One good loop or many agents? What actually works in practice

Everyone's building swarms of agents. Most of the time one well-built loop beats five agents talking past each other. We're working out where the line actually is.

FIRST TIME? This is a small, friendly group that meets twice a month at Schuler Books on 28th Street, in the Chapbook Cafe. Free, no cover, nothing to prepare. Turn up, grab a coffee, and talk shop.

WHAT WE'RE DIGGING INTO

The cases where splitting work across agents genuinely wins - mostly wide, read-heavy research - and the cases where it fragments context and produces mush. Why coding is the classic place multi-agent setups fall apart. The token cost multiplier nobody mentions. Orchestrator-and-workers versus a single agent with good tools. How these systems fail, which is usually coordination rather than capability.

WHO THIS IS FOR

Anyone building with agents who's wondered whether to add more of them. People who tried a multi-agent framework and found it worse than a plain loop. Beginners welcome - the useful version