AI workflow
How I build with AI.
AI makes me faster, not careless. Here is exactly where it helps in my projects, and where my own judgment takes over.
Tools
- Claude· Main tool: requirements drafts, feature code and debugging help
The workflow
Six steps, every project.
- 01
Requirements
I use AI to generate the basic project requirements, then check and correct every point myself before anything gets built.
AI + Me - 02
Project setup
I set up every project by hand: the framework, folder structure, tooling and configuration. No AI is involved in this step.
Me - 03
Code generation
AI, mainly Claude, writes most of the feature code, working from the requirements I have already verified.
AI - 04
Line-by-line review
I read and check every generated line, one by one. Code only stays in the project once I understand it and agree with it.
Me - 05
Security code
Authentication, authorization, tokens, cookies, payments and anything else security-sensitive is written by me by hand, not by AI.
Me - 06
Bugs and errors
When something breaks, I use AI as an assistant to find the cause and possible fixes, then I verify the fix myself.
AI + Me
Who does what
AI helps with
AI- Drafting the basic requirements
- Generating feature code (mainly Claude)
- Finding causes and fixes for bugs
I always do
Me- Check and correct the requirements
- Initialize and configure the project
- Review every line of generated code
- Write all security-related code
- Verify every bug fix
Principles
01
Requirements before code
AI drafts requirements, but nothing is built until I have checked them.
02
I read every line
Generated code is reviewed line by line. If I can't explain it, it doesn't ship.
03
Security stays human
Auth, tokens, cookies and payments are written by me, never generated.
04
AI suggests, I decide
AI helps me find bugs and options faster; I verify every fix and make the final call.
In my projects