The Ledger holds every act with its receipt: what ran, when, on whose authority, with machine evidence of the outcome. Review is reading, not reconstructing, and unverified acts are flagged as exactly that.
The management pattern that works: skim the Ledger the way you skim a team's standup, letting receipts carry the proof, and drill only where something looks off. Minutes a week, honestly spent.
This is what makes delegation to agents governable at scale, you audit evidence instead of re-doing work, and it is why receipts are non-negotiable in Heidi's design.