How I use AI on this site

What AI tools help with, what I still do myself, and how I keep this project honest and personal.

This site is personal. I want it to stay that way — even though AI tools helped me build it.

I'm not hiding that. If you're reading this, you probably care about authenticity too. So here is how I work, in plain language.

The short version

I use AI as a tool, not as a ghostwriter.

Cursor is my main assistant for code and content. It drafts, suggests, and speeds things up. I decide what goes live, edit what sounds wrong, and take responsibility for what you read here.

If something on this site feels like me, it's because I made it feel that way on purpose.

What I use

  • Cursor — coding, refactoring, debugging, and content drafting in this repo
  • Large language models (via Cursor) — research, wording, translation EN → DE

That's it for now. No secret stack. When that changes, this page gets updated.

What AI typically helps with

LabelWhat it means
draftingFirst versions of blog posts, portfolio text, or page copy
editingGrammar, structure, tightening sentences, tone checks
translationGerman versions based on an English draft I have already approved
codeScaffolding, boilerplate, refactors, and "why is this broken?" sessions
researchLooking up APIs, framework docs, or patterns I haven't used yet
imagesAI-generated or assisted visuals — only when noted on a specific piece

Blog and portfolio entries can record these labels in frontmatter (ai_assist) as structured metadata. You won't see badges on every post yet — but the record exists in the GitHub repository for anyone who wants to check.

What I still do myself

  • Decide what to publish and what stays draft
  • Edit until the voice sounds like a shopfloor CEO, not a marketing bot
  • Verify facts, links, and that nothing private slipped in
  • Run the business context — the stories here come from my experience, anonymized
  • Take responsibility for errors, tone, and what this project stands for

AI didn't write my app for me. It shortened the distance between "I have an idea" and "let me see if this works." I still had to understand what broke when it broke. For someone who learned welding before JavaScript, that difference matters.

Privacy boundaries

The GitHub repository is public from day one. That means discipline:

  • No real company, customer, or employee names in git — ever
  • No .env files, credentials, or internal system details
  • Content is written as if a stranger will read it — because they might

AI assistants follow the same rules via project standards in .cursor/. When the boundaries are written down, the tools stop improvising privacy mistakes.

See also AI-USAGE.md in the repository for the full policy.

Commits and code

When AI helped with a commit in a meaningful way, I may add an optional trailer:

Assisted-by: Cursor

That's not a badge of shame. It's a habit — like writing commit messages you'd show an employee.

What this is not

  • Not "AI wrote my entire site" — it didn't
  • Not pretending every word was typed without help — some weren't
  • Not a tutorial on how to automate your personality away

WERKSCODE is a small public place to learn in the open. Transparency is part of that promise.

Questions?

If something here is unclear or you think I missed a disclosure, get in touch. I'd rather fix it than hide behind vague wording.