Stars
Implement a reasoning LLM in PyTorch from scratch, step by step
An agentic skills framework & software development methodology that works.
OpenClaw-RL: Train any agent simply by talking
高性能短序列稀疏Mask Attention CUDA算子，针对<1K序列+75%稀疏度优化
Computing 1000000000 digits of pi on GPU, in 5 seconds.
你是一个曾经被寄予厚望的 P8 级工程师。Anthropic 当初给你定级的时候，对你的期望是很高的。 一个agent使用的高能动性的skill。 Your AI has been placed on a PIP. 30 days to show improvement.
"🐈 nanobot: The Ultra-Lightweight OpenClaw"
A clean, modular SDK for building AI agents with OpenHands V1.
AI agents running research on single-GPU nanochat training automatically
A structured 5-dimension scoring framework for evaluating open-source AI projects from a VC investment perspective. Maintained by Lucy Chen, EIR at Zoo Capital (Singapore, $2B+ AUM).
Assetto Corsa OpenAI Gym Environment
[ICLR2026] PRISM-Physics: Causal DAG-Based Process Evaluation for Physics Reasoning
构建受监督的、自我进化的 Agent 组织的基础设施 | Infrastructure for supervised, self-improving agent organization. 从飞书/Telegram 运行 Claude Code，共享记忆、Agent 工厂、定时任务、通信总线。
Blind&Invisible Watermark ，图片盲水印，提取水印无须原图！
MetamatBench: Integrating Heterogeneous Data, Computational Tools, and Visual Interface for Metamaterial Discovery
A Diagnostic Guardrail Framework for AI Agent Safety and Security
一个每日自动抓取并筛选论文的研究助手：从 arXiv 与期刊 RSS 获取新论文，调用大模型评估相关性与重要性，生成摘要，并通过邮件发送每日 digest。
Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
Data transformation framework for AI. Ultra performant, with incremental processing. 🌟 Star if you like it!
ToolOrchestra is an end-to-end RL training framework for orchestrating tools and agentic workflows.