Third Moment Research Operates 10,000 Agents Simultaneously with Artificial Intelligence Systems
Third Moment Research deploys ten thousand agents simultaneously for research lasting thirty-six to forty-eight hours using OpenAI models.
Third Moment Research is building a self-improving artificial intelligence system capable of operating ten thousand agents concurrently for comprehensive research lasting thirty-six to forty-eight hours by integrating OpenAI's newest models.
Self-Improving Artificial Intelligence
Third Moment Research believes that top agents must specialize and improve their capabilities through use.
The company is developing artificial intelligence systems to understand agent behaviors, identify weaknesses, and automatically generate synthetic scenarios.
Supporting Models and Processes
OpenAI models such as Codex, GPT-6 Astra, GPT-6 Sol, and GPT-6.1 Sol support the system, enabling ten thousand agents to perform tasks simultaneously in reviews lasting thirty-six to forty-eight hours.
Financial Data Analysis and Optimization
In the field of quantitative finance, Codex helped identify the bottleneck in testing a trading strategy over one year of order-level US stock data.
Applied caching optimization and memory packing techniques reduced testing time from approximately three hours to three to five minutes.
Docker Environments and Model Capabilities
Third Moment provides each agent with a Docker-based environment containing data, documentation, and a research repository, while GPT-6 models offer reasoning support.
Codex hardware allows agents to move, formulate plans, set goals, spawn sub-agents, and narrow context.
Future Plans
Moving forward, the company plans to transition from individual research to hundreds of thousands of collaborating sub-agents powered by trillions of tokens.