The Anatomy of Discovery: Our new paper in @MRSBulletin shows how GNNs + reasoning-driven massively agentic AI can autonomously design alloys - linking atomic vibrations, dislocation physics & macroscopic strength. Beyond prediction, this is an important step toward AI that

Paper: A. Ghafarollahi and M.J. Buehler, MRS Bulletin, 2025,

Really happy to share our new paper on using AlphaEvolve for mathematical exploration at scale, written with Javier Gómez-Serrano, Terence Tao, and @GoogleDeepMind's Bogdan Georgiev. We tested it on 67 problems and documented all our successes and failures. 🧵

This paper shows that even the most advanced LLMs/reasoning models cannot yet know what they believe as they collapse these into one surface. Until AI learns to differentiate, it cannot genuinely discover - science is, after all, the art of differentiation! The frontier is that

Is 3D dragging you down? Wish you could instead use the 2D ColabFold representation for all your work? 🤓 Introducing: py2Dmol 🧬

Speaking at #CDFAM tomorrow: how AI composes across scales - from protein and material design to music and manufacturing. A new nexus of creativity where biology, engineering, and human ingenuity collide!

We’ve built AI that learn from the past - from data, labels, and memory. But discovery begins where memory ends, and the next intelligence must learn through creation, not recollection, and explore never-before-seen data and phenomena - growing knowledge instead of replaying it.

Paper:

GitHub repo with code, part list, assembly instructions, etc.:

Nature produces incredible biomaterials - think plant fibers or insect exoskeletons - which we are often challenging to mold into shape and form for de novo engineering applications. Introducing BEAVER: An open-source and low-cost dual-extruder 3D printer for macroscale biotic

Excited to announce that we’re hosting the IROS 2025 AIR4S Workshop “Embodied AI and Robotics for Future Scientific Discovery” on October 24, 2025 (9:00 AM – 1:00 PM, GMT+8) at the Hangzhou International Expo Center – Pressroom A (210A). Invited Speakers - leading experts from

Link to source article:

Today’s LLMs excel at statistical pattern-matching but they often struggle to reason from first principles, to integrate data, physics, and symbolic knowledge into coherent scientific understanding. In my plenary at #SciFM2025, I outlined a framework for AI that can think -

DMSE together with @MIT_SCC seeks candidates for a tenure-track Assistant Professor with experience in combining fundamental scientific principles with algorithmic innovations to empower understanding and synthesis of materials. Applications close Nov 30:

Excellent paper and thought leadership! Thank you @DanHendrycks and collaborators. The work highlights many important features still missing in today’s AI such as interactive intelligence, perception, adaptability and memory. A key component is the ability to abstract and

I’m biased but I want to see way more of this AI + cancer research + open-source thinking long-term: science research has been most beneficial and successful when it involved a community exchanging and fostering ideas. We have to keep this alive in a world of increasingly

After a formative time at OpenAI, I’m launching Maison AGI (@maisonagi), a fashion house creating cultural artifacts for the AI era. Our first collection, Relic of Thought, is a collaboration with Ilya Sutskever (@ilyasut), featuring his original artworks alongside a signature

Nice work!

Do co-folding models learn physics—or just memorize pockets? Deep models like AlphaFold3 and RoseTTAFold All-Atom can place small molecules into protein pockets with eye-popping accuracy. But when chemistry changes, do they still obey sterics, electrostatics, and chemistry—or