From OpenAI · · 1m 20s
Third Moment Research runs 10,000 agents for 36+ hours
Third Moment Research uses AI agents to run 10,000 experiments for 36+ hours.
In brief
Third Moment Research uses AI agents to run 10,000 experiments for 36+ hours. Third Moment Research runs 10,000 AI agents for 36-48 hours using GPT-6 models to improve trading systems and security. Originally reported by OpenAI.
Third Moment Research
Third Moment Research builds self-improving AI agents that focus on specific tasks and improve with use.
10,000 agents running concurrently
Third Moment runs 10,000 AI agents simultaneously for 36-48 hours per research session.
Codex reduces backtest time
Codex cuts backtest time from 3 hours to 3-5 minutes by optimizing trading strategy simulations.
GPT-6 models for research
GPT-6 Astra, GPT-6 Sol, and GPT-6.1 Sol power agents to analyze data, write code, and improve trading systems.
Agents improve trading systems
Agents test trading strategies on historical data to find meaningful improvements in performance.
Daybreak for security testing
Daybreak platform automates security testing to ensure agents can’t breach research environments.
Scaling to hundreds of thousands
Third Moment plans to expand from 10,000 agents to hundreds of thousands for broader business applications.
Token efficiency with GPT-6.1 Sol
GPT-6.1 Sol completes tasks with the fewest tokens while delivering performance similar to GPT-6 Astra.
Impact on trading systems
GPT-6.1 Sol identified non-displayed trades affecting exchange volume, an unexpected result for the research team.
The one thing
Third Moment Research runs 10,000 AI agents for 36-48 hours using GPT-6 models to improve trading systems and security.
