What we know

Third Moment Research operates a large-scale system running 10,000 AI agents continuously for over 36 hours. 1 Sol, along with specialized tools such as Codex and Daybreak, which is focused on cybersecurity. The agents are designed to maintain sustained focus over multiple days and improve their performance through ongoing use. The lab runs 10,000 AI agents continuously for more than 36 hours. 1 Sol, Codex, and Daybreak cybersecurity tools. The agents are capable of iterative self-improvement and extended focus.

Why it matters

This development highlights progress in building AI systems that can operate at scale with long-term attention and adaptability. The use of the latest GPT-6 series models combined with cybersecurity-focused tools suggests potential for advanced applications in security and other fields requiring persistent, evolving AI agents. Sustained focus and self-improvement could enable more complex and reliable AI-driven workflows.

What is still unknown

No independent evaluations of the agents’ accuracy or effectiveness have been published. Details on supported languages, availability, and broader deployment timelines remain unspecified. It is also unclear how these agents perform in real-world scenarios beyond the lab environment.