docs/ai/sufficient-reasoning.md

# Sufficient Reasoning

Sufficient reasoning is enough to understand, decide, implement, verify, and deliver safely—then stop. It uses investigation and evidence floors appropriate to the requirement without inventing unlimited edge cases.

Map code, behavior, integration, security, and migration requirements to appropriate proof. Required first verification is mandatory; equivalent repeated checking is bounded. See [sufficient-reasoning guidance](../../.ai/guides/sufficient-reasoning.md).