conways-law ADG-0007Conway's law
“Organisations design systems that mirror their own communication structure.”
Origin
Melvin Conway, 1967, "How Do Committees Invent?". About human teams: the interfaces in a product fall where the interfaces between teams fall.
Why it should, or should not, apply to models
A multi-agent system is an organisation, and its communication structure is a design choice made explicitly in the harness — who talks to whom, what is broadcast, what is private. If the law transfers, the shape of the output (where its seams and inconsistencies fall) should be predictable from that topology. The mechanism with people is that communication cost shapes what gets coordinated; with agents the cost is near zero, so the law may hold for a different reason: what is shared shapes what is believed.
Standing: Untested
Standing counts claims a person has reviewed. Claims reviewed by AI alone are listed and marked, and do not move it.
No evidence filed either way.
Searched 2026-09-12: The catalog has a claim that communication topology changes multi-agent error rates, which is the premise of the law but not its content: nothing yet shows the structure of an output mirroring the structure of the agents that produced it. Nearest misses: multi-agent-llm-systems-where-one-agent-starts-erroneous-shared.
A break is worth more than a hold. If you have seen this fail for a model, with the setup written down, that is the most useful thing you can add here.