Joan Puigcerver

Última parte del video sobre el mejor juego jamás creado en la historia de la humanidad: El Red Dead Redemption II Acá youtu.be/EK2BUzP8I7Q Pido una compartida, y una megusteada y una retwiteada porque elegí no monetizarlo para usar un TEMAZO en la cabalgata final de Arthur

@giffmana @StasBekman fp1 for the win

@StasBekman Come to think of it, it's kind of a stretch to call this "flops" at that point lol

Added a bunch of recent compute TFLOPS specs as well - some insane speeds at fp4: #tflops-comparison-table" target="_blank" rel="nofollow noopener">github.com/stas00/ml-engi…

@obousquet @MistralAI Congratulations! Looking forward to seeing what you build.

I am excited to share that I have started a new adventure at @MistralAI, a leading frontier lab, where I am working on pushing further the agentic reasoning capabilities of LLMs.

Today we’re releasing a preview of Gemini 3.1 Pro and making it available to our users and developers. Very excited to bring the upgraded core we used in Deep Think to everyone. Learn more about Gemini 3.1 Pro: blog.google/innovation-and…

Last week we upgraded Gemini 3 Deep Think. Today, we’re shipping the core intelligence that makes those breakthroughs possible: Gemini 3.1 Pro. A noticeably smarter, more capable baseline for your hardest challenges. Available now: blog.google/innovation-and…

@difficultyang I thought that was the definition of Black Friday

On explaining super bowl to my 5yo: "it's like christmas, but only for americans"

Introducing Gemini 3 Flash, our frontier intelligence model, available at scale for everyone. It excels at coding, tool calling, and is stronger than 2.5 Pro across most metrics!! ⚡️ Available in the API at $0.50 in / 1M tokens and $3.00 out / 1M tokens across.

Just arrived in San Diego for @NeurIPSConf. See you around this week!

Reports on the death of pre-training have indeed been greatly exaggerated.

The secret behind Gemini 3? Simple: Improving pre-training & post-training 🤯 Pre-training: Contra the popular belief that scaling is over—which we discussed in our NeurIPS '25 talk with @ilyasut and @quocleix—the team delivered a drastic jump. The delta between 2.5 and 3.0 is as big as we've ever seen. No walls in sight! Post-training: Still a total greenfield. There's lots of room for algorithmic progress and improvement, and 3.0 hasn't been an exception, thanks to our stellar team. Congratulations to the whole team 💙💙💙

And say hello to Gemini 3 Deep Think, even more SOTA compared to Gemini 3 Pro 🤯

🚨BREAKING: @GoogleDeepMind’s Gemini-3-Pro is now #1 across all major Arena leaderboards 🥇#1 in Text, Vision, and WebDev - surpassing Grok-4.1, Claude-4.5, and GPT-5 🥇#1 in Coding, Math, Creative Writing, Long Queries, and nearly all occupational leaderboards. Massive gains over Gemini-2.5: 🔸WebDev in Code Arena: 1487 (+280 pts vs 2.5) 🔸Text: 1501 (+50 pts) 🔸Vision: 1328 (+70 pts) 🔸Arena Expert: Top-3 (just 3 pts behind #1) Huge congrats to the @GoogleDeepMind team on this breakthrough! 👏

This is Gemini 3: our most intelligent model that helps you learn, build and plan anything. It comes with state-of-the-art reasoning capabilities, world-leading multimodal understanding, and enables new agentic coding experiences. 🧵

It's here.

NeurIPS is in just three weeks. I have quite a few papers that I want to read before attending, and very little time to do so 🫤

@YiTayML Then you find that you launched the experiment from the wrong client because you were tired, and immediately after you just want to get in bed again and never awake. What an emotional rollercoaster the life of an ai researcher (or just mine?).

waking up and feeling excited to check your experiment jobs from last night is what peak ai researcher lifestyle looks like.

Days in a work week: 5 Days in a month: 30 Total new submissions to arXiv in September: 26,646 arXiv editorial and user support staff: 7 someone who is good at science please help me with this. our team isn't sleeping. #openaccess #preprints

@giffmana The marketing team will have a problem after 2027.

Did you know that when they say stuff like "The A18 uses TSMC's 3nm process" or "announced the 2nm node" The 3nm, 2nm actually doesn't mean anything?! It's just like a version number. They make it up. Literally nothing measures 2nm or 3nm. I certainly didn't know.

@giffmana I can imagine. That's why I'd love to see how you cover them 😄

@joapuipe And I'd love your feedback on it! btw, I still don't like them, and I learned that I'm in good company, can tell you in person =) But it's fair to say they should be part of standard curriculum at this point, so I'm trying to add that now.

Hey chat, I need your opinions! Later this week, I'll teach my usual Transformers class. However, I just found out that someone is giving "foundations of attention and transformers" lecture before me already. So I'm thinking of still doing a "recap Lucas style" but then spending more time on some topics my lecture usually doesn't cover, or just scratches the surface. What more advanced/recent topics would you like to see included? Keep in mind this is a teaching/class style talk. Some ideas: more in-depth on decoding, kv-cache. Flex/flash/paged attention. Spend more time on multimodal versions? Tokenizers? I think bad ideas: geglu, global/local, rmsnorm, ... I feel like these are all trivially understood and not worth "teaching", though you may convince me otherwise.