Lists (13)
Stars
The lightweight framework for building agents
Public Coworld CLI, Python helpers, manifest schemas, runner tooling, and reference worlds.
slime is an LLM post-training framework for RL Scaling.
Minimal codebase for efficiently training Natural Language Autoencoders (NLAs). Built on Celeste's nanoNLA: https://github.com/ceselder/nanoNLA
Interview questions for Computer Science faculty jobs
Frontier Models playing the board game Diplomacy.
Engineering Machine-Written Code: a modern software engineering textbook for building large, dependable systems with LLM coding agents
Natural Language Autoencoder (Anthropic NLA paper) on Qwen3-1.7B. With contrastive RL reward + batched sampling speedups.
✨ Personal plugins for Claude Code (& friends)... maybe worth sharing!
NVIDIA Object Oriented Agents: the Pythonic way to build AI Agents.
Convert sessions across harnesses to a unified trajectory format - designed to be consumed by agents (e.g. for memory formation, dreaming, search)
Data for "Value Leakage: An LLM's Answers Are Silently Shaped by Its Own Values"
A scaffold for automating science-of-evaluations research
A course for brains interested in brains (not tools interested in tools)
Companion code for the global workspace interpretability paper
An ARC-AGI solution using Agentica from Symbolica
A package for creating language-based environments.
Prompt Injection as Role Confusion
Efficient multi-token attribution for reasoning language models — Python package, CLI, and HTML token traces
minimal implementation of Natural Language Autoencoders (Anthropic, 2026)
NLA training pipeline for Qwen2.5-7B — full-layer activation verbalization with LoRA + TRL GRPO
Open-source library for scalable, reproducible evaluation of AI models and benchmarks.
Coding agent skills for Inspect AI and its ecosystem
Official codebase for "Next-Latent Prediction Transformers Learn Compact World Models"
AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI