It was an honor to speak about #AI to the Italian Parliament today.
I’m particularly excited about the applications that generative AI can have in healthcare and biology, by augmenting, not replacing the dwindling and overworked workforce in Italy.
Full video and summary ⬇️
@ArjitJaiswal7 It's back! I was planning to present live rendering from Claude Design .. after this experience...HECK NO. Will save pptx even if not pixel perfect.
Every drug × every target
📊 228 targets
💊 1,397 FDA-approved compounds
🔬 558K interactions
@EvEBiotech's largest data release ever adds 29 new GPCRs in the Gs pathway + 1 kinase (PDGFRa-D842V)
All on @huggingface 🤗
Today we’re launching a £100m call to double down on ARIA's AI Scientist and Activation Partners initiatives
We're looking for partners to bring advanced AI capabilities to R&D alongside translational expertise to help us turn speculative ideas into world-changing capabilities🧵
The coolest meeting I had this week with was Paul, who used ChatGPT and other LLMs to create an mRNA vaccine protocol to save his dog Rosie. It is amazing story.
"The chat bots empowered me as an individual to act with the power of a research institute - planning, education, troubleshooting, compliance, and yes, real scientific design work in converting genomic data to a vaccine prescription and designing the treatment protocol around it. But they worked alongside humans at every step. The combination is what made it possible."
It immediately got me thinking "this should be a company".
Also, Paul is an extraordinary guy. This should be easy to do, but it is not yet.
@awnihannun Feels like something more structured is needed that encodes relationships (eg a KG) but then how to update and have the LLM reason effectively on it?
@awnihannun How are you thinking about the data structure supporting the memory? Bunch of .md files feels suboptimal. I’m finding that in apps with complex relationships (eg, AI chief of staff, biology lit review) I can’t get hallucinations to go to zero no matter how I re-organize memory.
FDA is now open to Bayesian statistical approaches. A leap forward!
Bayesian statistics can help:
✅ Clinical trial design
✅ Finding the optimal dose
✅ Extrapolation to children
✅ Leveraging phase 2 results in phase 3
2026 will be the year of agents in biology. But we need better benchmarks.
We worked with scientists to turn real world analysis into verifiable problems. SpatialBench stratifies frontier models, shows harnesses matter, and reveals distinct failure modes between model families:
To me this is the _current_ AI superpower in science. Not out-reasoning experts after a decade of depth, but reclaiming the breadth that decade inevitably cost.
GPT-5 came back with an elegant solution. What struck me: the proof completely bypasses the rich, well-developed theory usually applied to these invariants, instead drawing on techniques from a different corner of algebraic geometry.
Sometimes an outsider's perspective helps.
Live from #NeurIPS2025!
We're connecting with top talent and showcasing organizations at the AI frontier: @ARIA, @AISecurityInst, UK SovAI, @HuggingFace, and more.
Visit our booth this week for meet-and-greets, lightning talks, and a special Q&A with @tkalil2050.
We are up to 688 LoCITT participants!
If we could enroll all 1,000 participants within a month of launch (by 11/30), it would send a powerful message about the demand for remote LC clinical trials.
We appreciate the community's help to spread the word:
https://t.co/K7yjhCfV3N