Ambiguity is insufficient clarity that permits multiple interpretations. In multi-agent systems, it changes from a minor inefficiency into a structural risk: each missing constraint can produce many divergent implementations. section 10.8.1
Ambiguity becomes dangerous when:
- Conventions rely on context or experience instead of explicit constraints. section 10.8.1
- Many agents interpret the same instruction differently, creating plausible but inconsistent code. section 10.8.1
- Individual inconsistencies compound across commits into architectural drift that nobody fully understands. section 10.8.1
The 90-percent-clear convention
A convention that is 90% clear may work for one developer, who infers the missing 10% from experience. Given to a hundred agents, it can yield ten subtly different implementations that compile, pass tests, and still move the architecture toward an unintended shape. section 10.8.1





