🧑‍🎄🎄 Christmas Foundation Model Release: scPRINT-2 *One-liner:* a 20M-active-param single-cell foundation model trained on **350M cells / 16 species / 300 tissues / 500 cell types**.

Every French🇫🇷 passenger train, on an ordinary Tuesday, from first departure to last

Fait notable : dans le calcul du "niveau de vie", on divise le revenu total du ménage par des unités de consommation (UC). Un retraité compte comme plus d'UC qu'un enfant. Donc, l'INSEE considère qu'un enfant DOIT recevoir moins qu'un retraité. #gérontocratie

We’re in! See you @NeurIPSConf! (thread coming soon) arxiv.org/abs/2605.28075

Genomes are full of dark matter of unknown functions. We present Minerva, a new method for discovery guided by genome language models. Minerva reveals the interactions hidden in non-coding DNA, pointing to hundreds of new putative RNAs and repetitive elements per bacterial genome. Minerva allows us to find and study elements invisible to traditional methods at orders of magnitude greater scale than before.

An example of the problems EU Inc will face by keeping "notarial control" in EU Inc article 14. Official response from the German Chamber of Notaries: If you're not German or not European living in Germany, you can't use online digital tools. Oh, and you can't use the digital tools you want: we also want a monopoly on software. How come the @EU_Commission authorizes a monopoly on software to notaries? How come EU Inc would reenforce and strengthen the notary monopoly and enable a huge company creation bottleneck by forcing a numerus clausus cartel to review EU Inc creations? Not on our watch Repost. We hope @EP_Legal cancels the "notarial control" in the final EU Inc text. PS: We believe MEPs and EU Commission are doing a great work on EU Inc. Yet this part of the draft is the one that needs to be fixed in the final voted law. We want them to open their eyes on all the issues this creates.

Yet another 🤯 AI moment

A year ago at ICML, I met with Dan and we discussed the future of virtual cell models, and in particular the gap between the promises of the field and the reality of generating the data needed to get there. When I discovered Somite, now Cellular Intelligence, I immediately saw a strong alignment with where we thought the field needed to go. I’m therefore very happy to share that I’ve joined Cellular Intelligence to help build better models of the cell and, ultimately, transform cell therapies. The challenges ahead are enormous, but so is the potential. Ad astra! ⭐

Second Ph.D paper published ✅ : iopscience.iop.org/article/1… After presenting this work at the ICML workshop on Foundation models for Life Science, I am happy for its open access publication. It is a paper of big ideas supported by yet small evidences. However, I believe it is the first work on this topic. For similar and richer works I would also now point to the PULSAR work from Pang, Rosen et al. pmc.ncbi.nlm.nih.gov/article…

Five pharma companies just fine-tuned OpenFold3 in a federated setting on 20,000+ structures... ... raising correct ligand-pose predictions from 29% to 47%

We got really good at DNA sequencing a long time ago. Ever since, any problem that can be turned into a sequencing problem can be magically solved, giving rise to *-seq methods that form the backbone of modern genomics. This team at Escalante has done this for protein binding kinetics. A million interactions in one tube, one overnight reaction, one sequencing run. Worth the read... blog.escalante.bio/how-to-se…

Actually , while we are on this subject. What are the worst threat we might face with AI, how would they pan out? According to Astra (and I would 100% agree) 1.Concentration of power and AI-enabled authoritarianism 2.Severe economic disruption and inequality 3.Cyberattacks and critical-infrastructure disruption They seem so obvious. Who gains from not focusing the discussions and policies on that?

I would add actually that nature is running an optimization algorithm 24/7 since the dawn of time on all living humans to find the worst virus we can think of. An LLM could find some too, but if this is an existential threat, this is one we have faced so many times and are going to face again and again, LLMs or not. They need to find a better existential threat bc this one doesn't really make sense.

Thank you for injecting us with a dose of realism vaccine against the AI doom virus.

Alice Weidel becomes the first leader in the world who is anti-immigration but has an immigrant wife, is anti-LGBTQ+ but is a lesbian herself, opposes LGBT adoption but has adopted two children, and is a hardcore German nationalist but lives in Switzerland

Navier Stoke results drama with openai seeming to scoop on purpose

🚨 BREAKING: OpenAI releases new Astra model, says it may represent AGI "Welcome to the AGI era" axios.com/2026/09/03/openai-…

New OpenAI repo with a Lean formalization by GPT-6-Astra proves that there are infinitely many pairs of consecutive primes whose distance is at most 186 github.com/openai/PrimeGaps1…

Probablement peu de temps après son élection 🤔