Third Moment runs 10,000 agents for 36-48 hours using GPT-6.1 Sol

Third Moment Research is building a self-improving AI system that uses OpenAI's latest models to understand agent behavior, identify recurring weaknesses, and automatically create thousands of synthetic scenarios around those problems. The team repeats the process as new issues emerge, aiming for agents that get better the more they are used.

The company's early focus is quantitative finance, which offers a clear measure of success. In one instance, Codex analyzed thousands of recorded simulation trajectories and uncovered a bottleneck: backtesting a trading strategy across a year of order-level US equity data took several hours. After identifying a caching optimization and a memory-packing technique, the backtest time dropped from approximately 3 hours to approximately 3 to 5 minutes, reducing token use and context pollution.

Third Moment's roadmap extends beyond individual investigations to hundreds of thousands of collaborating subagents, with trillions of tokens supporting chains of dozens of dependent experiments. Co-founder Adam Khakhar says this would let humans focus on how to do research and optimize the research itself instead of actually doing it.