I am really happy to announce our collaboration between Acellera Therapeutics and Psivant Therapeutics to push the frontiers of AI-driven drug discovery! businesswire.com/news/home/2025…

QuantumBind-RBFE: Accurate Relative Binding Free Energy Calculations Using Neural Network Potentials pubs.acs.org/doi/10.1021/ac… Gianni De Fabritiis Acellera #JCIM Vol65 Issue8 #compchem

At FutureHouse, we’ve noticed scientific agents are good at applying average intelligence across tasks. They always seem to make the obvious choices, which is good, but discovery sometimes requires more intuition and insight than average. We’ve made the first step today towards

Excited to unveil Boltz-2, our new model capable not only of predicting structures but also binding affinities! Boltz-2 is the first AI model to approach the performance of FEP simulations while being more than 1000x faster! All open-sourced under MIT license! A thread… 🤗🚀

Our development of machine-learned transferable coarse-grained models in now on Nat Chem! doi.org/10.1038/s41557… I am so proud of my group for this work! Particularly first authors Nick Charron, Klara Bonneau, Aldo S. Pasos-Trejo, Andrea Guljas.

Heading to the Structure-based drug discovery conference tomorrow. Among other things, I will present work on a co-scientist AI concept, hopefully interesting. See you at sbdd-congress.it/#welcome

Best getting to a conference ever

Heading to NVIDIA today for a workshop on molecular simulations and machine learning potentials, where I will talk about AceFF and TorchMD-Net. I will be in SF 8-12 Oct to chat with SF locals.

Amazing views going back

An exciting milestone for AI in science: Our C2S-Scale 27B foundation model, built with Yale University and based on Gemma, generated a novel hypothesis about cancer cellular behavior, which scientists experimentally validated in living cells.  With more preclinical and clinical tests,

Preprint released of "Speak to a Protein," an AI co-scientist that facilitates data gathering and analysis in an interactive collaborative session. It is quite amazing to use. Preprint: arxiv.org/abs/2510.17826

Our Huxley-Gödel Machine learns to rewrite its own code, estimating its own long-term self-improvement potential. It generalizes on new tasks (SWE-Bench Lite), matching the best officially checked human-engineered agents. Arxiv 2510.21614  With Wenyi Wang, Piotr Piękos,

Heading to Boston for two weeks. Write me if you would like to chat while I am there.