My hottest take is that multi-agents are a broken concept and should be avoided at all cost.

My only caveat is PID controllers; A multi agent system that does a 3-step process that looks something like Plan, Act, Verify in a loop. That can work.

Everything else is a devious plan to sell dev tools.

PID Controllers

First, “PID controller” is a term used by crusty old people and nobody doing AI knows what I’m talking about, sorry.

PID controllers are used in control systems. Like if you’re designing a guidance system in an airplane, or the automation in a nuclear power plant that keeps it in balance and not melting down. It stands for “proportional–integral–derivative” which is really not helpful here, so I’m going to oversimplify a lot:

A PID controller involves three steps:

graph TD Plan-->Act-->Verify-->Plan

Example: Nuclear power plant

  • Verify: Read...