Simple yet effective methods for AI4AI.

Website · Simple Long Horizon Agent · Long-Horizon Agent Survey

Simple Agent Lab is an open AI research collective. We are building toward a future where AI can help build AI, while the methods behind it remain understandable to humans.

AI will increasingly write code, generate data, run experiments, evaluate results, and improve models and systems. Improved AI can then participate in building the next generation, forming a recursive loop of improvement.

• AI4AI — using AI to improve models, data, training, evaluation, and AI systems.
• Self-Improving Systems — systems that turn feedback into improvements and carry what works into the next iteration.
• Recursive Self-Improvement (RSI) — the path toward systems whose improvements make them better at producing further improvements.

Simple does not mean unsophisticated. Simplicity is what allows humans to continue understanding and participating in a future built by AI. We want people to see how a method works, verify why an improvement happened, and help decide what comes next.

• Simple Long Horizon Agent — a minimal, understandable agent loop for real long-horizon work.
• Building Reliable Long-Horizon Agents — a survey of definitions, metrics, benchmarks, and system design for reliable agents.

We share code, experiments, evaluations, and failures so others can inspect, reproduce, and build on what we learn. Our members bring experience from ByteDance, UC Berkeley, Tsinghua University, Shanghai Jiao Tong University, and Tongji University.

AI4AI is our direction. Recursive improvement is our path. Simple is our principle.