Skip to main content

S4U Development Methodology

AI-Assisted Software Engineering at Production Quality

A systematic approach to building production software with an AI collaborator — treating the AI as a senior engineer under architectural supervision, with quality enforced by machinery you can run yourself. This site was built using the methodology it describes.

See the verifiable benefitsWhy a methodology?

What the methodology enforces — by machinery, not discipline:

Design-first
no code before a design exists
Verified
no unverified "it works"
Self-enforcing
quality lives in hooks + gates
Self-pruning
complexity retired on a schedule

Quality as Architecture

Quality is built into the infrastructure as automated structural constraints — hooks, gates, reviewer agents — not maintained by discipline. A violation must defeat four independent layers to land.

Evidence over Claims

"It should work now" is treated as "I have not verified it." No task is complete without fresh test output. The benefits are mechanisms you can run, not numbers you must trust.

Design before Code

Every non-trivial change flows through brainstorm, design, plan, then implement. The AI collaborator is a senior engineer under architectural supervision — not an autocomplete.

A reading path

  1. Why a Methodology — the failure modes of unsupervised AI code generation.
  2. Verifiable Benefits — each claim, the mechanism that enforces it, and how to run it.
  3. Philosophy — the eight principles, each from a specific failure mode.
  4. Development Lifecycle — the full cycle, the bug-fix cycle, the code-review cycle.
  5. The Machinery — instruction hierarchy, agents, memory, and quality gates.
  6. Adopt It — bring the methodology into your own project.