My Tools

My relationship with AI coding tools changed in late 2025. Claude Code writes much of the code now. I decide what to build, review its work, and test the result. I wrote more about that in You Are The Driver.

  • Claude Code is my main coding tool, from new projects to refactors. My CLAUDE.md setup tells it to flag uncertainty and push back when it sees a problem.

  • 💬

    Claude is for planning and talking through problems that don't require access to the code.

  • 🤝

    GitHub Copilot is still in VS Code for tab completion. I accept maybe 25% of its suggestions, mostly boilerplate.

  • I write the posts and opinions on this site myself. I sometimes use AI for editing suggestions, then decide what to keep.

The Driver

AI can write plenty of code without knowing whether it solves the right problem. From You Are The Driver, this is what I still need to provide:

🎯

Direction

What are we building? What problem are we solving? What are the constraints?

🔍

Validation

Does this actually work? Does it match the requirements? Is the code quality acceptable?

Architecture

How does this fit the larger system? What are the long-term implications?

Judgment

When to push back, when to accept, when to try a different approach entirely.

See my full CLAUDE.md configuration for how I set up this working relationship.

AIL Scale

I use the AI Influence Level framework to label how much AI was involved in content on this site. Most posts here are AIL 0 or 1.

0

All Human

Created entirely by hand, no AI involved.

1

Minor AI Assist

Human-written, AI helped with grammar or phrasing.

2

Major AI Augmentation

Human-written but significantly expanded with AI tools.

3

AI Created, Human Structure

Human described the full structure, AI filled it in.

4

AI Created, Human Idea

Human provided a basic idea, AI implemented it.

5

Mostly AI

AI generated with minimal human input.