AI FOR SCIENCE · ON THE RECORD
MIT
On the timeline →People filter →
Research on the record
- 2026-10-14A machine-learning correction on top of ECMWF's forecasts tops ECMWF's 2025 real-time contest for weather three and four weeks ahead, ahead of six operational forecast centres and 34 teamsLevel 3 · Checked against existing evidence
- 2026-10-08AutoDrude fits a polarizable force field straight from DFT, and catches the lithium transport in a polymer electrolyte that charge scaling missesLevel 3 · Checked against existing evidence
- 2026-10-08A doubly metastable synthesis method makes four phases, three of which match no structure in the databases its authors searchedLevel 4 · Tested in the physical world
- 2026-10-01MIT's VISTA harness lets Claude Opus 5.0 finish all 25 public ARC-AGI-3 games from screen images, with no new model trainedLevel 2 · Computed
- 2026-09-18MIT's Bayesian-optimization AI finds thermostable mRNA vaccine formulations in six iterationsLevel 4 · Tested in the physical world
- 2026-09-16ShEPhERD-2 designs drug-like molecules from a 3D interaction profile alone, 6.4x faster than its predecessorLevel 2 · Computed
- 2026-09-15Google, Google DeepMind and MIT FutureTech survey 637 scientists and analyse 15 million Gemini interactions: about 7 hours saved a week, and much of it spent checking the AINot graded
- 2026-09-08GPT-5.6 Sol runs experiments on a superconducting quantum chipLevel 4 · Tested in the physical world
- 2026-09-08MIT's IRIS reads which signals a cell received during development from its gene activity, and a mouse-tissue test confirmed one predictionLevel 4 · Tested in the physical world
- 2026-08-26MIT trains a crystal-and-text model on AI-written descriptions of 126,000 crystals; trained on sentences from papers instead, its property search does no better than chanceLevel 2 · Computed
- 2026-08-18ASI-Bench tests agents without step-by-step research guidanceLevel 2 · Computed
Graded by how far the evidence went, not how good the result is. See the methodology.
Read on AI4S.fyi
- 2026-10-05Same Model, Different Harness: What Three Studies Show About the System Around a ModelBrief
- 2026-10-05VISTA trains no new model, only changes how it sees and remembers: Claude Opus 5.0 goes from 40.68 to 100 on ARC-AGI-3's public gamesBrief
- 2026-09-26Codex Joins the Calibration of a Real Quantum Chip: What One Public Run ShowsBrief
- 2026-10-20The next bottleneck in AI for scienceCommentaryAnnelies Gamble · Mixture of Experts
- 2026-10-19AI for Science: six lessons from 60 papersCommentaryJorge Bravo-Abad · Discovery at Scale
- 2026-09-13Anthropic Just Gave AI Agents a Driver for Your Lab InstrumentsCommentaryJustin Johnson · Run Data Run
Also on the timeline
- 2026-10-26OpenAI arranged its examiners before announcing, and the group it arranged wrote to its field to say the involvement is not certificationPolicy
- 2026-10-20Eka Robotics, founded by MIT's Pulkit Agrawal and ex-DeepMind Tuomas Haarnoja, raises a $150M Series A for robots that grip by feelCapital
- 2026-10-14Google.org gives $36M to 17 organisations using AI in science, from fusion plasmas and sodium-ion batteries to African drug-metabolism genesCapital
People
Only people the AI4S.fyi record has named: in a round, a timeline entry, commentary, an interview or a sourced role. Not the full team.
Named on AI4S.fyi
Larry GuthClaude E. Shannon (1940) Professor of Mathematics A claimed proof of the local smoothing conjecture in every dimension carries the Kakeya conjecture with it, where the published record had reached three dimensions
Ryan WilliamsProfessor of Electrical Engineering and Computer Science OpenAI and Anthropic each claim a subexponential 3-SAT algorithm refuting the Exponential Time Hypothesis, each checked in Lean, neither checked by anyone outside
Pulkit AgrawalAssociate Professor, EECS; leads the Improbable AI Lab at CSAIL Eka Robotics, founded by MIT's Pulkit Agrawal and ex-DeepMind Tuomas Haarnoja, raises a $150M Series A for robots that grip by feel
Artur LyssenkoAuthor affiliation on a 2026 preprint (first author), Materials Science and Engineering AutoDrude fits a polarizable force field straight from DFT, and catches the lithium transport in a polymer electrolyte that charge scaling misses
Joseph G. CheckelskyAuthor affiliation on a 2026 preprint (corresponding author), Physics A doubly metastable synthesis method makes four phases, three of which match no structure in the databases its authors searched
Kyuho LeeAuthor affiliation on a 2026 preprint (first and corresponding author), Physics A doubly metastable synthesis method makes four phases, three of which match no structure in the databases its authors searched
Rafael Gómez-BombarelliAssociate Professor, Materials Science and Engineering AutoDrude fits a polarizable force field straight from DFT, and catches the lithium transport in a polymer electrolyte that charge scaling misses
Pavel EtingofProfessor of Mathematics; Chief Research Adviser of MIT PRIMES Mathematics of the Future
13 more people
Slava GerovitchPRIMES/MathROOTS Director; Senior Lecturer, Department of Mathematics Mathematics of the Future
Kaiming HeAssociate Professor with tenure, Department of EECS MIT's VISTA harness lets Claude Opus 5.0 finish all 25 public ARC-AGI-3 games from screen images, with no new model trained
Linlu QiuPhD student, EECS and CSAIL, advised by Yoon Kim and Jacob Andreas MIT's VISTA harness lets Claude Opus 5.0 finish all 25 public ARC-AGI-3 games from screen images, with no new model trained
Ana JaklenecNow: Principal Research Scientist and Principal Investigator · Koch Institute for Integrative Cancer Research, MIT MIT's Bayesian-optimization AI finds thermostable mRNA vaccine formulations in six iterations
Henry CohnProfessor of Mathematics The technical debt of AI-generated mathematics
Beatriz YankelevichGraduate Student, EECS Department GPT-5.6 Sol runs experiments on a superconducting quantum chip
Pulin LiAssociate Professor of Biology; Core Member, Whitehead Institute MIT's IRIS reads which signals a cell received during development from its gene activity, and a mouse-tissue test confirmed one prediction
Arya RaoNow: MD-PhD Candidate, Harvard/MIT · Harvard/MIT MD-PhD Program Measuring the Frontier of Scientific Capability in the Age of AI- EElsa OlivettiJerry McAfee Professor in Engineering; Professor, Department of Materials Science and Engineering MIT trains a crystal-and-text model on AI-written descriptions of 126,000 crystals; trained on sentences from papers instead, its property search does no better than chance
- JJu LiCarl Richard Soderberg Professor in Power Engineering; Professor of Nuclear Science and Engineering and of Materials Science and Engineering MIT trains a crystal-and-text model on AI-written descriptions of 126,000 crystals; trained on sentences from papers instead, its property search does no better than chance
Markus BuehlerJerry McAfee (1940) Professor in Engineering Unreasonable Labs' ProjectionBench asks models to predict a recent paper's results from its research question alone, then with more detail; GPT-5.4 scores highest
Tess SmidtAssociate Professor of Electrical Engineering; leads the Atomic Architects group Physics Is Different: Context, Culture & Craft in Effective AI for Physics- CCathy WuClass of 1954 Career Development Professor
Previously there
Jian MaNow: Founder · XtalPi XtalPi's LCMS2Yield predicts how strongly a reaction product absorbs UV light, so its yield can be measured without first purifying a reference sample
Shuhao WenNow: Founder · XtalPi XtalPi's LCMS2Yield predicts how strongly a reaction product absorbs UV light, so its yield can be measured without first purifying a reference sample
Nicholas T. HutchinsAuthor affiliation on a 2026 paper (first author), Department of Biological Engineering MIT's IRIS reads which signals a cell received during development from its gene activity, and a mouse-tissue test confirmed one prediction- YYang Jeong ParkPostdoctoral researcher, DMSE and NSE MIT trains a crystal-and-text model on AI-written descriptions of 126,000 crystals; trained on sentences from papers instead, its property search does no better than chance
Chenru DuanNow: Founder · Deep Principle Deep Principle · Series A3 · undisclosed
Haojun JiaNow: Founder · Deep Principle Deep Principle · Series A3 · undisclosed
Caida LaiNow: Founder · METiS TechBio METiS TechBio · IPO · about HK$2.113B gross
Wenshou WangNow: Founder · METiS TechBio METiS TechBio · IPO · about HK$2.113B gross
Lipeng LaiNow: Founder · XtalPi XtalPi · Placing · about HK$2,653.3M gross- TTheo SchaferNow: Founder · Tetsuwan Scientific Tetsuwan Scientific · Pre-seed · $2.7M
This page compiles what AI4S.fyi has recorded and links each item to its source. It is not a grade of the organization and not its complete history: a blank means the record has not caught up as often as it means nothing happened.