SUMMARY
It All Comes Down to Five Things
Fifteen principles. Five responsibilities they all serve.
CORE PRINCIPLE
Good AI-assisted development is not just about getting AI to write working code. It is about building something you can still understand when the conversation is gone.
- 01Describe the Outcome Before the Implementation
- 02Separate the Problem From the Solution
- 03State the Constraints
- 04Make the Repository Explain Itself
- 05Record Intent, Decisions and Current Reality
- 06Keep Documentation in Sync With Code
- 07Separate What Changes for Different Reasons
- 08Keep Routine Content Human-Readable
- 09Expose the Things You Expect to Change
- 10Use AI to Build Exploratory Tools
- 11Separate Product Code From Maintenance Tools
- 12Make Small Changes and Define What Must Stay Stable
- 13Compare Implementation With Intent
- 14Test Behaviour and Use Version Control
- 15Reduce AI Dependence as the Project Matures
Define
Know what the system is supposed to do.
Document
Record intent, decisions and current reality.
Structure
Separate content, logic, presentation and configuration.
Inspect
Make small changes, test behaviour and compare implementation with intent.
Decouple
Build toward a system that remains usable without AI.
COMPANION PRINCIPLE
The chat should help build the project; the project should not depend on the chat to explain itself.