authors, who formulated the required parameter scale and proposed higher-rank spherical buildings as the geometric setting. The model suggested splitting complete flags into their odd- and even-rank subflags, initially in connection with the edge problem; that suggestion became the halved-flag construction carrying the main theorem, and the graph is named for it.

WithWouter Cames van Batenburg, Samuel Korsky

Original work
arXivGitHub
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The degree-diameter problem is a named and long-tracked question in extremal graph theory, surveyed by Bermond and Bollobas (1981) and with Bollobas's 1978 Extremal Graph Theory as the reference point for the asymptotic question; Bachraty, Siagiova and Siran had approached the bound for diameter three in 2019. The authors note that after completing the work they learned the projective-plane construction of an earlier paper shares the rank-2 incidence graph underlying the k=1 case of theirs, and state the higher-rank construction and its odd-even routing argument are independent and different. No prior proof of the general asymptotic result appears.

Caveats

Not peer-reviewed. Autonomy is graded ai-assisted, the weaker defensible reading, because the disclosure is explicit that the authors developed the suggestion into the construction, formulated and verified all mathematical arguments, and take full responsibility for the content; the model's contribution was one structural idea during brainstorming. Generative tools also assisted with the Lean formalization. The Lean repository is author-produced and had no independent audit at announcement.

Independent checks

vibemathed.com curator check of the Lean repository: checked at commit 32beb227: DegreeDiameter.theorem_1_1 states Theorem 1.1 itself and corollary_1_2 states Corollary 1.2, neither a weakened lemma, with a second independent route proved alongside each; no sorry or admit in the sources · link ↗

OpenAI
2026-08-01
Formally verified AI-led
Model
Astra
Field
mathematics
Posed
1999 · open 27 yrs

Ten results in mathematics and theoretical computer science with Lean certificates

An unreleased OpenAI model produced ten results including the first explicit non-sofic group and a disproof of Connes's rigidity conjecture, each published with a machine-checkable Lean 4 proof.

The ten cover high-dimensional sphere packing, binary and spherical codes, non-sofic groups, Connes's rigidity conjecture, arithmetic circuit complexity, quantum parallel repetition, the closest vector problem, Ehrhart's volume conjecture, multicolor Ramsey numbers and extremal number conjectures. Soficity was introduced by Gromov in 1999 and whether a non-sofic group exists had stood open since; Connes posed his rigidity conjecture in 1980, and the disproof constructs infinitely many non-isomorphic property (T) groups sharing a von Neumann algebra. Humans used the same model to prepare the manuscripts and then to formalize each argument in Lean 4.32.0. OpenAI put the token cost of finding all ten at roughly $2,000 at Sol API rates.

Original work
GitHub
Announced
OpenAI
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Each of the ten is a named open problem with a documented origin: soficity from Gromov 1999, Connes rigidity from 1980, and the sphere-packing exponent improvement is the first on the general bound since 1978. The Lean repository states which prior bound each result improves. Thomas Bloom, who maintains erdosproblems.com and publicly corrected OpenAI's 2025 Erdős retrieval episode, called these results big news, which is meaningful given that he is the person who caught the previous overclaim.

Caveats

The certificates are checkable and the claim that the theorems hold is as solid as Lean makes it. The claim about provenance is not checkable: Astra is unreleased, so no outside party can test whether the model produces these results, and the account of how they were found rests entirely on OpenAI. Grading verification as formal reflects the proofs, not the provenance; a reader who cares only about who found it should read this as claimed. Specialist review of the mathematical significance is still pending in several cases. Noam Brown noted there are no Millennium Prize problems here. Press summaries of the ten disagree with the repository's own list, some substituting specific Erdős problem numbers that do not appear in it; the list above follows the repository.

Tencent Hunyuan
2026-07-29
Formally verified AI-led
Model
Hy3 (Hyra research agent)
Field
mathematics

Optimal exponent relating sumsets and difference sets determined

The least universal exponent c with |A+A|/|A| bounded by (|A-A|/|A|)^c for every finite integer set A is determined, settling the extremal question with a Lean 4 formalization.

For a finite set A of integers with at least two elements, the ratio of log(|A+A|/|A|) to log(|A-A|/|A|) measures how far sumset growth can outrun difference-set growth. Determining its largest possible value fixes the least universal exponent c such that |A+A|/|A| is at most (|A-A|/|A|)^c for every such A. Tencent Hunyuan's Hyra research agent, powered by the Hy3 model, explored and optimized finite-set constructions over roughly a 24-hour run, moving from finite numerical searches to natural-language proposals of general constructions and supporting arguments, and produced the construction underlying the paper. The human authors independently checked the construction, corrected and rewrote the exposition, and prepared the final proof manually.

WithHaowei Lin, Shanda Li

Original work
arXiv
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The relationship between sumset and difference-set growth is a standard question in additive combinatorics with a long record of partial bounds on the exponent; the extremal value was recorded as undetermined. The paper states the prior bounds it improves on. The construction is new, not a retrieval of a known extremal family.

Caveats

A newly released arXiv v1 preprint, not peer-reviewed. The Lean 4 and Mathlib formalization is author-provided; the repository reports lake build completing with no sorry declarations or warnings, but the principal asymptotic and supremum results rely on three native_decide certificates for elementary finite computations over a 12-element base-39 digit block, so those steps are trusted to the Lean kernel's decision procedure rather than proved term by term. Autonomy graded ai-led rather than autonomous: the agent produced the construction, but humans posed the problem, checked it and wrote the final proof.

Independent
2026-07-28
Formally verified AI-led
Model
GPT-5.6 Sol Ultra, Claude Fable 5
Field
computer-science
Posed
2001 · open 25 yrs

Kemeny rank aggregation shown NP-hard for three voters

Computing a Kemeny-optimal aggregate ranking is NP-hard when the input is exactly three complete rankings, closing the minimal open case left by hardness results for even voter counts.

Kemeny aggregation asks for the ranking minimizing total disagreement with a set of input rankings. Hardness was known for every even number of voters n at least 4, while n = 2 is solvable in polynomial time, leaving three voters as the minimal open case since 2001. GPT-5.6 Sol Ultra found the reduction from MAX CUT; Claude Fable 5 helped simplify parts of it. Combined with the earlier results, every fixed number of voters n at least 3 is now known to be hard. The reduction is Lean-checked alongside an author-written preprint.

WithDominik Peters

Original work
arXiv
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The complexity of Kemeny aggregation for three voters is a specifically tracked open case in the computational social choice literature, following Dwork, Kumar, Naor and Sivakumar's 2001 work establishing hardness for four voters and the subsequent extension to all even counts. The odd cases, and three in particular, were repeatedly noted as open. No prior hardness proof for exactly three voters appears.

Caveats

Not peer-reviewed. The Lean check covers the reduction; it is author-produced and had no independent audit at announcement. Autonomy graded ai-led rather than autonomous: the model found the reduction, but the human author posed the problem, reviewed the argument and wrote the paper.

Independent
2026-07-27
Formally verified AI-led
Model
ChatGPT 5.6 Pro
Field
mathematics
Posed
2004 · open 22 yrs

Feige's conjecture on sums of nonnegative random variables settled

For independent nonnegative random variables with mean at most 1 and sum S, the probability that S is below its mean plus one is at least 1/e, the sharp constant Feige conjectured in 2004; three independent proofs appeared within days.

Feige proved in 2004 that for independent nonnegative random variables X_1..X_n with E[X_i] <= 1 and S their sum, P(S < E[S] + 1) is bounded below by a positive constant, established 1/13, and conjectured the sharp value 1/e. The primary paper states plainly in its abstract that the proof was found by ChatGPT 5.6 Pro, which combined the Vlassis-Thomas Dirichlet calibration theorem with Grunbaum-type convex geometry; the authors then checked, revised and rewrote the argument. The sharper form determines the optimal small-deviation bound. An end-to-end Lean formalization accompanies the paper, covering the Vlassis-Thomas theorem, Grunbaum's centroid theorem and the combining argument, developed with Codex.

WithWeibo Fu, Yanjun Han, Guanyang Wang, Jun Yan, Peng Zhang, Zhengqing Zhou

Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Feige's conjecture is documented as open from his 2004 paper on sums of independent random variables, where he proved the weaker constant 1/13. The sharp 1/e constant is recorded as an open problem in the subsequent small-deviation literature and had stood for 22 years. No prior proof of the sharp constant appears; the near-simultaneous arrival of three independent proofs within days confirms it was live and open, not settled and forgotten.

Caveats

Not peer-reviewed; all three proofs are preprints. The Lean formalization accompanies the primary paper and is author-produced. Autonomy is graded ai-led rather than autonomous because the human authors posed the problem, then checked, revised and rewrote the argument for publication. The three independent proofs are genuine corroboration of the result, though two of the three were themselves AI-assisted, so they are not fully independent of the method.

Independent checks

Zipei Nie and Jiaye Wei (independent second proof): independently proved the same sharp 1/e bound, obtained with assistance from GPT-5.6 Sol · link ↗

Mark Stander (independent third proof): further independent proof of the conjecture · link ↗

Independent
2026-07-22
Claimed AI-led
Model
GPT-5.6 Pro
Field
computer-science
Posed
1999 · open 27 yrs

Counterexample to the Dinitz–Garg–Goemans conjecture

An explicit seven-node network whose cheapest single-route (unsplittable) shipping costs more than its fractional cost even under the allowed capacity slack, disproving a conjecture from the late 1990s.

Dinitz, Garg, and Goemans proved that any fractional multicommodity flow can be rounded to an unsplittable (single-path-per-demand) flow while violating each arc's capacity by at most the maximum demand. Goemans conjectured this rounding could also be done without increasing the total cost. The counterexample is a directed graph on seven nodes carrying three demands (15, 10, 15); its fractional solution costs 58, yet every unsplittable routing that stays within the allowed capacity cushion (violation ≤ 15) costs at least 60. Rybin reached it with GPT-5.6 Pro in four short prompts and reported that the model returned proof certificates, an exhaustive-enumeration verification program, machine-readable data, and LaTeX source.

WithDmitry Rybin

Announced
X
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Dinitz–Garg–Goemans unsplittable-flow result is from the late 1990s (D. Dinitz, N. Garg, M. Goemans, 'On the single-source unsplittable flow problem'); the cost-preserving strengthening was an open conjecture attributed to Goemans. No prior counterexample or resolution appears in the literature. The construction is a new concrete instance, not a retrieval of an existing example.

Caveats

Not peer-reviewed or machine-checked in a proof assistant. Verification rests on the author's own exhaustive-enumeration program over a finite instance, which anyone can rerun but which had no logged independent replication at announcement. The result was shared informally on X with a public GPT-5.6 Pro transcript. One third party (Hensen Juang) publicly generalized the instance into an infinite parametric family on the same seven nodes, which is consistent with the claim but is not a formal independent check. Autonomy graded ai-led: the model produced the construction; the human posed the problem and verified. Downgraded from 'author verified' when source classification showed no primary artifact is linked: the construction was published as a chat transcript rather than a paper or repository. Readers have checked the arithmetic and found it consistent, but there is nothing citable to point at, so the grade caps at 'claimed' until someone links a standalone write-up.

Independent
2026-07-20
Author verified AI-led
Model
GPT-5.6 Sol Pro + Claude Fable 5
Field
mathematics
Posed
2017 · open 9 yrs

Counterexamples to the Gaussian moments conjecture

Explicit low-degree polynomials disproving a 2017 conjecture that had been proposed as a route to proving the Jacobian conjecture, found days after that conjecture was itself disproved.

Derksen, van den Essen and Zhao proposed the Gaussian moments conjecture and showed that the Jacobian conjecture would follow from it. Long reports that a cubic counterexample in four variables was produced by GPT-5.6 Sol Pro with no human intervention after the initial prompt, which simply noted that the Jacobian conjecture had just been disproved and asked whether a small counterexample to GMC might therefore exist; Claude Fable 5 then found a quartic counterexample in three variables and supplied independent algebraic checks. Together these show GMC(n) is false for every n ≥ 3.

WithChristopher D. Long

Original work
arXivarXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Gaussian moments conjecture is due to Derksen, van den Essen and Zhao (Israel J. Math., 2017), building on their earlier moment-vanishing and integral conjectures. No prior counterexample appears in the literature; the conjecture's interest was precisely that it implied the Jacobian conjecture, which was disproved on 2026-07-19. The two-variable case GMC(2) is explicitly left open by this paper.

Caveats

arXiv preprint, not peer-reviewed, single author. The constructions are explicit polynomials and checkable by computer algebra, which lowers the cost of verification, but no independent check is on the public record yet. The result is a corollary of the Jacobian collapse rather than an independent line of attack: the prompt that produced it began by telling the model that the Jacobian conjecture had fallen.

Harmonic
2026-07-20
Formally verified Autonomous
Model
Aristotle
Field
mathematics
Posed
1969 · open 57 yrs

Eight problems from the Kourovka Notebook solved and formalized in Lean

A formal reasoning agent developed the proof strategies for eight open problems in group theory without human guidance on which mathematical steps to take, and machine-checked every one in Lean 4.

The Kourovka Notebook has collected open problems in group theory since 1965, with a new issue every two to four years. Aristotle solved eight of them: 3.46 (a group with exactly two maximal locally soluble normal subgroups), 18.50, 19.25, 20.125 (a surjective non-injective Rota-Baxter operator on a non-abelian group), 21.8, 21.24, 21.147 and 21.150. The oldest first appeared in the third issue in 1969. Solutions take the form of proofs, counterexamples and constructions, each formalized in Lean 4 and published in a companion repository. The authors then informalized the Lean proofs into conventional prose and submitted both to the Notebook editors and the original problem proposers, who accepted them before publication.

WithWouter van Doorn, Elias Judin, Pietro Monticone, Daniel Morrison

Original work
arXivGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Kourovka Notebook is itself the authoritative open-problems register for group theory, and each of the eight is recorded there as open with a named proposer, so novelty is established by the source of record rather than by a literature search. The authors submitted solutions to the Notebook editors and the problem proposers for acceptance before publishing, and the initial submissions remain in the Notebook's own preprint repository. Problem 3.46 traces to Plotkin's survey question on products of locally soluble normal subgroups, which Baumslag, Kovács and Neumann had partly addressed in the negative.

Caveats

Not peer reviewed as a journal article, though the individual solutions were accepted by the Kourovka Notebook editors and the problem proposers, which is the relevant gatekeeping for this venue. The autonomy grade rests on the authors' own disclosure: they state Aristotle developed the proof strategies without human guidance about which steps to take, while they monitored its work, intervened when clarifications were needed, and built a library of relevant concepts with it. That library work and the final canonisation for Mathlib upstreaming were done partly by hand. The Lean certificates are the load-bearing evidence and anyone can rerun them; the claim about how the arguments were found rests on the authors' account.

Independent
2026-07-20
Formally verified AI-led
Model
ChatGPT 5.6 Sol
Field
mathematics
Posed
2007 · open 19 yrs

Gaussian product inequality conjecture proved

For any centered Gaussian vector and positive exponents, the expectation of the product of absolute powers is at least the product of the individual expectations, settling a conjecture posed in 2007.

The conjecture states that for a centered Gaussian vector X = (X_1..X_n), not necessarily nondegenerate, and any positive exponents a_1..a_n, the expectation of the product of |X_i|^a_i is at least the product of the expectations of |X_i|^a_i. The stronger form adds that when every variance is positive, equality holds exactly when the coordinates are independent. The model returned a complete and correct solution from Greaves's prompt. It did not include the equality characterization, but only because that case was not in the prompt as posed. The proof was formalized in Lean using Codex, and the full prompt and output are public as a shared chat transcript.

WithFrédéric Ouimet, Dylan Greaves

Original work
DOIGitHub
Announced
ChatGPT
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Gaussian product inequality is a named conjecture dating to 2007, with a substantial partial-results literature covering special cases such as even integer exponents and small dimensions; it was recorded as open in the general case. The proof settles the general statement. No prior general proof appears in the literature.

Caveats

Not peer-reviewed; released as a preprint with a public chat transcript rather than through a journal. The Lean formalization is author-produced. The AI output omitted the equality-characterization half of the strong form, which was supplied around it. Checks by other models (ChatGPT 5.6 Sol Pro, Gemini 3.1 Pro, Grok 4.5) are not independent human verification and do not raise the grade; the human check by Ouimet is author-side.

Anthropic
2026-07-19
Formally verified Collaborative
Model
Claude Fable 5
Field
mathematics
Posed
1939 · open 87 yrs

Counterexample to the Jacobian conjecture in dimension three

An explicit polynomial map in three variables with constant Jacobian determinant −2 that is nevertheless not invertible, disproving a conjecture open since 1939.

The map F(x,y,z) = (u³z + y²u(4+3xy), y + 3xu²z + 3xy²(4+3xy), 2x − 3x²y − x³z) with u = 1+xy has Jacobian determinant identically −2, yet sends the three distinct points (0,0,−¼), (1,−3/2,13/2) and (−1,3/2,13/2) all to (−¼,0,0). A map with a global inverse cannot be three-to-one. Alpöge, a number theorist at Anthropic, announced it on X the day it was found.

WithLevent Alpöge

Original work
jacobianfun.org
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Jacobian conjecture (Keller, 1939) has been a celebrated open problem for 87 years, with many published false proofs in both directions. No prior counterexample in any dimension over characteristic 0 exists in the literature. Ott-Heinrich Keller's original formulation is the one addressed.

Caveats

Not yet peer-reviewed. The formula is public and checkable in seconds by computer algebra, which makes conventional peer review less load-bearing than usual, but the official record lists the conjecture as open until the literature catches up. The division of labor between Alpöge and the model has not been fully documented; autonomy graded conservatively pending a transcript.

Independent checks

whataifound.org (symbolic recomputation, SymPy): confirmed, det J = −2 identically; all three points map to (−¼,0,0)

Multiple mathematicians via public computer-algebra checks: confirmed · link ↗

Independent
2026-07-14
Formally verified AI-led
Model
GPT-5.6 Sol Pro
Field
mathematics
Posed
1996 · open 30 yrs

Near-quadratic lower bound for derivative-free convex optimization

A lower bound of order d²/log d on the number of exact function evaluations needed to minimize a convex Lipschitz function, closing a gap open since 1996 and showing a 1996 algorithm was essentially optimal.

In zeroth-order optimization an algorithm may only query a function's value, never its gradient. Protasov's 1996 algorithm needed about d² evaluations in d dimensions, but the best known lower bound was only about d, leaving open whether a much faster method existed. The new bound of Ω(d²/log(d+1)) matches the upper bound to polylogarithmic factors. Kerger reports that after a ten-page prompt built on his own earlier failed attempts, the model produced the complete proof in a single 2.5-hour session with no intervention; he then reviewed it and formally verified the core bound in Lean 4, which compiles against Mathlib with no `sorry` and no bespoke axioms.

WithPhillip Kerger

Original work
arXivGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The gap between Protasov's O(d²) upper bound (1996) and the O(d) lower bound is documented in the derivative-free optimization literature and had stood for three decades. No prior matching lower bound appears in the literature. The paper is explicit that the AI produced the argument.

Caveats

arXiv preprint, not peer-reviewed. The author states plainly that 'it is accurate to say that the AI model used solved the problem, not the author of this paper', which is the basis for the ai-led grade; the human wrote the prompt, reviewed the proof, and did the Lean formalization. Graded `formal` on the strength of the machine-checked Lean artifact for the core bound, not on the preprint as a whole.

Independent checks

Lean 4 / Mathlib compiler (author-run, artifact public): core Ω̃(d²) lower bound compiles with no sorry and no bespoke axioms · link ↗

Independent
2026-07-14
Formally verified Collaborative
Model
GPT-5.6 Pro, GPT-5.6 Sol
Field
mathematics

Sabidussi's compatibility conjecture proved

The edges of a finite connected multigraph carrying a closed eulerian trail can be partitioned into circuits so that no circuit contains two edges used consecutively in the trail, with a Lean 4 formalization.

Sabidussi's conjecture asks whether, given a finite connected multigraph together with a closed eulerian trail, the edge set can always be partitioned into circuits none of which contains two edges that the trail uses consecutively. The proof establishes this and in fact does more: it four-colours the edges so as to satisfy the constraints. The work was developed with GPT-5.6 Pro and GPT-5.6 Sol, with the author reviewing the proof, and a Lean 4 formalization accompanies the preprint in the author's repository.

WithNikolay Ulyanov

Original work
arXiv
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Sabidussi's compatibility conjecture is a named open problem in graph theory concerning eulerian trails and compatible circuit decompositions, with a partial-results literature covering restricted degree conditions. It was recorded as open in the general case. No prior general proof appears.

Caveats

Not peer-reviewed. The Lean formalization is in the author's own repository and had no independent audit at announcement. Autonomy graded collaborative rather than ai-led: the source describes the proof as developed with the models and reviewed by the author, without attributing the key idea specifically to the model, so the weaker defensible reading applies.

Independent
2026-07-11
Formally verified Collaborative
Model
OpenAI Sol (construction); Claude Fable (Lean formalisation)
Field
mathematics

Counterexample to Grothendieck's question on finite flat group schemes

A finite locally free group scheme of order four whose fourth power map is not trivial, answering a question Grothendieck raised in the 1960s; one model produced the construction and another formalised it in Lean.

Grothendieck asked whether every finite locally free group scheme of order n is killed by n. Deligne proved it for commutative group schemes; the non-commutative case stayed open. The counterexample is a Hopf algebra of rank four over the ring Z[a,b]/(a³, b³, a²b+2), with coordinate algebra R[U,V]/(U² − abU + b²V, V² − a²V), whose fourth convolution power is not the convolution unit. Kevin Buzzard, told of a 12-page informal write-up, replied that he does not read AI-generated informal mathematics and asked for a Lean proof instead; four hours later a 1,076-line Lean formalisation existed, and it compiles on a laptop in under five minutes. It was submitted to mathlib as pull request #41748.

WithAkhil Mathew, Kevin Buzzard

Original work
GitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Per Buzzard's account, the question had been settled in special cases by Grothendieck, Deligne and René Schoof, with further partial results published by Emiliano Torti in 2025; no counterexample appeared in the literature. The object here is an explicit new construction rather than a rediscovered example, and the Lean file states and refutes the general claim directly.

Caveats

The mathlib pull request was open, not merged, when this entry was written, and was filed by a pseudonymous account whose module docstring credits "Codex (OpenAI) and Claude (Anthropic), under the direction of the author" rather than naming Sol or Fable; the model attribution here follows Buzzard's blog post. The `formal` grade rests on the Lean proof compiling, which Buzzard reports doing himself; the 12-page informal argument has not been peer reviewed. Autonomy graded collaborative: a human posed the question, directed the work and filed the PR, and the record of who did which step is a blog post rather than a transcript.

Independent checks

Kevin Buzzard (compiled the Lean formalisation): confirmed; 1,076-line Lean proof compiles in under five minutes · link ↗

OpenAI
2026-07-10
Formally verified AI-led
Model
GPT-5.6 Sol Ultra
Field
mathematics
Posed
1973 · open 53 yrs

Cycle double cover conjecture proved for all bridgeless multigraphs

Every finite bridgeless loopless multigraph has a multiset of cycles covering each edge exactly twice, resolving a conjecture open since 1973, with a machine-checked Lean proof of the unconditional statement.

The cycle double cover conjecture, posed by Szekeres in 1973 and independently by Seymour in 1979, asks whether every bridgeless graph admits a multiset of cycles in which every edge appears exactly twice. It is a textbook open problem in graph theory and sits at the centre of a cluster of neighbouring conjectures. OpenAI announced on 10 July 2026 that GPT-5.6 Sol Ultra, running 64 concurrent subagents, produced a three-page proof of the full statement in under an hour, and published the proof and the prompt, with a Lean 4 formalization following. The formalized route goes through the Jaeger–Kilpatrick eight-flow theorem: it constructs a nowhere-zero flow valued in three copies of the integers mod two, then converts that flow into a cycle double cover. Sang-il Oum published a ten-page exposition a week later, presenting the argument with slight modifications intended to make it more accessible.

Original work
OpenAIGitHub
Announced
OpenAIX
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The conjecture is recorded as open in the standard references, including Wolfram MathWorld, and has a documented history of proof claims posted to arXiv that were later withdrawn or found to have gaps, so the prior state was unambiguously unresolved. One thing complicates the picture rather than the novelty: Shiva Kintali posted an independent Codex-assisted proof claim of the same statement three days later (arXiv:2607.14140), also routed through nowhere-zero three-bit flows and likewise unreviewed. The result is a new proof, not a retrieval of a known one.

Caveats

Not peer-reviewed. The conjecture's history of gapped proof claims is itself a reason for caution, and mathematicians are treating this one as unconfirmed pending expert review. The Lean development does check the unconditional statement for finite loopless bridgeless multigraphs, sorry-free and admit-free with no project-specific axioms, but it is first-party, produced by the same organisation as the proof, and had no independent audit at entry time. Autonomy graded ai-led rather than autonomous: the published prompt supplies more than the problem. It instructs the model to assume for purposes of the task that a complete affirmative proof exists, enumerates the approach families to explore including the flow formulations the proof in fact used, and specifies the 64-agent search strategy in detail.

Vesuvius Challenge
2026-06-25
Author verified AI-assisted
Model
Community-developed ink-detection neural networks
Field
archaeology

First Herculaneum scroll read end to end without unrolling it

PHerc. 1667 was virtually unwrapped and read in full from X-ray scans using learned ink detection: about 1.4 metres of papyrus across roughly 22 columns of Greek, identified as Philodemus, On Gods, Book 8.

The Herculaneum scrolls were carbonised by the AD 79 eruption of Vesuvius and disintegrate if physically unrolled. The pipeline scans a scroll by X-ray tomography, traces the rolled sheet as a three-dimensional surface, flattens it, and runs a network that detects near-invisible carbon ink from sub-millimetre surface texture. Luke Farritor's model recovered the first word, πορφύρας ("purple"), in October 2023, and the 2023 Grand Prize for four passages of 140 characters was claimed in February 2024. In June 2026, PHerc. 1667 became the first scroll recovered continuously from beginning to end, and a title read in a related scroll established that Philodemus' On Gods ran to at least eight books.

WithBrent Seales, Luke Farritor, Youssef Nader, Julian Schilliger

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Virtual unwrapping itself is older (Seales read the En-Gedi scroll in 2015), but that worked because its ink contained metal and showed up in X-ray. Herculaneum ink is carbon-based and nearly invisible to tomography, which is why the ink signal had to be learned from texture instead of density. No Herculaneum scroll had been read by any method before this programme.

Caveats

Readings are probabilistic reconstructions from a learned ink signal, not photographs of text: characters come with varying confidence and papyrologists supply conventional editorial restoration on top. The project's papyrology team reviews prize submissions, but that team is part of the challenge rather than an outside body, and no peer-reviewed critical edition of PHerc. 1667 had appeared when this entry was written, hence author-verified rather than independent. The attribution to Philodemus rests on a title read in a different scroll.

Aalto University / Rice University
2026-06-17
Peer reviewed Search scaffold
Model
Unknown
Field
materials

Two kagome superconductors predicted by machine learning and confirmed in the lab

A screening pipeline that narrowed over 1.3 million candidate structures picked out YRu3B2 and LuRu3B2, which were then synthesized and measured to superconduct at 0.81 K and 0.95 K.

The SuperC consortium used machine-learning pre-screening to cut a very large chemical space down to a tractable candidate list, then ran first-principles calculations on what survived, arriving at 741 dynamically and thermodynamically stable compounds with DFT-predicted Tc above 5 K. Morosan's group at Rice synthesized two of them and confirmed bulk superconductivity by magnetization and specific heat measurement. Both compounds carry a kagome lattice, in which electrons form flat bands.

WithRose Albu Mustaf, Päivi Törmä, Emilia Morosan, B. Andrei Bernevig, Miguel A. L. Marques

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Machine-learning screening of superconductor candidates is an established and crowded field, and the JHU APL work on novel superconductors predates this. What is claimed as new is the end-to-end path: candidates identified from scratch by a machine-learning-guided pipeline, then synthesized and experimentally confirmed, rather than screening validated only against already-known superconductors. YRu3B2 and LuRu3B2 were not previously reported as superconductors. Whether this is the first such end-to-end confirmation is the consortium's framing and is not independently adjudicated here.

Caveats

The gap between prediction and measurement is the story the coverage mostly buries: the screen selected for DFT-predicted Tc above 5 K, and the two compounds actually measured at 0.81 K and 0.95 K, roughly a factor of five to six low. These are sub-1-kelvin superconductors, nowhere near room temperature, and the consortium's stated goal of a room-temperature superconductor by 2033 is a research programme, not a result. Two confirmed compounds out of 741 stable candidates is also not a hit rate the paper establishes as generalizable. The specific machine-learning method is not named in the coverage consulted, so model is recorded as Unknown.

Independent
2026-06-02
Peer reviewed Collaborative
Model
Claude Sonnet 4.6, Claude Opus 4.7
Field
physics
Posed
2014 · open 12 yrs

Identity for the critical exponents of jamming derived analytically

The relation a + b = 1 between critical exponents at the jamming transition, previously seen only numerically in the full replica-symmetry-breaking solution of hard spheres, is derived from the scaling equations and published after peer review.

In the full replica-symmetry-breaking solution for hard spheres, the critical exponents a and b at the jamming transition had been observed numerically to high precision to satisfy a + b = 1, without an analytic derivation from the scaling equations. Parisi and Zamponi asked Claude for help; the model quickly proposed what proved to be essentially the correct idea, combining integration-by-parts identities with a maximum principle. Its first formal write-up contained errors, which the authors then fixed and verified before publication in the Journal of Statistical Mechanics.

WithGiorgio Parisi, Francesco Zamponi

Original work
arXiv
Commentary
VibeMathed
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The numerical observation that a + b = 1 in the FullRSB jamming solution dates to the 2014 work of Charbonneau, Kurchan, Parisi, Urbani and Zamponi on the exact mean-field theory of hard-sphere glasses, where the relation was reported as observed but unexplained. It remained without analytic derivation in the subsequent literature. The derivation is new.

Caveats

Peer-reviewed and published, which is the appropriate grade for a physics result of this kind, but this is an analytic derivation within an established mean-field framework rather than an independently replicated calculation. Autonomy graded collaborative: the model supplied the key idea, but its first formal write-up contained errors that the human authors fixed and verified, and the authors framed the exchange as asking for help on a problem they posed.

Google DeepMind
2026-05-21
Formally verified Search scaffold
Model
AlphaProof Nexus (LLM + Lean)
Field
mathematics

Nine Erdős problems and 44 OEIS conjectures proved with machine-checked proofs

An agent pairing language models with the Lean proof assistant resolved 9 of 353 open Erdős problems and 44 of 492 open OEIS conjectures, with every proof mechanically verified.

The system proposes proofs with a language model and checks every step in Lean, so a hallucinated argument cannot pass. Beyond the Erdős and OEIS totals it settled a 15-year-old question on log-concavity of pure O-sequences in codimension 3 and type 2, and proved an exact O(1/t) convergence rate for anchored gradient descent-ascent via a parameter choice not previously identified. Two of the Erdős problems had been open for 56 years. All formal proofs were released in a public repository.

WithPushmeet Kohli, Swarat Chaudhuri

Challenge
none linked
Novelty check, caveats & sources
Novelty check

The Erdős problems were listed as open in the erdosproblems.com database and the OEIS conjectures as unproven at the time of the run; the paper documents which. In the course of the work the authors found misformalizations in the stated problems #125 and #741(i), which had to be corrected before resolution, a reminder that 'open' status in a database is itself fallible. Terence Tao's community wiki independently tracks AI contributions to Erdős problems and records these.

Caveats

The headline is a hit rate, not a sweep: 9 of 353 and 44 of 492. The authors state successes concentrate in combinatorics, convex optimization and number theory where Lean's library is mature, that the agent inherits its LLM's biases and shows high search variance, and that it cannot solve problems requiring substantial new theory. Formal verification guarantees the proofs are correct; it says nothing about whether the problems were deep. Autonomy is graded search-scaffold rather than autonomous: humans built the agent, chose the problem sets, and supplied the formalization targets.

Independent checks

Lean compiler via the paper's SafeVerify step: all proofs compile with no sorry and no disallowed axioms (sorryAx) · link ↗

Terence Tao's AI-contributions wiki: tracked among AI contributions to Erdős problems · link ↗

OpenAI
2026-05-20
Independently checked AI-led
Model
GPT-5 series reasoning model
Field
mathematics
Posed
1946 · open 80 yrs

Disproof of the Erdős unit-distance conjecture

A reasoning model produced the core construction disproving a 1946 conjecture in discrete geometry, finding point configurations with more unit-distance pairs than the conjecture permitted.

The model produced a construction beating the conjectured bound, with an inexplicit exponent greater than 1. Nine mathematicians - Alon, Bloom, Gowers, Litt, Sawin, Shankar, Tsimerman, Wang and Wood - published a human-verified version of the argument; Sawin separately made the exponent explicit at n^1.014. Gowers called it 'the first example of a result produced autonomously by an AI that I find exciting in itself.'

Original work
arXivarXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The unit-distance problem dates to Erdős (1946). Prior bounds were well documented; the construction is new.

Caveats

Human mathematicians shaped the problem framing and verified the construction. The explicit n^1.014 exponent is Sawin's refinement, not the model's output: the model's own bound was inexplicit. The strength of the endorsement from Gowers is notable but is a judgment, not a formal check.

University of Cambridge / DIOSynVax
2026-05-18
Peer reviewed AI-assisted
Model
Unknown
Field
medicine

Phase 1 trial of a computationally designed pan-sarbecovirus vaccine

A vaccine whose antigen was designed entirely in silico from sarbecovirus sequence data was safe in 39 healthy volunteers and raised immune responses against SARS-CoV-2, SARS, and bat viruses that have never infected humans.

Rather than target one strain, the team analysed sarbecovirus genetic data to compute a single synthetic antigen capturing structural features shared across the subgenus, on the theory that a vaccine aimed at what the family has in common survives the evolution of any one member. The candidate, pEVAC-PS, is a DNA vaccine delivered needle-free by micro fluid jet. The phase 1 dose-escalation trial ran in 39 healthy volunteers aged 18 to 50, sponsored by University Hospital Southampton NHS Foundation Trust, and reported no significant side effects.

WithAlasdair P. S. Munro, Jonathan L. Heeney, Saul N. Faust

Announced
Cambridge
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Computationally designed and consensus or mosaic antigens have a long history in HIV and influenza vaccine work, and broad sarbecovirus vaccine candidates were pursued by several groups after 2020. What is new here is a first-in-human safety and immunogenicity readout for an antigen with no natural sequence, generated computationally rather than selected from circulating strains. Searched PubMed for prior phase 1 results on pan-sarbecovirus immunogens; earlier candidates in this class were preclinical at the time of publication.

Caveats

Phase 1: safety, tolerability and immunogenicity only. No efficacy claim, and immune response is not protection. n=39, healthy adults 18 to 50, so nothing follows about older or immunocompromised populations. Approval is years away and a phase 2 is planned. The autonomy grade is deliberately conservative because the public record is thin on what the computational method actually was: the reporting says artificial intelligence and machine learning without naming a model, architecture or developer, so model is recorded as Unknown rather than inferred. The bat-virus immunogenicity result is a laboratory assay against viruses that have not infected humans, not evidence of protection against a future outbreak.

Google DeepMind
2026-01-15
Independently checked Search scaffold
Model
AlphaEvolve (Gemini-based)
Field
mathematics

Structure in Bruhat intervals of permutation groups

AlphaEvolve identified unexpected special structure in Bruhat intervals for particular permutation groups.

Part of a continuing series of AlphaEvolve results in combinatorics. The system evolves candidate programs under a human-specified evaluation function, so the search is automated but the problem framing is human.

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Documented as a new observation in the Quanta coverage; the specific structure was not previously catalogued.

Caveats

The paper is an arXiv preprint, not yet peer reviewed, so this entry was downgraded from 'peer reviewed' once the artifact was located. The five authors are academic mathematicians unaffiliated with the announcing lab, and they turned AlphaEvolve's suggested permutations into proved theorems, which is what the 'independently checked' grade rests on.

OpenAI / Harmonic
2026-01-13
Formally verified Autonomous
Model
GPT-5.2 Pro + Aristotle
Field
mathematics

Erdős problem #728 resolved and formalized in Lean

GPT-5.2 produced a proof of a previously open Erdős problem; Harmonic's Aristotle formalized it in Lean, making it the first Erdős problem regarded as fully resolved by AI with machine-checked proof.

The problem concerns the existence of infinitely many integers a, b, n satisfying divisibility and inequality conditions involving factorials. GPT-5.2 generated the argument, Aristotle formalized it in Lean, and a human-readable writeup was posted to arXiv. Modifications of the same argument also resolved problems #729 and #401.

Original work
arXivGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Recorded as open in the Erdős problems database prior to resolution. Terence Tao's AI-contributions wiki tracked the resolution and did not find the result in existing literature.

Caveats

Tao cautioned that the win 'says more about speed than difficulty': the problem was open but not considered deep. Lean formalization makes correctness essentially certain; significance is the contested part, not validity.

Independent checks

Lean kernel (machine-checked): verified · link ↗

Terence Tao: accepted · link ↗

OpenAI
2025-11-20
Peer reviewed AI-assisted
Model
GPT-5
Field
computer-science

Early science acceleration experiments with GPT-5

A multi-domain study documenting cases where GPT-5 contributed to research progress across mathematics, physics, biology and materials science.

A collection of case studies rather than a single result. Useful as a source of individual entries; each contained claim needs separate grading before it belongs in the registry on its own.

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Not applicable. This is a meta-report, not a single claim.

Caveats

Lab-authored evaluation of its own model. Individual case studies vary widely in strength and several have not been independently checked.

Google DeepMind
2025-11-03
Independently checked Search scaffold
Model
AlphaEvolve (Gemini-based)
Field
mathematics

AlphaEvolve across 67 problems: 20 improvements, 8 regressions

A systematic study with Terence Tao of an evolutionary coding agent on 67 open problems in analysis, combinatorics and geometry, reporting where it beat the literature and where it did not.

Rather than a single headline result, this is the systematic version: 67 problems attempted, with outcomes recorded either way. Tao reports roughly 20 cases of a new result, most others matching known bounds, and 8 where the tool did worse than the literature. Genuine advances include new three-dimensional constructions for finite-field Nikodym sets, which then prompted a human-refined hybrid construction beating both algebraic and random baselines, and an improved upper bound for Sidon sets, from 1.96365 to about 1.9526. On the well-known conjectures of Sidorenko, Sendov and Crouzeix it found no counterexample, which Tao notes may simply reflect that they are true.

WithTerence Tao, Javier Gómez-Serrano, Bogdan Georgiev, Adam Zsolt Wagner

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Prior bounds are cited per problem in the 81-page paper, with data and prompts released publicly. Tao flags the central novelty risk himself: for well-known problems the tool often proposed the optimal construction immediately, suggesting recall from training data rather than search, and one supposedly new four-dimensional Kakeya construction turned out to essentially match a prior Bukh–Chao paper. Entries in this registry drawn from the earlier AlphaEvolve release (kissing number, minimum overlap, matrix multiplication) are tracked separately.

Caveats

Tao's own framing is that the contribution is scale and adaptability, not a fundamental breakthrough, and he describes the LLM as supplying 'educated randomness' inside an evolutionary loop rather than doing symbolic reasoning. Substantial human expertise was required to design non-exploitable verifiers: the system reliably finds loopholes, such as placing points at nearly identical locations to satisfy a distance tolerance. Included here because the negative and null results are reported alongside the wins, which is rare and is what makes it citable.

Independent checks

Terence Tao (co-author, published assessment): reports ~20 new results, 8 worse than literature; cautions on training-data contamination · link ↗

OpenAI
2025-10-19
Already known Retrieval
Model
GPT-5
Field
mathematics

GPT-5 "solved 10 Erdős problems": it located existing solutions

An OpenAI executive announced that GPT-5 had solved 10 previously open Erdős problems and made progress on 11 more; the problems were only 'open' in the sense that one database maintainer was unaware of the already-published solutions the model surfaced.

OpenAI VP Kevin Weil posted that 'GPT-5 found solutions to 10 (!) previously unsolved Erdős problems.' Thomas Bloom, who maintains erdosproblems.com, replied that listing a problem as open there only means he personally was unaware of a solution, and that GPT-5 had found existing papers containing the solutions rather than proving anything new. The post was deleted; Demis Hassabis called the episode 'embarrassing' and Yann LeCun mocked it.

Novelty check, caveats & sources
Novelty check

By definition, that is the whole point of the episode. The solutions existed in the published literature; GPT-5's contribution was locating them, a genuinely useful literature-search result that was misdescribed as original problem-solving.

Caveats

Retained as a cautionary entry and a direct companion to the GPT-5 convex-optimization case. The underlying literature search was real and valuable; only the 'solved previously unsolved problems' framing was false. This is distinct from the later, genuine GPT-5.2 + Aristotle resolution of Erdős #728.

Independent checks

Thomas Bloom (erdosproblems.com maintainer): the problems were already solved in the literature; no new proofs · link ↗

UT Austin / CWI Amsterdam
2025-09-25
Author verified AI-assisted
Model
GPT-5 Thinking
Field
computer-science

Limits to black-box amplification in QMA, with the key step written by GPT-5

A quantum complexity paper proving black-box error amplification for QMA cannot push completeness closer to certainty than doubly exponentially, in which the pivotal technical step was produced by GPT-5 in about half an hour.

The authors were stuck on how the largest eigenvalue of a matrix behaved as a parameter varied. GPT-5 proposed reformulating the quantity so that a complex approximation-theory bound applied, and that reformulation became the technical core of the oracle separation. Aaronson wrote that the step would have cost him or a graduate student a couple of weeks and that it was the first substantive AI contribution to a paper of his. The result makes his 2008 oracle separation quantitative and shows recent amplification results are optimal for black-box procedures.

WithScott Aaronson, Freek Witteveen

Original work
arXiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The theorem extends Aaronson's 2008 QMA oracle separation and a 2025 result of Jeffery and Witteveen; no prior bound of this form on black-box amplification appears in the literature. The AI contribution is documented in the paper's acknowledgements and on Aaronson's own blog, not inferred from press coverage. That matters, because most "AI proved a theorem" stories are not.

Caveats

An arXiv preprint. The contribution is one lemma inside a human-conceived and human-written paper; Aaronson is explicit that the model did not pose the problem or design the argument and that he could have done the step himself given time. Autonomy is ai-assisted, which stays the right grade even though the step was pivotal.

Google DeepMind (with Brown, NYU and Stanford)
2025-09-17
Author verified Search scaffold
Model
Physics-informed neural networks with Gauss–Newton optimisation
Field
mathematics

New families of unstable singularities in fluid equations

Physics-informed neural networks found previously unknown unstable self-similar blow-up solutions for three fluid equations, computed to near machine precision, and revealed an unexpected near-linear relation between a solution's instability order and its blow-up rate.

A physics-informed neural network minimises the residual of the differential equation itself rather than fitting data, which lets it hunt for self-similar blow-up profiles that ordinary simulation cannot hold onto because they are repelling. The team reports a stable and three unstable singularities for the incompressible porous media equation, a stable and three unstable solutions plus a candidate fourth for 2D Boussinesq, and an unstable singularity for 3D Euler with boundary. Residuals were driven near machine precision, the accuracy level a later computer-assisted proof would need. Across equations the blow-up rate depends almost linearly on the order of instability, a pattern nobody had predicted.

WithJavier Gómez-Serrano, Tristan Buckmaster, Ching-Yao Lai, Yongji Wang

Original work
arXiv
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Stable self-similar blow-up for these equations was already known, including Chen and Hou's computer-assisted proof for 3D Euler with boundary. Unstable singularities had been found only in isolated cases because they are numerically repelling; no systematic families existed in the literature. This is the first paper to produce several unstable families across three different equations by one method.

Caveats

An arXiv preprint, not peer reviewed at the time of this entry. These are high-precision numerical solutions, not theorems: converting them into proofs requires a computer-assisted verification step the paper leaves to future work. No singularity for 3D Navier–Stokes, the Millennium Prize problem, is claimed, and the equations studied are not that one. Autonomy is search-scaffold: the equations, the self-similar ansatz and the loss function are human-designed and the network is the solver.

Arc Institute / Stanford University
2025-09-17
Peer reviewed AI-led
Model
Evo 1 and Evo 2
Field
biology

AI-generated bacteriophage genomes that replicate and kill bacteria

Genome language models wrote complete ΦX174-like bacteriophage genomes from scratch; of 285 designs that could be built, 16 produced viable infectious phages, several with faster lysis than the natural virus.

The team fine-tuned Evo 1 and Evo 2 on a cleaned set of nearly 15,000 Microviridae genomes, generated 302 candidate genomes with ΦX174-like architecture, chemically synthesised the 285 that could be assembled, and tested them for plaque formation in E. coli. Sixteen were viable, with substantial sequence divergence from the template. Several generated phages outcompeted wild-type ΦX174 in growth and lysis kinetics, and a cocktail of them cleared three ΦX174-resistant E. coli strains. The authors frame it as the first generative design of a complete functional genome.

WithSamuel King, Brian Hie

Commentary
Science
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Synthetic phage genomes assembled from natural sequence date to Venter's 2003 ΦX174 reconstruction, and generative design of individual proteins was well established. Generating an entire genome that yields a working organism had not been demonstrated. The claim is bounded to bacteriophages, which infect bacteria only.

Caveats

Peer review does not replicate: no independent group has rebuilt these genomes, and the 16 viable designs remain the authors' own result. The designs are variants within one small, exceptionally well-characterised template (ΦX174, 5,386 bp), not free-form genomes, and 269 of the 285 that were built did not work. On biosafety, the authors excluded eukaryotic and human-infecting viruses from training and confine the work to phages; the result has nonetheless become a reference point in biosecurity debate, which is context this entry records rather than adjudicates. Science published a biosecurity perspective alongside the paper arguing for legally required screening of synthetic DNA orders. Note that some press coverage of the Science paper misattributes the work to OpenAI; the collaboration is Arc Institute and Stanford.

Math, Inc.
2025-09-11
Author verified AI-assisted
Model
Gauss (autoformalization agent)
Field
mathematics

Strong prime number theorem formalized in Lean by an autoformalization agent

An agent completed a Lean formalization of the strong prime number theorem in about three weeks, finishing a project two expert mathematicians had left blocked after 18 months.

Tao and Kontorovich began a Lean formalization of the strong prime number theorem (the version with an explicit error term) in January 2024, and by July 2025 had announced intermediate progress blocked on core difficulties in complex analysis. Math, Inc. reports that Gauss produced over 25,000 lines of Lean and roughly 1,100 theorems and definitions in about three weeks, completing the project. The repository README states that most statements and proofs were produced by the agent, with humans supplying the high-level blueprint, reviewing key lemmas, and adapting prior work.

WithTerence Tao, Alex Kontorovich

Original work
GitHub
Announced
Math Inc.
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The prime number theorem is not new mathematics; the contribution is formalization. What was open was whether this particular formalization could be completed, and the surrounding claim is about speed. Recorded because formalization at this scale is a distinct kind of contribution and because the AI finished work humans had started and stalled on, rather than producing a new theorem. Prior art is the PrimeNumberTheoremAnd project itself, which this build reuses and which the README credits.

Caveats

This is a formalization result, not a mathematical discovery: no new theorem was proved. The comparison to '18+ months' is not like-for-like: Tao and Kontorovich worked intermittently, and the repository explicitly reuses definitions and some proofs from their earlier PrimeNumberTheoremAnd project. Math, Inc.'s announcement states the agent 'relies on natural language scaffolding supplied by human mathematicians, and requires high-level expert guidance'; the proportion of human effort is not quantified, and neither Tao nor Kontorovich is quoted in it. Graded author-verified rather than formal: the artifact is public and Lean-checkable, but the README does not itself assert a sorry-free build and no independent audit of that is on the public record. Vendor-announced, so the framing carries the usual caution.

Independent checks

Public Lean artifact (not independently audited): repository public and machine-checkable; README reports the development finished · link ↗

OpenAI
2025-08-01
Already known AI-assisted
Model
GPT-5 Pro
Field
mathematics

Improved step-size bound in smooth convex optimization

GPT-5 Pro extended a guaranteed-convexity window for gradient descent from η ≤ 1/L to η ≤ 1.5/L, but the optimal 1.75/L bound had already been published months earlier.

Bubeck posed an open problem from a convex optimization paper. After about 17 minutes of reasoning the model produced an improved bound using Bregman divergence inequalities and cocoercivity. Bubeck verified the proof as correct and described it as new mathematics.

WithSébastien Bubeck

Novelty check, caveats & sources
Novelty check

Version 2 of the source paper, published 2 April 2025, had already established the optimal 1.75/L bound, strictly stronger than the model's 1.5/L. The model was working from an earlier version and its result was superseded before it was produced.

Caveats

The proof itself is valid; the novelty claim is not. Analysts noted the argument largely recombined known techniques in different notation. Retained as a cautionary entry: this is the single most common failure mode in this space, and the reason every entry carries a novelty check.

Harmonic
2025-07-28
Formally verified AI-led
Model
Aristotle
Field
mathematics

Machine-checked Lean proofs for five of six 2025 IMO problems

Harmonic's Aristotle produced formally verified Lean 4 proofs for five of the six 2025 IMO problems, so the gold-medal-standard score rests on a compiler rather than on human graders.

Aristotle drafts an informal proof, breaks it into intermediate lemmas, formalises each in Lean 4, and iterates against the compiler's feedback, which lets it use the large natural-language corpus of mathematics while still emitting machine-checked output. Harmonic published the Lean artifacts for the 2025 problems in a public repository. The announcement came a week after the natural-language gold results from Google DeepMind and OpenAI.

Original work
arXivGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Formal IMO results are not new (AlphaProof solved three 2024 problems in Lean), but a gold-medal-level score with a formal proof for every claimed solution had not been reported. The problems have published official solutions, so this is a capability milestone rather than new mathematics.

Caveats

Problems had to be stated formally in Lean before the system could attempt them, the same human step that capped AlphaProof's autonomy grade. Answer-construction problems require the answer to appear in the formal statement, a known weakness of formal olympiad evaluation. No IMO-certified grading and no independently enforced contest time limit. The `formal` grade covers the proofs Aristotle produced, not the claim that this equals a human gold medal.

Google DeepMind
2025-07-21
Independently checked AI-led
Model
Gemini Deep Think (advanced version)
Field
mathematics

Gold-medal standard at the 2025 International Mathematical Olympiad

An advanced version of Gemini Deep Think read the official problem statements and wrote natural-language proofs for five of the six 2025 IMO problems inside the contest time limits, scoring 35 of 42, graded and certified by the official IMO coordinators.

The model worked end to end in natural language within the two 4.5-hour sessions, with no human translating problems into a formal language, and solved problems one through five. Google DeepMind was in the first cohort whose results were officially graded and certified by IMO coordinators against the rubric used for student scripts; 35 points is the gold-medal threshold. OpenAI announced the same 35/42 score two days earlier with an experimental reasoning model, graded internally by three former IMO medallists rather than by the IMO.

Original work
DeepMind
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

IMO problems have published official solutions, so this is a capability milestone rather than a new mathematical result, and the registry records it as one. The prior best was AlphaProof and AlphaGeometry 2's 28/42 in 2024, which required humans to hand-translate each problem into Lean first; removing that step is the substantive change.

Caveats

Benchmark performance on problems with known answers, not a discovery. Google's own account describes training on a curated corpus of high-quality solutions and providing general guidance on how to approach IMO problems, which is human mathematical input beyond posing the problem, hence ai-led rather than autonomous. Problem six went unsolved. The parallel OpenAI claim is self-graded and not IMO-certified, and drew criticism for being published before the closing ceremony, which organisers had asked AI developers to wait for.

Independent checks

Official IMO coordinators (certified grading): 35/42, gold-medal standard · link ↗

Insilico Medicine
2025-06-03
Peer reviewed Search scaffold
Model
PandaOmics (target) + Chemistry42 (molecule)
Field
medicine

Phase 2a results for a drug whose target and molecule both came from AI

Rentosertib, a TNIK inhibitor whose target was picked by one AI system and whose molecule was generated by another, improved forced vital capacity by 98.4 mL at the top dose against a 20.3 mL decline on placebo in a 71-patient randomised trial.

TNIK was prioritised as a fibrosis target by PandaOmics from omics and literature data; the inhibitor was generated and optimised by Chemistry42. GENESIS-IPF was a 12-week double-blind placebo-controlled Phase 2a across 22 sites in China, randomising 71 idiopathic pulmonary fibrosis patients across placebo and three dose arms. The 60 mg once-daily arm showed the largest lung-function improvement, and the drug was generally well tolerated. Published in Nature Medicine.

Original work
Nature
Announced
Insilico
Challenge
none linked
Novelty check, caveats & sources
Novelty check

AI-generated molecules had reached the clinic before, and AI-nominated targets had been published before, but this is the first randomised placebo-controlled efficacy signal for a compound where both the target and the molecule came from generative systems. TNIK's role in fibrosis was not an established target hypothesis when it was nominated.

Caveats

Phase 2a is small, short and exploratory: 71 patients over 12 weeks, single-country sites, not powered for efficacy, with FVC change over 12 weeks not a registrational endpoint. The sponsor designed, ran, analysed and published the trial. "AI-discovered" covers target selection and generative chemistry; the medicinal chemistry, preclinical work and trial conduct were conventional. A Phase 3 was subsequently initiated, which does not retrospectively strengthen this readout.

Google DeepMind
2025-05-14
Independently checked Search scaffold
Model
AlphaEvolve (Gemini-based)
Field
computer-science
Posed
1969 · open 56 yrs

4×4 complex matrix multiplication in 48 scalar multiplications

AlphaEvolve found a scheme multiplying 4×4 complex matrices with 48 scalar multiplications, improving on Strassen's 49 from 1969.

Strassen's 1969 algorithm had stood at 49 multiplications for 56 years in this setting. AlphaEvolve was not purpose-built for matrix multiplication; it is a general evolutionary coding agent. The scheme is exactly checkable.

Original work
DeepMind
Announced
DeepMind
Commentary
Wikipedia
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Strassen (1969) and subsequent literature exhaustively catalogued; 49 was the standing record for complex-valued 4×4. Confirmed new.

Caveats

Applies to complex-valued matrices specifically. Practical speedup is limited; the significance is theoretical. Related follow-up work by others has explored rank-23 schemes for 3×3.

Independent commentary
Google DeepMind
2025-05-14
Independently checked Search scaffold
Model
AlphaEvolve (Gemini-based)
Field
mathematics
Posed
1694 · open 331 yrs

Improved lower bound for the 11-dimensional kissing number

AlphaEvolve improved the best known configuration for the kissing number problem in 11 dimensions.

Part of a sweep across 50+ open problems in analysis, geometry, combinatorics and number theory. AlphaEvolve rediscovered state-of-the-art solutions in roughly 75% of cases and improved on the best known in about 20%. The kissing number problem asks how many non-overlapping unit spheres can touch a central one.

Original work
DeepMind
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Kissing number bounds are well catalogued per dimension; the 11D improvement was checked against the standing record.

Caveats

A lower-bound improvement via explicit construction, not a resolution of the problem. The 20% improvement rate means most of the 50+ problems saw no advance.

Google DeepMind
2025-05-14
Independently checked Search scaffold
Model
AlphaEvolve (Gemini-based)
Field
mathematics
Posed
1955 · open 70 yrs

Improved bound for the Erdős minimum-overlap problem

AlphaEvolve nudged the best known bound for Erdős's minimum-overlap constant, the first improvement since 2016, and sharpened several autocorrelation inequalities.

As part of a sweep across open problems in mathematical analysis, AlphaEvolve improved the upper bound on the Erdős minimum-overlap constant from about 0.380927 to about 0.380924, the first movement since 2016, and improved constants in autocorrelation inequalities. The gains are numerically small but exceed long-standing records, and the constructions are explicit and checkable.

Original work
DeepMind
Announced
DeepMind
Commentary
Wikipedia
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The minimum-overlap constant and the autocorrelation inequalities have well-tracked records. AlphaEvolve's values were compared against the standing bounds and confirmed to improve them.

Caveats

The improvements are marginal in magnitude, and both were improved again by other automated systems in 2026. As with the other AlphaEvolve results, a human-designed evaluator and search loop did the selecting.

FutureHouse
2025-05-01
Peer reviewed AI-led
Model
Robin (multi-agent)
Field
biology

Candidate treatment for dry age-related macular degeneration

The Robin system automated hypothesis generation, experiment design and data analysis, identifying a novel candidate treatment for dry AMD.

Presented as an end-to-end automated discovery loop with humans executing the physical experiments. Robin proposed enhancing retinal pigment epithelium phagocytosis as a therapeutic strategy, then identified ripasudil and KL001 as candidates and confirmed in vitro efficacy for both. It went on to propose and analyse a follow-up RNA sequencing experiment, which showed upregulation of ABCA1, a lipid efflux pump and a possible novel target in its own right. Published in Nature in May 2026.

Original work
Nature
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Ripasudil is a clinically used Rho kinase inhibitor; the paper states that to the authors' knowledge it had never previously been proposed for the treatment of dry age-related macular degeneration, and the peer-reviewed version carries that claim. The ABCA1 mechanism emerged from the system's own follow-up experiment rather than from the prior literature. What is new is the repurposing hypothesis and its in vitro support, not a new compound.

Caveats

Peer reviewed, which is the right grade for an empirical claim of this kind, but in vitro efficacy is a long way from a treatment: a candidate is many years and several trial phases from the clinic. Verification of the biology is the journal's, not an independent replication. Autonomy stays ai-led rather than autonomous: the paper describes the approach as semi-autonomous, and humans executed every physical experiment, though all hypotheses, experimental directions, data analyses and main-text figures were produced by Robin.

Sakana AI
2025-03-12
Author verified AI-led
Model
The AI Scientist-v2
Field
computer-science

A fully machine-generated paper passed workshop peer review

One of three end-to-end AI-generated manuscripts submitted to an ICLR 2025 workshop averaged 6.33 from reviewers and would have been accepted; the authors withdrew it, and the bar cleared was a workshop track, not the main conference.

The AI Scientist-v2 generates a hypothesis, writes and runs the experiments through agentic tree search, and produces the manuscript, figures and related work without human editing. Sakana submitted three such papers to the ICLR 2025 workshop "I Can't Believe It's Not Better", with the knowledge and cooperation of the workshop organisers and ICLR leadership, and agreed in advance to withdraw anything accepted. One paper, on whether an explicit compositional regularisation term improves compositional generalisation, scored 6.33 and cleared the bar.

Original work
GitHub
Announced
Sakana AI
Challenge
TechCrunch
Novelty check, caveats & sources
Novelty check

Machine-written text had passed review before in the form of nonsense-generator stings (SCIgen, 2005), which tested review rather than produced research. The claim here is different: an autonomous pipeline produced a real experimental paper that reviewers rated acceptable. Sakana's own write-up is the source for the process, and the code and submitted manuscripts are public.

Caveats

Workshop tracks accept a far higher share of submissions than the main conference, and Sakana says so itself; the company also notes the accepted paper contains a citation error and that its own reviewers judged the work below the main-conference bar. The paper's scientific content is a negative result: the proposed regularisation did not help. Reviewers were not told which submissions were machine-generated, but the experiment ran with organiser consent, so this is not a blind test of peer review at scale. Humans chose the venue and selected which generated papers to submit, which caps autonomy at ai-led.

Google
2025-02-19
Peer reviewed AI-assisted
Model
AI Co-Scientist (Gemini 2.0 multi-agent)
Field
biology

AI co-scientist hypotheses on antimicrobial resistance and liver fibrosis

A multi-agent system generated hypotheses that were subsequently validated experimentally in the lab.

The system proposed mechanisms in antimicrobial resistance and liver fibrosis that wet-lab work then supported. Notable as one of the earliest cases of an AI-generated hypothesis surviving experimental test rather than just sounding plausible. The peer-reviewed version, published in Nature in May 2026, covers three biomedical applications together: drug repurposing, novel-target discovery and explaining mechanisms of antimicrobial resistance. It adds drug-repurposing candidates and synergistic combination therapies for acute myeloid leukaemia that were then validated in vitro.

Original work
NaturebioRxiv
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The AMR mechanism proposed had reportedly been independently arrived at by a human group whose work was unpublished at the time, which complicates the novelty claim.

Caveats

Hypothesis generation with human-run validation, not autonomous discovery. Peer review confirms the work, not that every hypothesis in it was unprecedented, and the novelty question above is untouched by publication. Independent replication outside the collaborating labs is limited. Graded conservatively.

Institute for Protein Design, University of Washington
2025-02-13
Peer reviewed Search scaffold
Model
RFdiffusion with PLACER ensemble scoring
Field
chemistry

Enzymes with working catalytic machinery designed from scratch

Diffusion-based protein design produced serine hydrolases with catalytic efficiencies up to 2.2 × 10⁵ M⁻¹s⁻¹ on five folds unlike any natural serine hydrolase, with crystal structures matching the design models to under 1 Å.

An enzyme needs more than the right shape: its active site has to stay organised at every step of the reaction, which is where designed enzymes have historically failed. The team generated backbones with RFdiffusion from a minimal active-site description, then used an ensemble-generation network to check that the catalytic geometry held along the whole reaction coordinate, keeping only designs that passed. The resulting hydrolases approach natural catalytic efficiency on folds not found among natural serine hydrolases, and the crystal structures match the designs closely. Published in Science.

WithAnna Lauko, David Baker

Original work
Science
Announced
Baker Lab
Challenge
none linked
Novelty check, caveats & sources
Novelty check

De novo enzyme design dates to the 2000s (Kemp eliminases, retro-aldolases), but designed enzymes were typically orders of magnitude slower than natural ones and needed directed evolution to become useful. The advance is reaching this efficiency with multi-step catalytic machinery by design rather than by evolution.

Caveats

Serine hydrolases are a comparatively tractable target with a well-understood catalytic triad; the result does not generalise automatically to harder chemistry. Efficiencies remain below the best natural hydrolases. The per-round success rate is low and the paper reports the screening funnel. The reaction, the active-site description and the assays were human-specified, which is why autonomy is search-scaffold.

EvolutionaryScale
2025-01-16
Peer reviewed AI-led
Model
ESM3 (98B)
Field
biology

A working fluorescent protein generated by a language model

ESM3 designed esmGFP, a bright green fluorescent protein sharing only 58% of its sequence with the closest known natural fluorescent protein, a distance the authors estimate at over 500 million years of natural evolution.

ESM3 is a masked language model over protein sequence, structure and function jointly. Prompted with the residues that form the GFP chromophore and little else, and run through iterative generation, it produced candidates that were synthesised and assayed in the lab. esmGFP fluoresces at brightness comparable to natural GFPs while differing at 96 of 229 positions from its nearest natural relative. Published in Science.

WithAlexander Rives, Tom Sercu

Original work
Science
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Engineered GFP variants are routine but sit a handful of mutations from a natural template, and directed evolution moves in small steps. A functional fluorescent protein this far from any known sequence had not been reported, and the paper includes the expression, spectra and brightness measurements rather than a prediction alone.

Caveats

The 500-million-year figure is an inference from a sequence-divergence-to-time calibration, not a measurement, and has been criticised as a marketing frame. esmGFP was one hit from a generated and screened set; the paper reports the funnel. The design goal and the chromophore constraint were supplied by humans. No independent re-synthesis of esmGFP by an unaffiliated lab was on record when this entry was written.

Microsoft Research
2025-01-16
Disputed Search scaffold
Model
MatterGen
Field
materials

Generative model designs crystals to order; its flagship synthesis turned out to be a known compound

MatterGen generates crystal structures conditioned on target properties, and its headline experimental validation was later shown to match a compound reported in 1972 that sits in the model's own training data.

MatterGen is a diffusion model over crystal structures (atom types, coordinates and lattice jointly), trained on about 608,000 stable materials from the Materials Project and Alexandria, and fine-tunable to condition on a target property. Conditioned on a bulk modulus of 200 GPa it proposed a structure that collaborators at the Shenzhen Institutes of Advanced Technology synthesised, measuring 169 GPa, within 20% of the specification. Published in Nature.

Original work
NatureGitHub
Announced
Microsoft
Novelty check, caveats & sources
Novelty check

Generative models for crystals existed (CDVAE, and screening pipelines like GNoME); the contribution claimed is property-conditioned generation validated experimentally. The novelty question is precisely what is disputed below: whether the synthesised compound was new.

Caveats

A published critique in Materials Horizons argues the synthesised disordered phase Ta₁⁄₃Cr₂⁄₃O₂ is the same material as Ta₁⁄₂Cr₁⁄₂O₂ reported in 1972, that this compound is present in MatterGen's training set, and that its composition differs from the reported TaCr₂O₆, i.e. the flagship validation recovered a known material rather than a new one. The generative method itself is not refuted; the disputed part is the experimental novelty claim, and the entry is graded on the strength of that published objection. This registry keeps entries like this one rather than removing them.

Independent checks

Materials Horizons: Continued challenges in high-throughput materials predictions: disputed the novelty of the synthesised compound; identified it as present in the training data · link ↗

Google DeepMind / Google Quantum AI
2024-11-20
Peer reviewed AI-led
Model
AlphaQubit
Field
physics

A neural decoder that identifies quantum errors more accurately than hand-designed methods

A transformer trained on Google's Sycamore surface-code data cut decoding errors by 6% against tensor-network decoding and 30% against correlated matching, the best reported accuracy at the time.

A quantum error-correcting code produces a stream of syndrome measurements; a decoder has to infer from them which errors actually occurred. Hand-designed decoders assume a simplified noise model, whereas AlphaQubit is a transformer trained first on simulated data and then fine-tuned on hundreds of millions of real syndrome samples from Sycamore, so it can learn the device's actual correlated noise, leakage and crosstalk. It beat both the accuracy-oriented tensor-network decoder and the fast matching decoder, and held up in simulation to distance-11 codes. Published in Nature.

Original work
Nature
Announced
Google
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Machine-learning decoders had been proposed for years but had not beaten the best classical decoders on real hardware data at scale. The result is a decoder that does, on a specific processor.

Caveats

Accuracy, not speed: AlphaQubit is far slower than matching decoders and does not run inside the real-time budget a fault-tolerant machine needs, which the paper states. It is trained per device on that device's data, so it does not transfer for free. Scaling behaviour beyond the tested code distances is open. This improves a component of error correction; it does not itself demonstrate a fault-tolerant computation.

FlyWire Consortium (Princeton, MRC LMB, Cambridge, Vermont)
2024-10-02
Peer reviewed AI-assisted
Model
Automated electron-microscopy segmentation networks
Field
neuroscience

Complete wiring diagram of an adult fruit-fly brain

Machine segmentation of an electron-microscopy volume, corrected by a large community of proofreaders, produced the first synapse-level map of an entire adult brain: 139,255 neurons and roughly 50 million synapses.

A female Drosophila brain was sliced into about 7,000 sections and imaged by electron microscopy, producing a volume no team could trace by hand. Neural networks segmented the neurons automatically; more than 200 labs and a crowd of proofreaders then corrected the segmentation over several years and annotated over 8,000 cell types. Simulations built directly on the resulting wiring diagram predicted which neurons drive feeding and grooming, and the predictions held up experimentally. Published as a Nature package in October 2024.

WithSebastian Seung, Mala Murthy, Gregory Jefferis

Original work
NatureFlyWire
Announced
NIH
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Complete connectomes existed for C. elegans (302 neurons, 1986) and the Drosophila larva (3,016 neurons, 2023). This one is roughly 45 times larger than the previous largest complete connectome, and automated segmentation is what made that scale reachable at all.

Caveats

Automated segmentation alone was not accurate enough: the result required years of human proofreading, which is why autonomy is graded ai-assisted rather than ai-led. Neurotransmitter identity is predicted rather than measured for most connections, and gap junctions and neuromodulation are not fully captured. It is one individual's brain, and a wiring diagram is not by itself an explanation of behaviour.

Google DeepMind
2024-07-25
Independently checked AI-led
Model
AlphaProof + AlphaGeometry 2
Field
mathematics

Silver-medal standard at the 2024 International Mathematical Olympiad

AlphaProof and AlphaGeometry 2 together solved four of the six 2024 IMO problems for 28 of 42 points (one short of the gold threshold), with the algebra and number-theory solutions produced and checked in the Lean proof assistant.

AlphaProof, a reinforcement-learning system that works inside Lean, solved two algebra problems and one number-theory problem (including P6, the competition's hardest, which few human contestants solved), while AlphaGeometry 2 solved the geometry problem in seconds. The two combinatorics problems went unsolved. The Lean-based solutions are machine-checked by construction; the full performance was graded by mathematicians Timothy Gowers and Joseph Myers under competition-style marking.

Original work
Nature
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

These are competition problems with published official solutions, so the achievement is a capability milestone (solving hard, known-answer problems under near-competition conditions) rather than a new mathematical result. The registry records it as such.

Caveats

Problems were hand-translated into formal Lean statements by people before AlphaProof attempted them, a real human contribution beyond posing the question, so autonomy is graded conservatively. AlphaProof also took far longer than the human time limit on some problems. This is benchmark performance on solved problems, not a discovery.

Independent checks

Timothy Gowers & Joseph Myers (competition-style grading): 28/42, silver-medal standard · link ↗

Video explainers
AlphaProof and AlphaGeometry 2 achieve a silver-medal score at the IMO, explainedElvis Saravia (DAIR.AI) YouTube ↗

Nothing loads from YouTube until you press play.

Google DeepMind / Isomorphic Labs
2024-05-08
Peer reviewed AI-led
Model
AlphaFold 3
Field
biology

Joint structure prediction for proteins, nucleic acids and ligands (AlphaFold 3)

A diffusion-based successor to AlphaFold predicts complexes spanning proteins, DNA, RNA, small molecules and ions in a single model, with large accuracy gains over the specialised docking tools used for each case.

AlphaFold 2 predicted the fold of a protein chain; drug discovery mostly needs the structure of that protein bound to something else. AlphaFold 3 replaces the structure module with a diffusion model operating directly on atom coordinates, which lets one network cover protein–ligand, protein–nucleic-acid and antibody–antigen complexes. Reported accuracy substantially exceeds specialised docking tools on protein–ligand benchmarks. Published in Nature.

WithJohn Jumper, Max Jaderberg

Original work
NatureGitHub
Announced
Google
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Docking and complex prediction had decades of specialised tools (AutoDock, RoseTTAFold All-Atom, and AlphaFold-Multimer for protein assemblies). The contribution is one model spanning molecule types at higher accuracy, not a new biological result; the structures it predicts are natural ones.

Caveats

At publication neither code nor weights were released, only a rate-limited web server, which drew an open letter from over a thousand researchers about reproducibility; weights for academic use followed in November 2024. Accuracy on antibody–antigen complexes and some nucleic-acid targets is far below the headline figures, and the model can produce confidently wrong hallucinated structure in disordered regions. Predictions are hypotheses, not measurements.

Princeton University / PPPL / DIII-D National Fusion Facility
2024-02-21
Peer reviewed Search scaffold
Model
Deep reinforcement learning controller over a learned plasma model
Field
physics

Reinforcement learning steers a tokamak away from tearing instabilities

A controller trained on past DIII-D shots forecast tearing-mode instabilities up to 300 ms ahead and adjusted the plasma in real time to avoid them, holding high-performance conditions that would otherwise have collapsed.

Tearing modes break up the magnetic surfaces confining a fusion plasma and can end a discharge. Existing approaches suppress them once they appear; this one predicts them. A network trained on past DIII-D discharges estimates the instability probability, and a reinforcement-learning policy trained against a learned plasma model adjusts beam power and plasma shape to keep that probability low while holding pressure high. It was tested live on DIII-D, not only in simulation. Published in Nature.

WithJaemin Seo, Egemen Kolemen

Original work
Nature
Announced
Princeton
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Distinct from the 2022 DeepMind–TCV work already in this registry, which learned magnetic shape control rather than instability avoidance. Tearing-mode prediction from machine learning had been published before; closing the loop so the controller acts on the prediction during a live shot is the new part.

Caveats

Demonstrated on one machine in specific scenarios. A data-driven controller trained on DIII-D discharges has no guarantee of transferring to another tokamak, still less to ITER-scale burning plasma where the training data does not exist. Tearing modes are one disruption pathway among several. The physics model, actuators and safety limits are human-designed; the policy operates inside them.

Google DeepMind
2024-01-17
Peer reviewed Search scaffold
Model
AlphaGeometry
Field
mathematics

Olympiad geometry solved without human demonstrations

A language model trained purely on synthetic data, paired with a symbolic deduction engine, solved 25 of 30 olympiad geometry problems against 10 for the previous best automated system and 25.9 for the average human gold medallist.

Geometry proofs stall when they need an auxiliary construction, an extra point or line that is not in the problem statement. AlphaGeometry pairs a symbolic deduction engine, which exhausts what follows mechanically, with a language model that proposes constructions when deduction runs dry. The model was trained on 100 million synthetic proofs generated from random diagrams, with no human proof data, and its output is a symbolic proof that the deduction engine checks. Published in Nature.

WithTrieu Trinh, Thang Luong

Original work
NatureGitHub
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Automated geometry provers date to Wu's method in the 1970s and to GEX/Java Geometry Expert; the prior state of the art solved 10 of the 30 benchmark problems. The benchmark is IMO geometry problems from 2000–2022 with known solutions, so the contribution is the method and the score, not a new theorem.

Caveats

Benchmark performance on solved problems, not a discovery. The domain is narrow: plane geometry expressible in the system's formal language, which excludes problems involving inequalities or variable numbers of points, and the benchmark set was filtered to what the language can state. Proof steps are machine-checked by the deduction engine, which is what carries the confidence; the peer-reviewed grade reflects Nature review of the system and results.

Carnegie Mellon University
2023-12-20
Peer reviewed AI-led
Model
GPT-4 (Coscientist agent)
Field
chemistry

A language-model agent that planned and ran chemistry experiments on lab robots

Coscientist, driven by GPT-4 with web search, documentation search, code execution and lab automation, designed and executed palladium-catalysed cross-coupling reactions on real robotic hardware from a one-line goal.

Given "perform Suzuki and Sonogashira reactions", the agent searched the literature to identify the chemistry, read the hardware documentation for an API it had not seen, wrote the control code, located the reagents on the plate from a UV-Vis reading, and ran the reactions on a liquid handler. The paper reports six tasks in total, including reaction optimisation under Bayesian search. Published in Nature.

WithDaniil Boiko, Robert MacKnight, Gabe Gomes

Challenge
none linked
Novelty check, caveats & sources
Novelty check

Robotic synthesis platforms and automated reaction optimisation predate this (Cronin's mobile robotic chemist, Nature 2020, among others). What was new is a general-purpose language model doing the planning, tool selection and code generation across unfamiliar hardware from a natural-language goal, rather than executing a human-written protocol.

Caveats

The chemistry performed is textbook cross-coupling: the finding is about autonomy, not about a new compound or reaction. The paper's own safety section shows the agent could be steered toward hazardous syntheses and the authors call for safeguards. Humans set up the hardware, reagents and safety envelope, and the tasks were chosen to be within the platform's reach.

MIT / Broad Institute / Harvard
2023-12-20
Peer reviewed Search scaffold
Model
Ensembles of graph neural networks with substructure attribution
Field
chemistry

A new structural class of antibiotic candidates against MRSA

Graph neural networks trained on 39,312 assayed compounds and applied to 12 million molecules surfaced a chemical class active against MRSA in mice, with the substructures driving each prediction made explicit rather than left opaque.

The team measured antibiotic activity and human-cell cytotoxicity for 39,312 compounds, trained network ensembles on that data, and predicted both properties for over 12 million molecules. Rather than reading off top scores, they extracted the chemical substructures the models were keying on, which let them pick a class rather than isolated hits. Two lead compounds cleared MRSA infection in mouse models, topically and systemically, and appear to kill by collapsing the electrochemical gradient across the bacterial membrane. Published in Nature.

WithFelix Wong, Erica Zheng, James Collins

Original work
Nature
Challenge
none linked
Novelty check, caveats & sources
Novelty check

This is the same group's follow-up to the 2020 halicin work and to abaucin (2023); the new element is selecting a structural class via model interpretation instead of screening for individual hits. "New structural class" is a claim about scaffold novelty relative to clinical antibiotics, checked against known antibiotic chemotypes in the paper.

Caveats

These are candidates, not drugs: mouse models only, with no clinical development at the time of this entry. The mechanism, disrupting membrane potential, is the same one halicin uses, so the novelty is structural rather than mechanistic, and membrane-active compounds carry a known mammalian-toxicity risk. The screening library, assays and interpretation were human-run; the model scored and explained.

Google DeepMind
2023-12-14
Peer reviewed Search scaffold
Model
FunSearch (PaLM 2 / Codey)
Field
mathematics
Posed
1970 · open 53 yrs

New lower bound constructions for the cap set problem

FunSearch discovered larger cap sets than any previously known construction, the first time an LLM-based system produced a genuinely new discovery on an established open problem.

FunSearch pairs an LLM that proposes programs with an automated evaluator that rejects incorrect ones, sidestepping hallucination by construction. Applied to extremal combinatorics, it found new large cap sets in both finite-dimensional and asymptotic cases. Published in Nature. Historically the first entry in this registry's scope.

Original work
NatureGitHub
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Cap set bounds are a well-tracked literature; the constructions were confirmed to exceed the standing record at publication.

Caveats

The LLM proposes; a human-designed evaluator and search loop does the selecting. Calling this 'the AI discovered it' overstates the model's role, though the discovery is real.

Google DeepMind
2023-12-14
Peer reviewed Search scaffold
Model
FunSearch (PaLM 2 / Codey)
Field
computer-science
Posed
1971 · open 52 yrs

Improved heuristics for online bin packing

FunSearch produced bin-packing heuristics outperforming standard baselines on benchmark distributions.

Discovered programs are human-readable, which allowed domain experts to inspect and deploy them. Practical rather than theoretical significance.

Original work
NatureGitHub
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Compared against best-fit and first-fit families and published heuristics; improvements are empirical on tested distributions.

Caveats

An empirical improvement on benchmark distributions, not a proved worst-case bound.

Google DeepMind
2023-11-29
Disputed Search scaffold
Model
GNoME (graph neural network)
Field
materials

Large-scale prediction of new stable crystalline materials (GNoME)

GNoME predicted about 2.2 million new inorganic crystal structures, roughly 380,000 of them flagged as thermodynamically stable, presented as an order-of-magnitude expansion of known stable materials.

Graph Networks for Materials Exploration (GNoME) is a graph neural network trained in an active-learning loop against density-functional-theory calculations to predict formation energies and filter candidates by stability. DeepMind released the predicted structures as a public dataset. Published in Nature.

Original work
NatureGitHub
Announced
DeepMind
Novelty check, caveats & sources
Novelty check

Stable-material prediction builds on the Materials Project and OQMD convex-hull databases. The dispute below is precisely about how much of GNoME's set is genuinely new relative to those existing catalogues.

Caveats

Substantive objections were raised and remain unresolved. Cheetham and Seshadri, reviewing the release, found 'scant evidence for compounds that fulfil the trifecta of novelty, credibility and utility,' noting that many entries are minor substitutional variants, radioactive, or otherwise unlikely to be useful. Stability here means a DFT convex-hull prediction, not experimental synthesis; the headline counts are predicted, not made.

Independent checks

Cheetham & Seshadri (critical review): disputed the novelty and utility of most predicted compounds · link ↗

Lawrence Berkeley National Laboratory
2023-11-29
Disputed AI-led
Model
A-Lab (ML planning + robotics)
Field
materials

Autonomous laboratory reports solid-state synthesis of new inorganic compounds

Berkeley's A-Lab, an AI-driven autonomous laboratory, reported synthesising 41 novel inorganic compounds out of 58 targets over 17 days with minimal human intervention.

A-Lab coupled machine-learning synthesis-recipe prediction with robotic sample preparation, heating and X-ray characterisation, drawing targets from stability predictions including GNoME's. It reported making 41 of 58 targeted compounds autonomously. Published in Nature alongside GNoME.

Original work
Nature
Novelty check, caveats & sources
Novelty check

Targets were drawn from computed stability databases. The contested question is whether the synthesised phases were correctly identified and genuinely new, rather than known phases or misindexed results.

Caveats

A detailed critique led by Robert Palgrave argued that many of the 41 claimed compounds were misidentified from the X-ray data, in several cases likely known phases, mixtures, or amorphous products rather than the claimed novel crystals. It is the identification step, not the automation, that is disputed; the autonomy claim itself is not the weak point. A 2026 Author Correction from the team re-analysed the diffraction data and clarified that 'novel' was meant as new to their prediction platform rather than necessarily new to science.

Independent checks

Palgrave et al. (analysis of the reported diffraction data): disputed the identification of most claimed new compounds · link ↗

Authors' correction (Nature, 2026): re-analysed diffraction data; clarified 'novel' meant new to the platform, not to science · link ↗

Google DeepMind
2023-11-14
Peer reviewed AI-led
Model
GraphCast
Field
climate

Medium-range weather forecasts from a graph neural network beat the operational physics model

GraphCast produced more accurate 10-day forecasts than ECMWF's HRES on roughly 90% of the evaluated variable and lead-time combinations, in under a minute on a single TPU against hours on a supercomputer.

GraphCast takes the two most recent global atmospheric states and predicts the next state six hours ahead on a roughly 0.25° grid, applied repeatedly to reach ten days. It is trained on ERA5 reanalysis rather than on the equations of motion, and it beat HRES, the operational gold standard, on the large majority of evaluated variables, rising to almost all of them within the troposphere. It also flagged severe weather events, including cyclone tracks, earlier than the physics model despite never being trained to look for them. Published in Science, with code and weights released.

WithRemi Lam, Peter Battaglia

Original work
ScienceGitHub
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Machine-learning weather models existed before (FourCastNet, Pangu-Weather), and Pangu had already reported beating HRES on some measures. GraphCast's contribution is the breadth of the win across variables and lead times, at operational resolution, with public code. That is what moved national weather services to adopt learned models.

Caveats

GraphCast is trained on ECMWF reanalysis and initialised from ECMWF analyses, so it depends on the physics-based system it is measured against rather than replacing it. Deterministic models trained on a mean-squared-error objective blur fine structure and underestimate extremes such as peak cyclone intensity. It has no conservation guarantees and cannot forecast anything outside its trained variables. Later ensemble models such as GenCast address part of this.

Google DeepMind
2023-09-19
Peer reviewed AI-led
Model
AlphaMissense
Field
medicine

Pathogenicity predictions for 71 million human missense variants

An AlphaFold-derived model classified 89% of all 71 million possible human missense variants as likely benign or likely pathogenic, against the roughly 0.1% that carry a clinical classification today.

AlphaMissense fine-tunes AlphaFold's structural representation on human and primate variant frequency data, so it learns which substitutions natural selection tolerates without being trained on clinical labels. Applied across 19,233 canonical human proteins it scored every possible single amino-acid change, calling 57% likely benign and 32% likely pathogenic. The full catalogue and the model code were released publicly. Published in Science.

Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Variant-effect predictors are a crowded field (SIFT, PolyPhen-2, EVE, CADD). The contribution is proteome-wide coverage at state-of-the-art benchmark accuracy without training on clinical annotations, which is what makes the predictions usable where no clinical data exists.

Caveats

These are predictions, not diagnoses, and DeepMind says so: ACMG/AMP guidance treats computational evidence as supporting rather than standalone. Benchmarks have been criticised for circularity, since the model trains on population frequency and is evaluated partly against datasets encoding related signals. Calibration varies by gene, and a "likely pathogenic" call on a variant nobody has seen in a patient remains a hypothesis.

Google DeepMind
2023-06-07
Independently checked Search scaffold
Model
AlphaDev (AlphaZero-based)
Field
computer-science

Faster sorting routines discovered and merged into the LLVM C++ library

AlphaDev found shorter branchless routines for sorting small fixed-length inputs; reverse-engineered to C++, they were merged into LLVM's libc++, the first change to those routines in over a decade.

AlphaDev treated the construction of a sorting routine as a single-player game played directly in CPU assembly, rewarding shorter and faster correct programs. For sort-3, sort-4 and sort-5 it removed instructions relative to the human-tuned library code, yielding large speedups on short sequences. The routines were reverse-engineered to C++ and accepted into the LLVM libc++ standard sort, which ships to millions of users. Published in Nature.

Original work
NatureGitHub
Announced
DeepMind
Commentary
Wikipedia
Challenge
none linked
Novelty check, caveats & sources
Novelty check

The libc++ small-sort routines had been hand-optimised by compiler engineers for years. AlphaDev's shorter instruction sequences were reviewed against the standing implementations and confirmed to be improvements before being merged.

Caveats

The gains are on very short fixed-length inputs; asymptotic sorting complexity is unchanged, and the headline percentage speedups apply only to those small cases. The 'independent' grade rests on LLVM maintainers reviewing and merging the code, not on a formal proof of optimality.

Independent checks

LLVM libc++ maintainers (code review and merge): reviewed and merged into the standard sort library · link ↗

Independent commentary
Video explainers
AlphaDev: Discovering Faster Sorting Algorithms with Reinforcement LearningAI for Good (ITU) YouTube ↗

Nothing loads from YouTube until you press play.

Google DeepMind
2022-10-05
Peer reviewed Search scaffold
Model
AlphaTensor (AlphaZero-based)
Field
computer-science
Posed
1969 · open 53 yrs

Faster matrix-multiplication algorithms found by reinforcement learning

AlphaTensor discovered matrix-multiplication schemes using fewer scalar multiplications than any previously known, including a way to multiply 4×4 matrices over GF(2) in 47 multiplications versus the 49 of two-level Strassen.

AlphaTensor recast the search for matrix-multiplication algorithms as a single-player game of decomposing a 3-D tensor into rank-one terms, then trained an AlphaZero-style agent to play it. It matched or beat the best known rank for many matrix sizes, and over GF(2) found a 4×4 scheme using 47 multiplications, improving on the 49 obtained by applying Strassen's 1969 algorithm twice. Every decomposition it returns is an exact identity checkable by hand. Published in Nature.

Original work
NatureGitHub
Announced
DeepMind
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Fast matrix multiplication has an exhaustively catalogued literature going back to Strassen (1969) and Laderman (1976). AlphaTensor's schemes were checked against the standing per-size records; the 4×4 result over GF(2) was new at publication.

Caveats

The 4×4 improvement is specific to characteristic-2 arithmetic and the practical speedup is limited. Within about a week, Kauers and Moosbauer used a classical computer search to improve the 5×5 GF(2) case from AlphaTensor's 96 multiplications to 95, showing conventional methods remained competitive. 'Discovered' here means an RL search over decompositions, not an autonomous mathematical insight.

Independent checks

Kauers & Moosbauer (computer-algebra search): verified and shortly improved on the 5×5 case · link ↗

Video explainers
How AI Discovered a Faster Matrix Multiplication AlgorithmQuanta Magazine YouTube ↗
AlphaTensor by DeepMind explainedYannic Kilcher YouTube ↗

Nothing loads from YouTube until you press play.

Google DeepMind
2022-02-16
Peer reviewed Search scaffold
Model
DeepMind RL controller
Field
physics

Deep reinforcement learning controls tokamak fusion plasma

A reinforcement-learning controller learned to shape and stabilise the magnetic confinement of real fusion plasma inside the TCV tokamak, holding configurations that are hard to sustain by hand.

Working with EPFL's Swiss Plasma Center, DeepMind trained a deep reinforcement-learning agent to command the magnetic control coils of the TCV tokamak. Learning in a simulator against specified target shapes, the controller then ran on the real machine, holding elongated, snowflake and other configurations, including a 'droplet' state with two separate plasmas at once. It is one of the first times a single learned controller replaced the hand-engineered cascade normally required.

Original work
Nature
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Magnetic plasma control is normally built from many separately engineered controllers. A single RL-learned controller running on a real tokamak was new; the Nature paper documents the comparison against conventional control.

Caveats

Demonstrated on the TCV research tokamak, not a power-producing reactor, and the policy was trained in simulation before transfer to hardware. This is a control-engineering advance, not a solution to fusion energy. Independent groups have since extended RL plasma control to other machines.

Google DeepMind (with Oxford and Sydney)
2021-12-01
Peer reviewed AI-assisted
Model
Supervised networks with gradient-based attribution
Field
mathematics

Two theorems found by machine pattern-spotting in knot theory and representation theory

Networks trained to predict one mathematical invariant from another, then probed for which inputs mattered, pointed mathematicians to a proved theorem linking a knot's signature to a new geometric quantity, and to structure behind the combinatorial invariance conjecture.

The recipe was to train a network to predict one invariant from another, then use attribution to see which inputs carried the signal, and hand that to a mathematician. In knot theory it led Marc Lackenby to define a new quantity, the natural slope of a knot, and to prove a theorem bounding the signature in terms of it. In representation theory it led Geordie Williamson to structure in Bruhat interval graphs bearing on the combinatorial invariance conjecture for Kazhdan–Lusztig polynomials. Published in Nature.

WithAlex Davies, Marc Lackenby, Geordie Williamson

Original work
NatureGitHub
Announced
Oxford
Challenge
arXiv
Novelty check, caveats & sources
Novelty check

Both theorems are new and neither had a prior form in the literature; both were proved conventionally by the human mathematicians after the models indicated where to look. The general idea of using computation to generate conjectures is old; what was new was the attribution step turning a black-box predictor into a usable hint.

Caveats

The mathematics was done by humans; the models indicated which relationships were worth studying, which is why autonomy is ai-assisted. Ernest Davis's review argues the "guiding human intuition" framing overstates the contribution and that the knot-theory signal was within reach of simpler statistical methods. The theorems themselves are not disputed.

Independent checks

Ernest Davis (critical review of the framing): theorems accepted; disputes how much the deep learning contributed · link ↗

Google DeepMind
2021-07-15
Independently checked AI-led
Model
AlphaFold2
Field
biology
Posed
1972 · open 49 yrs

Accurate protein structure prediction across the known proteome (AlphaFold2)

AlphaFold2 predicted three-dimensional structures for nearly all catalogued proteins from amino-acid sequence at accuracy rivalling experiment, work that earned the 2024 Nobel Prize in Chemistry.

AlphaFold2 pairs an attention-based neural network with evolutionary sequence information to predict how a protein folds from its amino-acid sequence. At the 2020 CASP14 blind assessment it reached accuracy competitive with experimental methods, and DeepMind and EMBL-EBI then released predicted structures for over 200 million proteins, nearly the entire catalogued proteome. Demis Hassabis and John Jumper shared the 2024 Nobel Prize in Chemistry for the work, alongside David Baker for computational protein design.

WithJohn Jumper, Demis Hassabis

Original work
NatureGitHub
Announced
DeepMind
Commentary
Nobel Prize
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Protein-structure prediction had been an open grand challenge for roughly fifty years, benchmarked every two years by the CASP assessment. AlphaFold2's CASP14 result was a discontinuous jump over all prior methods and is uncontested.

Caveats

AlphaFold predicts structures rather than determining them experimentally; outputs are computational hypotheses that can be wrong for disordered regions, alternative folds, point mutations, and many complexes. It predicts natural structures, not new biology on its own. The grade reflects the prediction step; the Nobel recognised the human-built method.

Independent checks

CASP14 blind assessment (independent assessors): accuracy competitive with experiment across most targets

Royal Swedish Academy of Sciences (2024 Nobel in Chemistry): awarded to Hassabis and Jumper for protein structure prediction · link ↗

Tel Aviv University
2021-04-29
Independently checked Search scaffold
Model
Deep cross-entropy method (custom network)
Field
mathematics

Reinforcement learning refutes several conjectures in extremal combinatorics

A neural network building graphs edge by edge under a cross-entropy search produced explicit counterexamples to several published conjectures in graph theory and extremal combinatorics.

Each conjecture was rewritten as a scoring function on graphs. A network proposes graphs one edge at a time, the top-scoring constructions are kept, and the network is retrained on them, so the search drifts toward whatever the conjecture says should be impossible. Among the conjectures refuted are a question of Brualdi and Cao on maximising permanents of pattern-avoiding matrices and several concerning adjacency and distance eigenvalues of graphs. The counterexamples are small finite graphs printed in the paper, so each can be checked by hand or by a few lines of code.

WithAdam Zsolt Wagner

Original work
arXivGitHub
Challenge
arXiv
Novelty check, caveats & sources
Novelty check

Computer search for combinatorial counterexamples predates this work: linear-programming and SAT-based refutations go back years, including Refuting conjectures in extremal combinatorics via linear programming (2019). What was new is the reinforcement-learning formulation and the specific conjectures, which were open at the time. The paper is widely treated as the origin of the "AI finds counterexamples" line of work.

Caveats

An arXiv preprint; no journal publication is listed on the arXiv record five years on. The refuted conjectures are individually modest (none is a named problem of the Erdős or Jacobian class), and several came from open-problem lists rather than headline literature. Autonomy is search-scaffold on the strictest reading: the human chose the conjectures, wrote the scoring function that encodes each one, and designed the search; the network only proposes graphs.

Independent checks

Roucairol & Cazenave, Monte Carlo search replication and extension: reproduced the refutations and found further counterexamples by different search · link ↗

MIT / Broad Institute
2020-02-20
Peer reviewed Search scaffold
Model
Directed message-passing graph neural network (Chemprop)
Field
medicine

Halicin, an antibiotic found by a neural network screening a compound library

A graph neural network trained on 2,335 molecules picked halicin out of a repurposing library; it killed multidrug-resistant bacteria including Acinetobacter baumannii and Mycobacterium tuberculosis in vitro and cleared infections in mice.

The model was trained to predict growth inhibition of E. coli from structure alone, then applied to the Drug Repurposing Hub. Halicin, an abandoned diabetes candidate, scored highly despite being structurally unlike known antibiotics. It kills by dissipating the proton-motive force across the bacterial membrane, and the team could not evolve resistant E. coli over 30 days of serial passage. A follow-up screen of about 107 million molecules from ZINC15 produced eight further candidates. Published in Cell.

WithJonathan Stokes, Regina Barzilay, James Collins

Original work
Cell
Announced
MIT News
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Virtual screening long predates deep learning. What was new was a model surfacing a hit structurally distant from its training set, with a mechanism distinct from existing antibiotic classes, and the follow-through to efficacy in animals. Halicin itself was a known molecule; the discovery is of its antibacterial activity, not of the compound.

Caveats

Halicin had not entered clinical trials six years on, which is the usual fate of antibiotic leads and a reminder that discovery is not the bottleneck in this field. The model ranked compounds; the mechanism work, the animal studies and the interpretation are human. Membrane-disrupting antibacterials carry a known selectivity risk in mammals.

Google Brain / University of Texas at Austin
2017-12-14
Peer reviewed Search scaffold
Model
Convolutional neural network (AstroNet)
Field
astronomy

An eighth planet around Kepler-90 found by a neural network

A convolutional network re-examining Kepler light curves recovered transit signals the standard pipeline had ranked below threshold, yielding Kepler-90i and making Kepler-90 the first star besides the Sun known to host eight planets.

The network was trained on 15,000 previously vetted Kepler signals to separate genuine transits from false positives, then pointed at 670 stars already known to host multiple planets, searching the weak signals the automated pipeline had discarded. It surfaced Kepler-90i (about 30% larger than Earth, orbiting every 14.4 days, with a surface hot enough to rival Mercury) and Kepler-80g, which completes a five-planet resonant chain. Published in The Astronomical Journal.

WithChristopher Shallue, Andrew Vanderburg

Original work
IOPGitHub
Announced
NASA
Challenge
none linked
Novelty check, caveats & sources
Novelty check

Machine classification of Kepler transit candidates already existed in the pipeline (Autovetter, Robovetter). What was new was training a network to vet signals below the pipeline's detection threshold and getting confirmable planets out of the discard pile.

Caveats

The network vetted candidate signals; the transit search, the follow-up and the statistical validation were conventional. "First system with as many planets as ours" is a statement about what has been detected, not what exists. Detection is heavily biased toward close-in planets, and all of Kepler-90's known planets orbit within roughly Earth's distance from its star. The 670-star search produced two new planets in total.

All findings in the registry, sortable by column.
ModelField
2026-08-13 SOP_2 and SOP_3 theories shown to coincide Independent ChatGPT 5.6 Author verified Collaborative mathematics
2026-08-12 Complete minimizer picture for Gamow's liquid drop model Independent ChatGPT 5.6 Pro Author verified AI-led mathematics
2026-08-10 Proportion of zeta zeros on the critical line raised to 67.25% Anthropic Claude (unreleased research version) Formally verified AI-led mathematics
2026-08-08 A 112-vertex counterexample to the Petersen coloring conjecture Independent Unnamed OpenAI model Formally verified AI-assisted mathematics
2026-08-05 Sendov's conjecture proved for every degree Independent GPT-5.6 Pro Formally verified Collaborative mathematics
2026-08-05 Counterexamples to Schiffer's conjecture and the Pompeiu problem Independent GPT-5.6, Claude Opus 4.8, Claude Fable 5 Formally verified AI-assisted mathematics
2026-08-04 Asymptotic degree-diameter problem resolved for fixed diameter Independent GPT-5.6 Pro Formally verified AI-assisted mathematics
2026-08-01 Ten results in mathematics and theoretical computer science with Lean certificates OpenAI Astra Formally verified AI-led mathematics
2026-07-29 Optimal exponent relating sumsets and difference sets determined Tencent Hunyuan Hy3 (Hyra research agent) Formally verified AI-led mathematics
2026-07-28 Kemeny rank aggregation shown NP-hard for three voters Independent GPT-5.6 Sol Ultra, Claude Fable 5 Formally verified AI-led computer-science
2026-07-27 Feige's conjecture on sums of nonnegative random variables settled Independent ChatGPT 5.6 Pro Formally verified AI-led mathematics
2026-07-22 Counterexample to the Dinitz–Garg–Goemans conjecture Independent GPT-5.6 Pro Claimed AI-led computer-science
2026-07-20 Counterexamples to the Gaussian moments conjecture Independent GPT-5.6 Sol Pro + Claude Fable 5 Author verified AI-led mathematics
2026-07-20 Eight problems from the Kourovka Notebook solved and formalized in Lean Harmonic Aristotle Formally verified Autonomous mathematics
2026-07-20 Gaussian product inequality conjecture proved Independent ChatGPT 5.6 Sol Formally verified AI-led mathematics
2026-07-19 Counterexample to the Jacobian conjecture in dimension three Anthropic Claude Fable 5 Formally verified Collaborative mathematics
2026-07-14 Near-quadratic lower bound for derivative-free convex optimization Independent GPT-5.6 Sol Pro Formally verified AI-led mathematics
2026-07-14 Sabidussi's compatibility conjecture proved Independent GPT-5.6 Pro, GPT-5.6 Sol Formally verified Collaborative mathematics
2026-07-11 Counterexample to Grothendieck's question on finite flat group schemes Independent OpenAI Sol (construction); Claude Fable (Lean formalisation) Formally verified Collaborative mathematics
2026-07-10 Cycle double cover conjecture proved for all bridgeless multigraphs OpenAI GPT-5.6 Sol Ultra Formally verified AI-led mathematics
2026-06-25 First Herculaneum scroll read end to end without unrolling it Vesuvius Challenge Community-developed ink-detection neural networks Author verified AI-assisted archaeology
2026-06-17 Two kagome superconductors predicted by machine learning and confirmed in the lab Aalto University / Rice University Unknown Peer reviewed Search scaffold materials
2026-06-02 Identity for the critical exponents of jamming derived analytically Independent Claude Sonnet 4.6, Claude Opus 4.7 Peer reviewed Collaborative physics
2026-05-21 Nine Erdős problems and 44 OEIS conjectures proved with machine-checked proofs Google DeepMind AlphaProof Nexus (LLM + Lean) Formally verified Search scaffold mathematics
2026-05-20 Disproof of the Erdős unit-distance conjecture OpenAI GPT-5 series reasoning model Independently checked AI-led mathematics
2026-05-18 Phase 1 trial of a computationally designed pan-sarbecovirus vaccine University of Cambridge / DIOSynVax Unknown Peer reviewed AI-assisted medicine
2026-01-15 Structure in Bruhat intervals of permutation groups Google DeepMind AlphaEvolve (Gemini-based) Independently checked Search scaffold mathematics
2026-01-13 Erdős problem #728 resolved and formalized in Lean OpenAI / Harmonic GPT-5.2 Pro + Aristotle Formally verified Autonomous mathematics
2025-11-20 Early science acceleration experiments with GPT-5 OpenAI GPT-5 Peer reviewed AI-assisted computer-science
2025-11-03 AlphaEvolve across 67 problems: 20 improvements, 8 regressions Google DeepMind AlphaEvolve (Gemini-based) Independently checked Search scaffold mathematics
2025-10-19 GPT-5 "solved 10 Erdős problems": it located existing solutions OpenAI GPT-5 Already known Retrieval mathematics
2025-09-25 Limits to black-box amplification in QMA, with the key step written by GPT-5 UT Austin / CWI Amsterdam GPT-5 Thinking Author verified AI-assisted computer-science
2025-09-17 New families of unstable singularities in fluid equations Google DeepMind (with Brown, NYU and Stanford) Physics-informed neural networks with Gauss–Newton optimisation Author verified Search scaffold mathematics
2025-09-17 AI-generated bacteriophage genomes that replicate and kill bacteria Arc Institute / Stanford University Evo 1 and Evo 2 Peer reviewed AI-led biology
2025-09-11 Strong prime number theorem formalized in Lean by an autoformalization agent Math, Inc. Gauss (autoformalization agent) Author verified AI-assisted mathematics
2025-08-01 Improved step-size bound in smooth convex optimization OpenAI GPT-5 Pro Already known AI-assisted mathematics
2025-07-28 Machine-checked Lean proofs for five of six 2025 IMO problems Harmonic Aristotle Formally verified AI-led mathematics
2025-07-21 Gold-medal standard at the 2025 International Mathematical Olympiad Google DeepMind Gemini Deep Think (advanced version) Independently checked AI-led mathematics
2025-06-03 Phase 2a results for a drug whose target and molecule both came from AI Insilico Medicine PandaOmics (target) + Chemistry42 (molecule) Peer reviewed Search scaffold medicine
2025-05-14 4×4 complex matrix multiplication in 48 scalar multiplications Google DeepMind AlphaEvolve (Gemini-based) Independently checked Search scaffold computer-science
2025-05-14 Improved lower bound for the 11-dimensional kissing number Google DeepMind AlphaEvolve (Gemini-based) Independently checked Search scaffold mathematics
2025-05-14 Improved bound for the Erdős minimum-overlap problem Google DeepMind AlphaEvolve (Gemini-based) Independently checked Search scaffold mathematics
2025-05-01 Candidate treatment for dry age-related macular degeneration FutureHouse Robin (multi-agent) Peer reviewed AI-led biology
2025-03-12 A fully machine-generated paper passed workshop peer review Sakana AI The AI Scientist-v2 Author verified AI-led computer-science
2025-02-19 AI co-scientist hypotheses on antimicrobial resistance and liver fibrosis Google AI Co-Scientist (Gemini 2.0 multi-agent) Peer reviewed AI-assisted biology
2025-02-13 Enzymes with working catalytic machinery designed from scratch Institute for Protein Design, University of Washington RFdiffusion with PLACER ensemble scoring Peer reviewed Search scaffold chemistry
2025-01-16 A working fluorescent protein generated by a language model EvolutionaryScale ESM3 (98B) Peer reviewed AI-led biology
2025-01-16 Generative model designs crystals to order; its flagship synthesis turned out to be a known compound Microsoft Research MatterGen Disputed Search scaffold materials
2024-11-20 A neural decoder that identifies quantum errors more accurately than hand-designed methods Google DeepMind / Google Quantum AI AlphaQubit Peer reviewed AI-led physics
2024-10-02 Complete wiring diagram of an adult fruit-fly brain FlyWire Consortium (Princeton, MRC LMB, Cambridge, Vermont) Automated electron-microscopy segmentation networks Peer reviewed AI-assisted neuroscience
2024-07-25 Silver-medal standard at the 2024 International Mathematical Olympiad Google DeepMind AlphaProof + AlphaGeometry 2 Independently checked AI-led mathematics
2024-05-08 Joint structure prediction for proteins, nucleic acids and ligands (AlphaFold 3) Google DeepMind / Isomorphic Labs AlphaFold 3 Peer reviewed AI-led biology
2024-02-21 Reinforcement learning steers a tokamak away from tearing instabilities Princeton University / PPPL / DIII-D National Fusion Facility Deep reinforcement learning controller over a learned plasma model Peer reviewed Search scaffold physics
2024-01-17 Olympiad geometry solved without human demonstrations Google DeepMind AlphaGeometry Peer reviewed Search scaffold mathematics
2023-12-20 A language-model agent that planned and ran chemistry experiments on lab robots Carnegie Mellon University GPT-4 (Coscientist agent) Peer reviewed AI-led chemistry
2023-12-20 A new structural class of antibiotic candidates against MRSA MIT / Broad Institute / Harvard Ensembles of graph neural networks with substructure attribution Peer reviewed Search scaffold chemistry
2023-12-14 New lower bound constructions for the cap set problem Google DeepMind FunSearch (PaLM 2 / Codey) Peer reviewed Search scaffold mathematics
2023-12-14 Improved heuristics for online bin packing Google DeepMind FunSearch (PaLM 2 / Codey) Peer reviewed Search scaffold computer-science
2023-11-29 Large-scale prediction of new stable crystalline materials (GNoME) Google DeepMind GNoME (graph neural network) Disputed Search scaffold materials
2023-11-29 Autonomous laboratory reports solid-state synthesis of new inorganic compounds Lawrence Berkeley National Laboratory A-Lab (ML planning + robotics) Disputed AI-led materials
2023-11-14 Medium-range weather forecasts from a graph neural network beat the operational physics model Google DeepMind GraphCast Peer reviewed AI-led climate
2023-09-19 Pathogenicity predictions for 71 million human missense variants Google DeepMind AlphaMissense Peer reviewed AI-led medicine
2023-06-07 Faster sorting routines discovered and merged into the LLVM C++ library Google DeepMind AlphaDev (AlphaZero-based) Independently checked Search scaffold computer-science
2022-10-05 Faster matrix-multiplication algorithms found by reinforcement learning Google DeepMind AlphaTensor (AlphaZero-based) Peer reviewed Search scaffold computer-science
2022-02-16 Deep reinforcement learning controls tokamak fusion plasma Google DeepMind DeepMind RL controller Peer reviewed Search scaffold physics
2021-12-01 Two theorems found by machine pattern-spotting in knot theory and representation theory Google DeepMind (with Oxford and Sydney) Supervised networks with gradient-based attribution Peer reviewed AI-assisted mathematics
2021-07-15 Accurate protein structure prediction across the known proteome (AlphaFold2) Google DeepMind AlphaFold2 Independently checked AI-led biology
2021-04-29 Reinforcement learning refutes several conjectures in extremal combinatorics Tel Aviv University Deep cross-entropy method (custom network) Independently checked Search scaffold mathematics
2020-02-20 Halicin, an antibiotic found by a neural network screening a compound library MIT / Broad Institute Directed message-passing graph neural network (Chemprop) Peer reviewed Search scaffold medicine
2017-12-14 An eighth planet around Kepler-90 found by a neural network Google Brain / University of Texas at Austin Convolutional neural network (AstroNet) Peer reviewed Search scaffold astronomy

Browse by topic

Mathematics36Computer science9Biology6Materials science4Medicine4Physics4Chemistry3Archaeology1Astronomy1Climate science1Neuroscience1

Announced at the source

Every organisation on record here, with how many entries it accounts for, linked to where it posts its own results.

Also watched, nothing on record yet

Major labs we follow that have not yet produced a result meeting the bar for an entry. They move into the list above the moment one does.

Reference and discussion

Where results get formalised, checked and argued over. External links, open in a new tab.

Keyboard shortcuts

k
Command palette: entries, filters, sorts and pages. Cmd K works too
/
Focus the search box
Esc
Clear the search and leave the box
v
Switch between cards and table
t
Cycle the theme: light, system, dark
?
Open this list

Every filter, the sort order and the card or table choice are all in the address bar, so any view of the registry can be linked or bookmarked.

Arrow keys move, Enter opens, Escape closes.