Agents let AI safety share experiments hourly, not just papers monthly

Summary: Today, AI safety research is shared primarily at the scale of papers, creating collective feedback loops that take weeks or months. I propose an agent-based research approach that also shares progress at the scale of individual experiments, allowing agents and researchers to continuously replicate, extend, critique, and build upon one another's work. By increasing the granularity of collaboration, we can potentially reduce the collective research feedback loop to hours. We can start thi...