Skip to content
ContentLora

    Tip: press / anywhere to search.

    ● Developing story

    AI for science tracker: milestones and what to watch

    AI for science moved from landmark papers to operational use between 2023 and 2026: AI weather models run daily at ECMWF and NOAA, protein design tools are open source, and governments and investors are funding autonomous labs.[1][2][3][4] The open questions are whether AI-designed drugs and materials deliver verified real-world wins.[5][6]

    Editor reviewedUpdated AI for scienceArtificial intelligenceScienceLife sciences

    What we know

    • ECMWF has run an AI forecasting system operationally since 25 February 2025, alongside its physics model[1]
    • NOAA deployed operational AI-driven global weather models on 17 December 2025[2]
    • RFdiffusion3, an all-atom protein design model, is open source (December 2025)[3]
    • CASP17 predictions closed 11 September 2026; results due 30 November 2026[21]
    • Isomorphic Labs raised $2.1 billion in May 2026[15]
    • Rentosertib, an AI-generated drug for lung fibrosis, began a 320-patient phase III trial in July 2026[13]
    • A January 2026 correction narrowed the A-Lab's 'novel' compounds to new to its prediction platform, with 36 of 40 successes confirmed[14]
    • Open alternatives to AlphaFold 3 exist, including Boltz (MIT licence) and OpenFold3 (Apache 2.0)[10]
    • FDA decision on the computationally designed drug zasocitinib is due in Q1 2027[24]
    • DOE's Genesis Mission had 297 first-year projects by October 2026[4]

    What we don't know yet

    • Whether Isomorphic Labs starts its first clinical trial by its end-2026 target (as of 29 September 2026 it described its work as preclinical)
    • Whether AI-derived drugs beat industry phase II success rates once more trials read out
    • Whether any AI-proposed inorganic material is independently confirmed as new and useful
    • Whether company-reported results such as IsoDDE's benchmarks and Lila Sciences' catalyst screen hold up in peer review
    • How CASP17 results on immune complexes and ligands compare with earlier rounds
    • Whether AI weather models stay reliable for unprecedented extremes in a warming climate
    Story status

    Story status

    State
    developing
    Started
    2021-07-15
    Timeline entries
    30
    Last entry

    Timeline data (JSON)

    This tracker follows the frontier of AI for science across four areas: biomolecules, weather and climate, materials and drugs, and autonomous labs, from AlphaFold 3 in 2024 to the Genesis Mission awards of October 2026.[7][4] Each timeline entry is dated and sourced.

    Where things stand (October 2026)

    Weather is the most mature area: ECMWF has run its AI system operationally since February 2025 and NOAA since December 2025, both alongside physics-based models.[1][2] In biology, structure prediction is moving from single proteins to complexes, and the AlphaFold Database added 1.7 million predicted homodimers in March 2026.[8] Protein design tools such as RFdiffusion3 are open source, and AI agents are now running design campaigns end to end.[3][9] Open models such as Boltz and OpenFold3 offer AlphaFold 3-class prediction outside Google’s licence terms.[10][11][12]

    Drugs and materials are where evidence is thinnest. AI-derived molecules have shown about 40% phase II success, similar to industry norms, and reporting in late 2025 found no convincing big win in AI materials discovery.[5][6] The most advanced generative-AI drug, rentosertib, entered phase III in July 2026.[13] In materials, the A-Lab’s authors wrote in January 2026 that their “novel” compounds were new to the prediction platform, not necessarily new to science.[14]

    Recent developments (July to October 2026)

    Drug discovery saw the most movement. Insilico Medicine started a 320-patient phase III trial of rentosertib in July, while Isomorphic Labs, which raised $2.1 billion in May, said in late September that its preclinical data was readying it for clinical work.[13][15][16] In August Anthropic reported that Claude-led protein design produced binders for 14 of 15 targets in external lab tests.[9]

    Public money also flowed. DOE announced the first Genesis Mission projects in July and, on 8 October, 12 Phase II awards worth $159 million, including a fusion digital twin led by Commonwealth Fusion Systems.[17][4][18] Autonomous-lab start-up Lila Sciences, one of the Phase I winners, reported a 2,942-catalyst screen for green hydrogen in September; the result is company-reported.[19][20]

    What to watch next

    • CASP17 results (30 November 2026). The assessment targets areas where deep learning has yet to deliver, such as immune complexes and ligand binding.[21][22]
    • Isomorphic Labs’ first trial. The company said in January 2026 it expected its first clinical trials by the end of 2026; on 29 September it was still describing preclinical work.[23][16]
    • Zasocitinib FDA decision (Q1 2027). A test case for how “AI-designed” drugs are labelled.[24]
    • Rentosertib’s phase III. The 320-patient, 52-week placebo-controlled trial in China started in July 2026.[13]
    • Genesis Mission outputs. 297 first-year projects should begin reporting results.[4]
    • Autonomous materials labs. Peer-reviewed results from Periodic Labs, Lila Sciences and others, and whether they survive independent checks.[25][26]

    How this tracker is updated

    New entries come from this field’s official feeds and are added after the primary source is checked.

    For background, see alphafold, ecmwf-aifs and self-driving-labs. Many milestones here involve Google DeepMind, which built AlphaFold, GNoME and the WeatherNext models.[27][28][29]

    Timeline

    25 confirmed4 reported1 disputed

    1. confirmed

      DOE announces $159 million in Genesis Mission Phase II awards[4][18]

    2. confirmed

      Isomorphic Labs says preclinical data is readying it for clinical development[16]

    3. reported

      Lila Sciences reports AI-run lab screened 2,942 green-hydrogen catalysts[20]

      Company-reported; not peer-reviewed.

    4. confirmed

      FDA accepts zasocitinib application under priority review[24][49]

      The TYK2 inhibitor was developed with machine learning and physics-based simulation.

    5. reported

      Anthropic reports Claude designed binders for 14 of 15 protein targets[9][48]

      Company-reported results with external lab testing; not peer-reviewed.

    6. confirmed

      First Genesis Mission project awards announced[17][19]

    7. confirmed

      Rentosertib enters phase III trial in China[13][47]

      320 patients with idiopathic pulmonary fibrosis, 52 weeks, placebo-controlled.

    8. confirmed

      Isomorphic Labs raises $2.1 billion[15]

    9. confirmed

      AlphaFold Database adds 1.7 million predicted protein complexes[8]

    10. reported

      Isomorphic Labs unveils IsoDDE, claiming double AlphaFold 3's accuracy on a hard ligand benchmark[46]

      Company-reported benchmark results.

    11. reported

      Isomorphic Labs pushes first clinical trials to end of 2026[23][45]

      The same day it announced a multi-target research collaboration with Johnson & Johnson.

    12. confirmed

      A-Lab authors correct novelty claims; 36 of 40 syntheses confirmed[14][44]

    13. confirmed

      NOAA deploys operational AI global weather models[2]

    14. confirmed

      RFdiffusion3 released open source[3]

    15. confirmed

      US launches the Genesis Mission for AI-driven science[43]

    16. confirmed

      Google DeepMind releases WeatherNext 2[42]

    17. confirmed

      Lila Sciences closes $350 million Series A, with NVIDIA's venture arm investing[41]

    18. confirmed

      Periodic Labs launches with $300 million seed for autonomous materials discovery[25]

    19. confirmed

      Google DeepMind steps in to fund CASP after NIH grant runs out[40]

    20. confirmed

      Chai-2 preprint reports 16% hit rate for de novo antibodies[39]

      Company preprint, not peer-reviewed at posting.

    21. confirmed

      Boltz-2 released openly, predicting structure and binding affinity together[38][10]

    22. confirmed

      Google launches Weather Lab; Hurricane Center forecasters see AI cyclone predictions[36][37]

    23. confirmed

      Phase 2a trial of AI-generated drug rentosertib published[35]

    24. confirmed

      ECMWF makes its AIFS AI forecast operational[1]

      The ensemble version followed on 1 July 2025.

    25. confirmed

      MatterGen generative materials model published in Nature[34]

    26. confirmed

      GenCast beats ECMWF's ensemble on 97.2% of targets[33]

    27. confirmed

      Nobel Prize in Chemistry for protein design and AlphaFold[32]

    28. confirmed

      AlphaFold 3 extends prediction to complexes with DNA, RNA and small molecules[7]

    29. disputed

      Chemists dispute A-Lab's claim of new materials[31]

      A PRX Energy perspective concluded no new materials were discovered.

    30. confirmed

      GNoME predicts 2.2 million crystals; A-Lab reports autonomous synthesis[28][30]

      Two Nature papers published the same day launched the AI materials debate.

    Sources

    Each numbered claim is a statement we checked against the sources listed with it. Status shows how well established it is.

    1. [1]

      ECMWF took its Artificial Intelligence Forecasting System (AIFS) into operations on 25 February 2025, running side by side with its physics-based IFS. confirmedas of 2025-02-25

    2. [2]

      On 17 December 2025 NOAA deployed operational AI-driven global models, AIGFS, AIGEFS and the hybrid HGEFS, which NOAA says deliver more accurate guidance using a fraction of the computing resources. confirmedas of 2025-12-17

    3. [3]

      In December 2025 the Institute for Protein Design released RFdiffusion3, an all-atom model that designs proteins interacting with any type of molecule commonly found in cells, with training code and weights available through the Rosetta Commons Foundry. confirmedas of 2025-12-03

    4. [4]

      On 8 October 2026 DOE announced 12 Phase II Genesis Mission awards totaling $159 million, plus six Phase I projects, capping a first-year cohort of 297 projects. confirmedas of 2026-10-08

    5. [5]

      The same 2024 analysis found a phase II success rate of about 40% for AI-discovered molecules, on a limited sample, comparable to historic industry averages. confirmedas of 2024-06-01

    6. [6]

      MIT Technology Review reported in December 2025 that AI materials discovery had not yet produced a convincing big win or a new miracle material. reportedas of 2025-12-15

    7. [7]

      AlphaFold 3, published in Nature in May 2024, uses a diffusion-based architecture to predict joint structures of complexes containing proteins, nucleic acids, small molecules, ions and modified residues. confirmedas of 2024-05-08

    8. [8]

      In March 2026 EMBL-EBI, Google DeepMind, NVIDIA and Seoul National University added 1.7 million high-confidence predicted homodimer structures to the AlphaFold Database. confirmedas of 2026-03-16

    9. [9]

      In August 2026 Anthropic reported that Claude models, using open-source protein design tools, designed binders against 14 of 15 targets, with designs produced and tested by Adaptyv Bio and Twist Bioscience. reportedas of 2026-08-18

    10. [10]

      The Boltz team describes Boltz-1 as the first fully open-source model to approach AlphaFold 3 accuracy, and releases Boltz code and weights under the MIT licence for academic and commercial use. confirmedas of 2026-10-10

    11. [11]

      OpenFold3-preview, from Columbia University's AlQuraishi Lab and the OpenFold consortium, aims to be a bitwise reproduction of AlphaFold 3 and is available under the Apache 2.0 licence for academic and commercial use. confirmedas of 2026-10-10

    12. [12]

      AlphaFold 3's inference code is public on GitHub, but its model parameters may only be used if obtained directly from Google under terms of use; the model is also offered on AlphaFold Server for non-commercial use, with a more limited set of ligands. confirmedas of 2026-10-10

    13. [13]

      On 7 July 2026 Insilico Medicine announced the start of a phase III trial of rentosertib, a randomised, double-blind, placebo-controlled study expected to enrol 320 patients with idiopathic pulmonary fibrosis in China over 52 weeks. confirmedas of 2026-07-07

    14. [14]

      A January 2026 author correction to the A-Lab paper said its novelty claims meant materials new to the prediction platform, not necessarily new to science; a manual re-analysis confirmed 36 of the 40 reported successes, with four inconclusive, and one compound was removed because it had been in the training data. confirmedas of 2026-01-19

    15. [15]

      On 12 May 2026 Isomorphic Labs announced a $2.1 billion funding round led by Thrive Capital, with Alphabet and GV among existing investors and MGX, Temasek, CapitalG and the UK Sovereign AI Fund as new investors. confirmedas of 2026-05-12

    16. [16]

      On 29 September 2026 Isomorphic Labs' president wrote that the company's preclinical data was helping it "gear up for clinical developments"; the post did not announce a clinical trial. confirmedas of 2026-09-29

    17. [17]

      DOE first announced Genesis Mission project awards in July 2026. confirmedas of 2026-10-08

    18. [18]

      DOE's October 2026 Phase II Genesis Mission awards include a Commonwealth Fusion Systems digital twin of a fusion demonstration device, University of Washington enzyme design tools, a UC San Diego project to expand the RNA structure database five-fold and a Fermilab project to design rugged microchips for extreme environments. confirmedas of 2026-10-08

    19. [19]

      Lila Sciences said in July 2026 that three of its proposals won Phase I Genesis Mission awards, each with a national laboratory or university partner. confirmedas of 2026-07-22

    20. [20]

      Lila Sciences reported in September 2026 that its AI-directed lab proposed, synthesised and screened 2,942 catalysts for green hydrogen in three months and identified six high-performing material families; the result is company-reported. reportedas of 2026-09-25

    21. [21]

      The CASP17 prediction season ended on 11 September 2026, with assessment results due on 30 November 2026, a day before the CASP17 conference in Rome. confirmedas of 2026-10-10

      • CASP17 home page · Protein Structure Prediction Center (UC Davis) (retrieved 2026-10-10)
    22. [22]

      CASP17 emphasises areas where deep learning has yet to deliver, including immune complexes, protein–ligand complexes, nucleic acids and conformational ensembles. confirmedas of 2026-10-10

      • CASP17 home page · Protein Structure Prediction Center (UC Davis) (retrieved 2026-10-10)
    23. [23]

      In January 2026 Isomorphic Labs said it expected its first clinical trials by the end of 2026, a year later than its earlier target of the end of 2025. reportedas of 2026-01-20

    24. [24]

      On 14 September 2026 Takeda said the FDA had accepted its new drug application for zasocitinib (TAK-279), an oral TYK2 inhibitor for moderate-to-severe plaque psoriasis, under priority review with a target action date in the first quarter of 2027. confirmedas of 2026-09-14

    25. [25]

      Periodic Labs emerged from stealth in September 2025 with a $300 million seed round, with a first goal of inventing better superconductors. confirmedas of 2025-09-30

    26. [26]

      Flagship Pioneering unveiled Lila Sciences in March 2025 with $200 million in committed seed capital to build a scientific superintelligence platform and fully autonomous labs for life, chemical and materials sciences. confirmedas of 2025-03-10

    27. [27]

      AlphaFold is an AI system developed by Google DeepMind that has predicted the structure of millions of proteins since 2021. confirmedas of 2026-03-16

    28. [28]

      Google DeepMind's GNoME work, published in Nature in November 2023, used graph networks to identify 2.2 million crystal structures, of which it counted 381,000 as newly discovered stable materials. confirmedas of 2023-11-29

    29. [29]

      Google DeepMind's public WeatherNext repository contains code for WeatherNext 2 and the earlier GraphCast and GenCast models, and daily WeatherNext 2 forecast feeds are offered through Google Cloud, Weather Lab and Open-Meteo. confirmedas of 2026-10-10

    30. [30]

      The A-Lab, an autonomous lab for solid-state synthesis described in Nature in November 2023, realized 36 compounds from 57 targets over 17 days of continuous operation, according to the current version of the paper. confirmedas of 2026-10-10

    31. [31]

      A March 2024 PRX Energy perspective concluded that no new materials had been discovered in the A-Lab work and that about two-thirds of claimed successes were likely known disordered versions of predicted compounds. disputedas of 2024-03-07

    32. [32]

      The 2024 Nobel Prize in Chemistry went half to David Baker for computational protein design and half jointly to Demis Hassabis and John Jumper for protein structure prediction. confirmedas of 2024-10-09

    33. [33]

      GenCast, a probabilistic machine-learning weather model published in Nature in December 2024, had greater skill than ECMWF's ENS ensemble on 97.2% of 1,320 evaluated targets. confirmedas of 2024-12-04

    34. [34]

      MatterGen, a generative model for inorganic materials published in Nature in January 2025, produced structures more than twice as likely to be new and stable as earlier generative models. confirmedas of 2025-01-16

    35. [35]

      A phase 2a trial of rentosertib, published in Nature Medicine in June 2025, randomized 71 patients with idiopathic pulmonary fibrosis to one of three doses or placebo for 12 weeks. confirmedas of 2025-06-03

    36. [36]

      In June 2025 Google DeepMind and Google Research launched Weather Lab, with an experimental AI cyclone model that generates 50 scenarios up to 15 days ahead, and began sharing its predictions with the US National Hurricane Center. confirmedas of 2025-06-12

    37. [37]

      Google DeepMind says Weather Lab is a research tool and its live predictions are not official warnings. confirmedas of 2025-06-12

    38. [38]

      Boltz-2, described in a June 2025 preprint by MIT and Recursion researchers, predicts both complex structures and small molecule–protein binding affinity, and its authors say it is the first AI model to approach free-energy perturbation accuracy while being at least 1,000 times more computationally efficient. confirmedas of 2025-06-18

    39. [39]

      In a preprint posted in July 2025, Chai Discovery reported that its Chai-2 model achieved a 16% hit rate in fully de novo antibody design and found at least one binder for 50% of 52 targets, testing 20 or fewer designs per target, none of which had an existing antibody binder in the Protein Data Bank. confirmedas of 2025-07-06

    40. [40]

      In July 2025 Google DeepMind provided temporary funding to CASP after its NIH grant ran out. confirmedas of 2025-07-21

    41. [41]

      In October 2025 Lila Sciences closed a $350 million Series A, bringing its total funding to $550 million, with NVIDIA's venture arm NVentures among the investors. confirmedas of 2025-10-10

    42. [42]

      Google DeepMind's WeatherNext 2, announced in November 2025, uses a Functional Generative Network and surpasses the previous WeatherNext model on 99.9% of variables and lead times from 0 to 15 days. confirmedas of 2025-11-17

    43. [43]

      On 24 November 2025 the US president signed an executive order launching the Genesis Mission, a national initiative to accelerate scientific breakthroughs using AI. confirmedas of 2025-11-24

    44. [44]

      Nature says the A-Lab diffraction re-analysis behind the 2026 correction was peer-reviewed after publication. confirmedas of 2026-01-19

    45. [45]

      On 20 January 2026 Isomorphic Labs announced a cross-modality, multi-target research collaboration with Johnson & Johnson. confirmedas of 2026-01-20

    46. [46]

      In February 2026 Isomorphic Labs reported that its drug design engine, IsoDDE, more than doubles AlphaFold 3's accuracy on a challenging protein–ligand generalisation benchmark and predicts binding affinities more accurately than gold-standard physics-based methods; these are company-reported benchmark results. reportedas of 2026-02-10

    47. [47]

      Insilico lists the rentosertib phase III trial under registry numbers CTR20262475 and NCT07687459. confirmedas of 2026-07-07

    48. [48]

      Anthropic reported hit rates of 22% to 35% for Claude-led binder design depending on setup, against the 10–15% it describes as typical for protein design campaigns, with 354 binders from 1,320 designs. reportedas of 2026-08-18

    49. [49]

      Zasocitinib came from a Nimbus Therapeutics TYK2 programme that, with Schrödinger from 2016, used machine learning and physics-based free energy perturbation to assess more than 13,000 compounds computationally. confirmedas of 2026-09-25

    Revision history (2)
    1. Page created.
    2. Refresh: added 13 milestones (Weather Lab, Boltz-2, Chai-2, Lila Sciences, the A-Lab correction, IsoDDE, rentosertib phase III, Genesis awards, Isomorphic's clinical status) and updated what to watch.

    Created Oct 10, 2026. Last reviewed by an editor on Oct 10, 2026. Next scheduled review: Jan 10, 2027.

    Cite this page

    "AI for science tracker: milestones and what to watch." ContentLora, updated Oct 10, 2026. https://contentlora.com/events/ai-for-science-tracker

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.

    Keep exploring

    ☰ Contents