Habemus GPT-5!

Little Bobby “Ignore All Previous Instructions.”

What it feels like to select o3 in the Legacy Models menu.

Non-cytotoxic RNA editor just dropped.

AI efficiency is important. Today, Google is sharing a technical paper detailing our comprehensive methodology for measuring the environmental impact of Gemini inference. We estimate that the median Gemini Apps text prompt uses 0.24 watt-hours of energy (equivalent to watching an

One Gemini request ~ 9s of TV. I’ll take that trade any day! Kudos to GDM for publishing detailed environmental impact data for their AI products.

Lead paint was displaced not by regulation but by a cheaper and better substitute. The same dynamic is the only realistic path for a global energy transition.

💯 this!

LLMs are absolutely wild. Even if all progress stopped *right now* it would take us years to probe the insane quirks of how these systems work.

Very helpful on long runs! 🏃🏻⛰️

Every bit of reality around you would not have seen the light but for creative engineering efforts, and sheer, dogged persistence. A museum of passion projects. Fascinating.

There is so much to do in AI/ML it's inspiring.

The French PM, Francois Bayrou, is the first European politician of his cohort, as far as I can tell, to successfully frame the debate about fiscal sustainability as a matter of intergenerational equity. The generation that has had a great run leaving a huge bill for their

If you're not following Kexin Huang you're missing out on some of the most interesting work in AI for science, from Biomni to his fantastic hypothesis generation/invalidation paper with Jure Leskovec. Congratulations Dr Huang!!! 🥳🎓

The goosebumps 🤯

Tom Evans (@ASICSEurope) wins the 2025 #UTMB in 19:18:58. Enjoy our pre-race interview with Tom: i-rn.fr/25UTMB-Tom

Watching the finish line of HOKA UTMB® Mont-Blanc is always special. What a race! What incredible runners! 😍🏃‍♀️🏃‍♂️🏔️

#Otd in 1955, the term “artificial intelligence” was coined in a conference proposal: stanford.io/2WJJJGN

On a quand même drôlement progressé depuis Victor Hugo, non ? Voyez comment il décrivait l'Assemblée : « Elle est presque entièrement composée d'hommes qui, ne sachant pas parler, ne savent pas écouter. Ils ne savent que dire, et ils ne veulent pas se taire. Que faire ? Ils font

Reinforcement learning leads to better AI scientist agents! 🚀 By training models end-to-end with multi-turn RL, we’re seeing breakthroughs in reasoning and problem-solving for real biomedical research. Excited to introduce Biomni-R0 — an agentic LLM trained with this approach.