Core Concepts
This section explains the mental model behind this system.
It is written for humans, not for AI. Reading it is optional, but it helps you understand why the system is structured the way it is.
If you want to start immediately, you can skip this section and follow Getting Started.
1. The System Mental Model
This system is designed around a simple idea:
AI does not replace professional software development practices. It enforces them.
AI is treated as a capable but literal collaborator. It does not hold long-term intent, architectural memory, or product context on its own.
Therefore:
- Intent must be explicit
- Decisions must be written down
- Quality must be enforced through structure
Speed is a by-product of clarity - not the primary goal.
2. Artifact‑Driven Development
Work in this system happens around artifacts, not conversations.
Artifacts are durable, inspectable documents that define intent and constraints, such as:
- Project Briefs
- Feature Descriptions
- Refactoring Plans
- Standards and Definitions of Done
Why this matters:
- Chat history is ephemeral.
- Artifacts survive context loss.
- Artifacts make intent auditable.
If work feels unclear, the solution is usually to create or update an artifact — not to "explain it better in chat".
3. The "AI contextFlow Operating System"
The system separates concerns deliberately to prevent context overload.
| Concept | Definition | Question it answers |
|---|---|---|
| Roles (Agents) | How AI thinks | "Who am I right now?" |
| Standards | What rules apply | "What is 'good' code?" |
| Workflows | How work is executed | "What is the next step?" |
| Templates | Which artifacts exist | "Where do I write this?" |
4. Explicit over Implicit
AI performs best when boundaries are clear, constraints are visible, and ambiguity is removed early.
This system favors:
- Explicit scope over assumptions
- Written decisions over tribal knowledge
- Small, reviewable changes over large leaps
If AI produces inconsistent results, it usually means something important was implicit. Make it explicit.
5. Why This Is Not Overhead
Creating artifacts and following playbooks may feel slower at first.
In practice, it:
- Reduces rework.
- Prevents architectural drift.
- Makes refactoring safe.
- Keeps AI aligned over time.
The goal is not ceremony. The goal is sustained velocity without loss of quality.