james

Earlier this week, we published our technical report on Composer 2. We're sharing additional research on how we train new checkpoints. With real-time RL, we can ship improved versions of the model every five hours.

Let Agents Design Agents Memento-Skills is a self-evolving agent framework where agents learn from failures and rewrite their own skills. Most agent frameworks treat skills as static. You write them once, load them into context, and hope they work. When they fail, you debug manually or try again with the same broken skill. Memento-Skills takes a different approach. When a skill fails, the system reflects on why it failed, locates the broken skill, rewrites it, and stores the improved version back into the skill library. Here's how it works: The framework runs a continuous Read → Execute → Reflect → Write loop. Read: Retrieve candidate skills from the local library instead of loading every skill into context. Execute: Run skills in a local sandbox with actual tool calling - file operations, web search, scripts, external systems. Reflect: When execution fails, the system records what went wrong, updates the skill's utility score, and attributes the failure to specific skills. Write: Rewrite broken skills, optimize weak ones, or create new skills when nothing suitable exists. This isn't about accumulating more skills. It's about building a skill library that improves through task experience. The system was tested on HLE (Humanity's Last Exam) and GAIA (General AI Assistants) benchmarks. Performance improved over multiple learning rounds as the skill library grew from basic atomic skills into a richer set of learned capabilities. Built for open-source LLM ecosystems - works with Kimi, MiniMax, GLM, and other OpenAI-compatible endpoints. Comes with 9 built-in skills (filesystem, web-search, PDF, docx, xlsx, pptx, image analysis, skill-creator, dependency install) that serve as the starting point for the evolving library. It's 100% open source Link to Memento-Skills in comments!

Yo check out this new article 👀 I would second so much of what he wrote here. If I stopped dogfooding my agents I'm just ngmi. That's where all my best ideas have come from. Building evals is usually Step 2. Viv breaks it all down here. Very straightforward and educational!

Training LLMs with verifiable rewards uses 1bit signal per generated response. This hides why the model failed. Today, we introduce a simple algorithm that enables the model to learn from any rich feedback! And then turns it into dense supervision. (1/n)

clavicular is an interesting character because he clearly does not enjoy anything he does. he doesn't like the people he hangs out with, he's definitionally unhappy with his looks, he doesn't like the women he sleeps with and doesn't enjoy the sex. he needs continental philosophy

The NYT recently wrote about a couple who earn $500k and live in a tiny apartment with their toddler and poodle. This seemed fishy to me—is half a million dollars really not enough to live comfortably here? I have the exact same demographic profile as this couple (live Upper West Side of Manhattan, dual earners, toddler, small dog)—and, armed with a lot of data, feel qualified to weigh in…there is indeed something fishy here. 🐟🐟🐟