AI Driven Development Methodology

Human as Architect → AI as Builder

Humans own intent, scope, architecture, delegation and final decisions. AI agents perform bounded work.

One workflow, four steps

PRD → PLAN → TASK → IMPLEMENTATION

1. 01 · PRDDefine intent, scope, constraints, non-goals and acceptance criteria.
2. 02 · PLANTurn approved intent into a reviewable approach with dependencies and completion conditions.
3. 03 · TASKDelegate known, bounded work with context, scope, criteria and expected evidence.
4. 04 · IMPLEMENTATIONBuild within scope, review, test and use evidence for a human decision.

Evidence is not an agent's completion claim. Review and testing belong inside IMPLEMENTATION. The human judges the evidence against intent, constraints and acceptance criteria before deciding what happens next. Delegating execution does not delegate accountability.

The methodology supports reviewable, evidence-backed work. It does not guarantee production-ready software or automatically authorize release or deployment.