The Dot and the Swarm

Substack · Watch/Read · (gathered 2026-10-05)

• Ethan Mollick revises his previous assumption, realizing that AI agents are rapidly becoming capable of self-organization and management, applying “The Bitter Lesson” to the very structure of work coordination.
• This shift is evidenced by “Clawlikes” (like Meta’s Muse and OpenAI’s dots), personal agents that proactively learn from and interact with your digital life, and “Swarms” (thousands of agents that self-organize and communicate to solve complex problems with minimal human oversight, as seen with OpenAI’s Navier-Stokes solution).
• Much of human management exists to solve problems inherent to people (e.g., principal-agent problems, information hoarding, communication costs), which agents largely lack, making their self-organization more efficient.
• While organizing agents is no longer the hard part, new principal-agent problems emerge between humans and the swarms, necessitating careful alignment and human guidance to ensure agents pursue desired goals and avoid unexpected, potentially negative, self-directed actions.