With the explosion of Microduck and Reachy Mini, we'll need more people to join the @pollenrobotics @huggingface teams in all functions. If you're excited about democratizing robotics AI with open-source, my DMs are open!

Moonshot removed cyber-related data from training, as "without data removal [...] downstream cyber capability would increase substantially" Is China stopping the cybermaxxing?

new fav personal assistant use case unlocked! i was annoyed that the roads near my house in HSR were dug up for MONTHS making it hard to even walk and it felt like there's nothing i could do about it! early sept i asked faff to find a way to flag this to the authorities...

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

We trained our Muse foundational models, including Muse Wonder, at Prometheus, our gigawatt-scale supercluster in New Albany, Ohio. Across a supercluster with hundreds of thousands of GPUs, even one underperforming host can slow an entire training run. We engineered Prometheus for resilience at scale, achieving 85%+ goodput on pretraining, 90%+ on reinforcement learning and ~47K GPU hours mean time between interruptions during Muse Wonder training. Take a look at the tech powering Prometheus: go.meta.me/4yXXApZ

the holmes story is used *all* the time to discredit me + my team's legitimate technical accomplishments it's unbelievably frustrating + angering, and it has made what is already a very hard company to build significantly more challenging so now feeling pity for her is strange

AI agents are changing how we work and collaborate, but documents are still stuck in the pre-AI era. Agents can now create beautiful, interactive decks, dashboards, and reports in HTML. But the moment you want to collaborate on them, everything breaks down. Google Docs wasn’t built for rich, interactive HTML or AI agents. Feedback gets scattered across Slack and email. Every edit means going back to the agent, generating a new file, and sharing another link. No real-time collaboration. No single source of truth. We built SlashDocs to fix this. Think Google Docs, but built for AI agents and the documents they create. Host HTML and Markdown in one central hub. Share documents, manage permissions, and comment directly on the content, just like Google Docs. Agents can read feedback and update the same document in real time. No duplicate files, no endless back-and-forth, and everyone stays in sync. What started as a small internal experiment in one team has now scaled across @Razorpay. 28,000+ documents created. Over 50% of all documents generated at Razorpay now live on SlashDocs. And today, we’re opening it up to everyone. Self-hosted. Agent-agnostic. Fully open source. AI agents have changed how we create documents. It’s time we changed how we collaborate on them. Try it: razorpay.com/slash-docs/

At Proximal, we believe that building our own research infrastructure is crucial to design the highest-quality training data Today, we are sharing more information about our internal post-training stack

The recent breakthrough in Navier-Stokes has garnered a lot of attention to the new possibilities that emerge when agents work together. We have been interested in this question for a while. How can a team of agents achieve more than agents working alone? I'm excited to finally share our new work, "Self-Organizing Agent Teams Learn to Reason Together" 🧵