

Fantopy
639 posts





Agent traces are the new fuel. Looking fw to announce `trl` official support for agent traces for training💥 (w/ `datasets` v5, coming out tmr?) Pick your local, synthetic, or community traces and train your own specialized Agent 🔜trl sft --dataset-name julien-c/synthtraces



NVIDIA Nemotron 3 Ultra now available in Command Code! Strongest US open model yet! 🍀 • 1M context • 5x faster inference • 550B MoE frontier-intelligence open model DEAL 2.3x usage 🎟️ $1 Go plan gets you ~$23 usage on Nemotron Woah, it's fast x taste compliance is great!






Really timely research from the ML systems community. It resonates with me a lot recently, it feels like building agent-friendly frameworks is quickly becoming a top priority. Curious to see how the open-source ecosystem evolves here(cc @PyTorch @vllm_project @sgl_project). Maybe human reviewers are the last line of defense, or maybe agents will eventually distill and replace reviewers altogether. idk again, really nice work, congrats @ruihanglai @tqchenml










This month, Poolside’s Laguna XS.2 is free to train on Prime Intellect Lab. First come, first serve while reserved capacity lasts.










This is the model of data center development we should all be advocating for. Google is building 1.4 GW of solar and wind capacity to power their 840 MW data center and using basically no water. This project also helps explain why Google paid $4.75 billion for Intersect Power, the developer of Meiter Energy Center. Intersect has gigawatts of clean energy capacity in the Texas interconnection queue. Thanks to a new rule passed last year, they can effectively share that interconnection with data centers. So rather than wait 5+ years to bring online a 840 MW data center, Google is going to bring it online in October 2027.


The great project of crypto is building an everything exchange. The point at which crypto ceases to be a category unto itself and declares victory is when there is a single permissionless venue for global access to all assets with portfolio margin and best execution