Third Moment Research says its agents can run for 36+ hours
OpenAI says Third Moment Research builds long-running AI agents using GPT-6 models for cybersecurity and workflow improvement. (Read More)
Asanat Analysis — Why it matters
Third Moment Research's claimed 36+ hour agent runtime represents a meaningful step toward autonomous systems capable of handling multi-day workflows without human intervention. In crypto contexts, this matters for security monitoring, arbitrage execution, and liquidation management—tasks where continuous operation without checkpointing reduces latency and operational overhead. The use of GPT-6 indicates these aren't simple state machines but reasoning systems that can adapt mid-stream, a significant shift from rigid automation.
The cybersecurity angle is particularly relevant to DeFi infrastructure. Longer-running agents could enable more sophisticated threat detection and response, potentially reducing the response time to exploits from hours to minutes. However, the 36-hour claim warrants skepticism around actual runtime stability, memory efficiency, and whether these agents maintain decision quality under extended operation—problems that have historically plagued autonomous systems in production environments. If validated, this capability could reshape how protocols manage security and treasury operations, but deployment in high-stakes DeFi contexts will require rigorous testing and fallback mechanisms.