Incredibly excited to say I am going to be joining Orbital Materials to work on driving the adoption and advancement of the amazing tools they've been building. I've been using their recently released Orb model (a universal neural network potential) and it's blowing me away. I

So I think I've found another pretty incredible example of the generalisability of neural network potentials: this is a problem I've been dreaming of tackling for a decade but never felt I had the the tools to get at until now: How do potassium ion channels work. 1/n These

To date, I've found @OrbMaterials pre-trained models to have the best trade-off w.r.t speed and chemical accuracy, with MACE being the next best choice.

So Orb has blown me away again. I simulated the carbonic anhydrase enzyme with it: one of the most important and well studied enzymes in biology. (It converts CO2 to bicarbonate and is involved in many diseases and could also be useful for carbon capture.) Remarkably, despite

Excited to announce Orb-v3, a new family of universal Neural Network Potentials from me and my team at Orbital, led by Ben Rhodes and @sanderhaute! These new potentials span the Pareto frontier of models for computational chemistry.

We’re excited to share that Civo is deploying our #carbonremoval technology at its UK data center – a crucial step for the #tech industry and revolutionizing AI. More in Semafor: bit.ly/4i672yb Learn more about our data center products: bit.ly/3RKwFK4

Delighted to see our paper published: CHIPS-FF: Evaluating Universal Machine Learning Force Fields for Material Properties pubs.acs.org/doi/10.1021/ac… #CHIPSact #jarvisnist #mlff

A project we started a long time ago with Michael Pieler, Philippe Schwaller (he/him) and many others is now on Arxiv: The ChemPile - a massive dataset for training chemical foundation models.

Great leaderboards challenge the whole field to move faster. Thanks to all the authors on Matbench Discovery for your contributions!

A new evaluation shows our AI simulation model, Orb, stands out in many tests. 👏 In this paper (arXiv:2506.01860) by Bowen Han & Yongqiang Cheng, Orb v3 was benchmarked against ~5,000 inorganic crystals and outperformed other MLIPs such as MatterSim and SevenNet-MF-ompa. Take

In high-density racks, every watt spent on cooling is a watt you can’t spend on compute. Our two-phase direct-to-chip cooling system for next-gen GPUs means: ➡️ More power for compute ➡️ Lower energy use and data center PUE ➡️ More hardware, bigger models See how it works

Fighting over a shrinking pie is the cause of much of the UK's political and social woes, but as Matt reminds us - we have the talent and human capital to grow it. It's just a matter of will. So... where can I vote for this?

There's a new poster ranking, quickest time for their coined phrase to become Word of the Year. Well done to the GOAT Andrej Karpathy

This great Ruxandra Teslo 🧬 post about drug discovery applies directly to material discovery. Better hypotheses not more hypotheses. The equivalent of human trials for materials is scaling it to economic viability, which is harder than finding good candidates

new Centre for British Progress piece w/ Pedro Serôdio on exit tax: tl;dr: - exit tax pushes founders abroad, killing new economic engine just as it matures - briefing is like causing bank run & kills any upside - tax rent-seekers, not those rebuilding British dynamism britishprogress.org/articles/kill-…

The Wylfa announcement today means it's time to commission excellent cross-sections again.

Fantastic to see an MP bring this kind of energy, and refreshing to see a minister getting hands-on with their portfolio. Captures the vibes without being cringe too, well done

Britain needs nuclear power. Our nuclear projects are the most expensive in the world and among the slowest. Regulators and industry are paralysed by risk aversion. This can change. For Britain to prosper, it must. Earlier this year, the Prime Minister appointed me to lead a

Reflecting the gut feeling of many, Ilya says “something important” is missing from current AI models. But what is the concrete nature of this chasm? One candidate: the difference between fractured entangled representation (FER) and unified factored representation (UFR).

Meet Mofasa, a step change in generative AI for Metal-Organic Frameworks 👏 Our all-atom generative model achieves SoTA fidelity on crystal structures as large as 500 atom, generating 40-46% valid, novel, and unique MOFs. By “rediscovering” experimental topologies and metal