A Million Agents Is a Distributed Systems Problem
A Million Agents Is a Distributed System Problem
One agent is a worker; a thousand agents form an organization of machines, which is a distributed system. Drawing on Google Research's 180-configuration study, Silo-Bench, and AIOS, the author argues that adding agents helps parallel tasks by up to 80.9% but hurts sequential work by 39–70%, and that error amplification drops from 17.2× to 4.4× with an orchestrator. The durable thing should be the state, not the agent: schedule agents like processes and recover them like nodes.
A single agent is an intelligence problem. A million agents is a distributed systems problem.
- duhhhhh1212
https://www.pangram.com/history/b85fd1a5-1797-4347-9bfa-23de...
Don’t waste your time.
- tolugenius
This leaves me two questions
1. Will we ever reach (or have we already) where we can more empirically measure the cost of the adding the nth agent. Like the way businesses of different sizes can know when the nth employee add diminishes returns overall, can we measure the front with agents depending on task scope, resources, and other factors.
2. On the other end, can we see point where we can reduce how many agents are needed down the minimal set? For example many chains have reduced staff down to the minimal number of employees (note I do not agree with this) and still the balance sheet is in the black. I think in the hyper optimization and efficiency society we're in I think this will also happen, although the reality of "I can do 100 agents work with 10" might not be in providers and labs best interest economically.
- whearyou
I’ve got no problem with LLM copywriting but the LLMisms just drip off this thing, to the detriment of its clarity. Shame.