Method
Speed comes from clarity.
I combine product understanding, explicit architecture, short increments and instrumentation. AI accelerates execution, but decisions and responsibility stay human.
-
01
Understand the real
Users, context, constraints, business goals and success metrics. I reformulate the problem before proposing a solution.
Possible deliverables
- user journeys
- business rules
- edge cases
- acceptance criteria
- product metrics
-
02
Structure the solution
Architecture, component responsibilities, data model, API contracts and testing strategy. Important choices are documented before they become expensive to change.
Possible deliverables
- architecture diagram
- technical decisions
- integration plan
- offline strategy
- migration plan
-
03
Ship in increments
Short, testable and demoable batches. AI accelerates some steps, but every proposal stays reviewed, tested and placed back into the product's architecture.
Possible deliverables
- vertical features
- code reviews
- tests
- regular demos
- shippable builds
-
04
Observe and improve
After release, I use feedback, metrics, incidents and real behaviours to prioritise what comes next. Production is where learning begins, not where the project ends.
Possible deliverables
- analytics instrumentation
- error tracking
- funnel analysis
- performance improvements
- iteration roadmap
Definition of done
A feature is not done because the code compiles. It's done when it answers an understood need, works in the expected cases, can be tested, is observable in production and can be picked up by the team.
AI in my method
AI as a copilot, never as owner of the code.
AI tools can accelerate exploration, documentation, test generation or certain implementations. They don't replace understanding of the system. I check assumptions, review changes, run tests and keep an architecture that is understandable without the tool.
To speed up specification and execution, I use a structured-documentation method inspired by BMAD. The product and architectural context is versioned so that humans and AI tools work from the same base.