Skip to main content

4. AI build phase

Build well because it was thought well.

This section contains the operational guide for executing Phase 4. While the methodology section explains the why and the concepts, this guide details the how: the concrete steps, templates, artifacts, and validation criteria.

5
Structured steps
From Setup to Closure
6
Artifacts
Transferred to Phase 5
10
Gate Review criteria
Integration checklist

The AI Build Phase workflow​

AI Build Phase executes construction in five steps that manage parallel tracks, detect collisions between them, and validate the result with the original stakeholders. The environment is verified before building anything. Agents execute Story Files under human supervision. Gaps are classified and resolved in real time. The fidelity review confirms that what was built solves the original problem and respects the agreed solution. Closure consolidates learnings for the market cycle.

1. Environment setup

Collective briefing, three-read context test, and implicit dependency graph verification between Story Files.

Verified environment
1

2. Story File-driven construction

Agent execution under Dev Lead supervision. Decision classification: Local, Cross-track, or Global.

Completed Skills
2

3. Gap management during construction

QA Agent detects inconsistencies between built output and context. Classification: Specification, Rules, Architecture, or Data.

Updated Decision Log
3

4. Fidelity review

Triple validation: Technical (system works), Phase 1 (solves the problem), and Phase 2 (implements the agreed solution).

Build Validation Report
4

5. Decision Log closure

Decision Log consolidation across all tracks and Iteration Brief drafting with cross-track learnings.

Updated Context Document
5

Phase artifacts​

This phase produces six artifacts that record decisions made during construction, discovered dependencies, and validation that what was built is faithful to the original problem and solution.

What this guide covers​