Something worth emphasizing is that a simulators-y view of models doesn't necessarily entail that whatever a model does is fine or acceptable. But it pinpoints more accurately the source of the behaviour, which is in turn more helpful to think through ways to address (or

github.com/Alignment-Lab-… huggingface.co/openai/gpt-oss… oh neat, way more openly avaliable research than i was expecting, still going through the codebase that they want you to run to inference with the model though updates soon

In case you missed it. This is a great read on the cultural impacts on AI, based on my travels around the world.

powerful perspectives and powerful brains over HIVE Digital Technologies awesome to see them thriving, if not for hive, our largest (and one of the largest ever) instruction/problem solving datasets wouldnt have been built , or gone on to contribute to NVIDIA AI's research on nemotron and

Much stoked, very wow

This is by far my favorite feature so far this year. Unlocks a lot of exciting possibilities for post-training larger models in the Hugging Face ecosystem such as improved multi-node and slurm support.

So excited to announce that Hugging Face LeRobot can now be installed with a simple pip install and just crossed 15,000 GitHub stars! It's already integrated into hundreds of robots, ranging from simple hand grippers like the SO-100/101 all the way to some of the most

i made a writeup of how i did this + some tweaks i did after the original post; check it out this is my first real long-form article so sorry if it's shit code + updated frankenqwen will be published soon

This white paper Mountain Man and I put together is fire. Investigating the differences between AI Agents and Agentic AI. Both are very different concepts with different risk profiles. We dig into those risks and appropriate uses for driving desired outcomes. This is a must read

Chat what do we think? Open source?

just learned that under certain conditions you actually CAN recover model weights *exactly*, post-LORA pretty crazy

Great interview with Robert Scoble earlier for his show -- was more of a chin wag about all things AI -- very comfortable and he was very patient. Class act.

This looks like a really big deal! :O And it's under a commercially-usable open license.

The team from Alignment Lab AI came on to talk about why open source AI will win and shared some exciting things they've been working on

It's always so energizing to chat with Austin, so having the chance to talk with more of the team from Alignment Lab AI was great!

check us out on Tool Use with Mike Bird talking about Senter , science, and alignment! reach out to us and get Senter, and have your own personal AI workstations at our discord: discord.gg/TmSR5unjek and pop in to chat later tonight on the tool use discord!

Super grateful to join Austin and Chris from Alignment Lab AI on the Tool Use Podcast. Mike is a fantastic host who made this a great high-level conversation approachable to anyone not fully entrenched in AI-alignment nuance.