UMA is launching to build general-purpose mobile and humanoid robots from Europe. Join us: uma.bot

I’m joining UMA as a Founding Research Scientist. Already, LLMs broaden what robots can do and improve their robustness in the real world. General-purpose robots are coming. Our job is now to find the best path. Follow our journey at UMA Join us uma.bot

Un nouveau venu européen 🇪🇺 (et même français 🇫🇷 !) à suivre de près dans l'industrie des robots 🥳 : UMA, créée par des anciens de Tesla, HuggingFace, Google, avec une équipe internationale... "Nous concevons des robots mobiles et humanoïdes polyvalents dotés d'une

Big day: UMA is live @uma_robots 🦾 Super excited to be on the founding team! Robotics is one of the most interesting challenges of today, and it is riding a massive tailwind. Feels like AI is finally leaving the screen! On a mission to gather data, build infra and train those

We’re only scratching the surface of general-purpose robots doing economically valuable work. LLMs have exploded once scaled. Robotics hasn’t had this moment yet. New ideas, properly scaled, can get us there and unlock a huge amount of value. If that excites you, reach out !

I’m really happy to share that we’re launching UMA. Together with Remi Cadene, Simon Alibert, Rob Knight, and an exceptional founding team, we’re building general-purpose mobile and humanoid robots. If you want to be part of this adventure, reach out at uma.bot

debug mode with Cursor is crazy, period.

debug mode with Cursor is crazy, period.

lol

VL-JEPA is out! A non-generative vision-language model, based on JEPA. Different from typical data-space autoregressive VLMs, VL-JEPA is trained to predict continuous embeddings in the latent space. (1/N)

We’re releasing Action100M: the largest open dataset of ~15 years of video with dense action + caption annotations. It's a key ingredient behind VL-JEPA, now open to fuel the next generation of VLMs, World Models, and Robotics policies. Dataset on HF: huggingface.co/datasets/faceb…

🔥I am super excited for the official release of an open-source library we've been working on for about a year! 🪄interpreto is an interpretability toolbox for HF language models🤗. In both generation and classification! Why do you need it, and for what? 1/8 (links at the end)

We opened a new role at UMA! Apply on our website 😇

squill.dev is ready for me to share: it's a canvas for writing SQL. It's free and fully open source. I made what I wished I had all these years of churning out SQL. Enjoy! There is a pro for AI features and such, but it's still early days.

It took me a while to find a proper way to have a Plaid index that can be updated incrementally. Finally found the recipe. When updated one document at a time, NextPlaid mostly behave the same way as when updated in batch