Chapter one,
Humans lead. AI amplifies.
People are the architects of vision, meaning, and experience. AI is the accelerant: it handles complexity, flags what breaks, and does the routine work so a person does not have to. It does not decide what a product is for, and it does not merge code. In Forge, agents generate, test, and self-review; a human reads the diff and rules on every checkpoint. Different models do different jobs. Every action is logged, so accountability never blurs.
Unit 214 · forge/unit-214-rounding
Invoice totals round half-up
Agent self-review · Builder agent
- Plan match: 3 of 3 acceptance criteria covered
- Tests: 12 added, 0 skipped, all green
- Proof recorded: 00:41, two flows, no console errors
- Open questions: None. ADR-014 applied as written
Diff summary
Shows this file's share of the whole diff.
- src/invoices/total.ts+18-6
- src/invoices/total.test.ts+640
- docs/adr/014-rounding.md+90
- 3 files+91-6