i do think the future of work is like starcraft or age of empires. you have 200 microagents you’re directing to fix problems, gather information, reach out to people, design new systems, etc.

Mistral releases Magistral, their new reasoning models! Magistral-Small-2506 excels at mathematics and coding. You can run the 24B model locally with just 32GB RAM by using our Dynamic GGUFs. GGUFs to run: huggingface.co/unsloth/Magist…

My team at IGN France French Geographic Institute just released new public dataset FLAIR-HUB on Hugging Face Large-scale Multimodal Dataset for Land Cover and Crop Mapping (aerial, satellite, multi annotations) with code, pretrained models, data paper! Huge work! Congrats

🔥Big exciting news - I've started a new company! 🚀 We are building AI agents that take actions in the real world by orchestrating the movement of physical goods. We're working with our first partners and are now growing the founding engineering team. We're building in NYC,

Exciting PhD position open at FAIR in Paris. We are looking for a candidate to join our team and contribute to advancing the field of AI, especially reinforcement learning. Find more details and apply below. Feel free to reach out to me by email. metacareers.com/jobs/192266079…

For RL folks in Paris

Some of the links: - My slides as keynote: drive.google.com/file/d/1a0h1mk… - Software 2.0 blog post from 2017 karpathy.medium.com/software-2-0-a… - How LLMs flip the script on technology diffusion karpathy.bearblog.dev/power-to-the-p… - Vibe coding MenuGen (retrospective) karpathy.bearblog.dev/vibe-coding-me…

feels like a big issue with chat interface for llm usage is the lack of clear affordance, i.e. what the user can or cannot do with the tool. What's the current "SOTA" on building interface with clear affordance for llm based service ?

Our new state-of-the-art AI model Aeneas transforms how historians connect the past. 📜 Ancient inscriptions often lack context – it's like solving a puzzle with 90% of the pieces lost to time. It helps researchers interpret and situate inscriptions in their past context. 🧵

casually dropping a box link to DL their OST in mp3 for free is the gaming equivalent to mistral tweeting a torrent to DL a SOTA model. Insane community contribution

i'm looking to scale the humans in my team @ Mistral AI eng. only, no AI or science involved dm me with credentials if you're interested RTs appreciated

love this rich format blog post with actual data points in the article

Starting today, Psyche will train 6 new models in parallel in pursuit of creating world class open source AI. These runs serve as the starting point for future experiments and a more thorough and empirical training process.

interesting to see Z dot ai propose quarterly plan for their claude (clawd?) code API like code agent. No one serious is committing to a yearly plan on a model provider given monthly SOTA rotations among labs

Thrilled to announce Mistral AI's new formal math team following our $2B funding round. AI for formal math researchers wanted. We offer: - Work on creating the state-of-the-art prover, autoformalizer, and automatic proof agent all in one model - World class team (with

interesting to see jetbrains trajectory with their products being severely challenged by a IDE-less future with terminal based dev - they're reaching out to their user for feedback on in-IDE LLM tools usage and pretty happy with their progress last semester tbh

big indeed. A new dimension for training data collection with clearer reward definition. Exciting times for the oai RL folks

"reinforcing the absolute criticality for insanely high talent density"