Building it
🏠The everyday picture
A good builder works in stages: foundation, walls, roof, wiring, finishing, with an inspection at each stage. Rushing ahead and hiding problems behind plaster leads to cracks later.
📘What it really is
-
Build in vertical slices (Module 98): get one thin flow working end to end early (for example, log in → create one record → see it listed), deployed to staging, then grow it.
-
Frontend and backend:
- a typed React (or your chosen) frontend that's accessible, responsive and fast on mobile (Levels 3 and 4);
- a typed backend with validation, layered modules and clean core rules (Modules 46 and 92);
-
the database through migrations and parameterised queries in repositories (Level 5).
-
Authentication and authorisation: proven libraries or services; roles and object-level checks on every request; MFA for admins (Module 74).
-
Security controls: the OWASP Top 10 handled (Module 75); headers, CSRF and rate limits; secrets outside code (Module 77); dependency and code scanning (Module 78).
-
Tests:
- unit tests for domain rules and value objects;
- API tests including permission refusals and validation;
- database integration tests (Module 59);
- a few end-to-end tests of key flows (Module 39);
-
evaluations for any AI feature (Module 89).
-
Working with AI tools: give clear, small tasks with context (Module 83 and the Builder Path); read and understand every change; run the tests; never accept code you can't explain. Keep a short log of where AI helped and what you had to correct; it's part of your assessment (Lesson 5).
-
Work in your process: sprints or Kanban, a Definition of Done, commits and pull requests with reviews (Module 98).
🖱️Try it
Build your MVP in vertical slices, deploying to staging from the first slice. Keep your board, pull requests and test reports up to date. Keep an AI usage log for the project: the task, what you asked, what you changed, what you rejected and why.
✅Quick check
Why get a thin end-to-end slice working and deployed early?
Show the answer
It proves the whole path works (frontend, backend, database, deployment) and exposes integration problems early.
What must you be able to do with any AI-written code you keep?
Show the answer
Understand and explain it, and have tests showing it works.
