← Show my work
LESSON How I build [DATE] · 5 min read

Can a non-engineer build real tools with AI?

Short answer: yes — if you treat it like an operations project, not a coding project.

Taylor Walburg · Walburg Co.

I'm not a software engineer. My career has been quality, operations, and leading teams in West Michigan manufacturing. Even so, I've built a small-group Bible reading app and a 40+ lesson Spanish audio course, and I'm getting both ready for other people to use.

I didn't get there by learning to code. I used skills I already had: defining the problem, writing down how the process should work, and testing with real people. AI does the typing. The thinking is still mine.

The process, start to finish

Five steps. The same ones I'd use to fix a process on a plant floor.

  1. 1

    Name the real problem

    One sentence, one person. If I can't say who it's for, I don't start.

  2. 2

    Write the spec first

    What it does, what it doesn't, and what "done" looks like — before any code.

  3. 3

    Build in small phases

    With Claude Code, one working piece at a time, tested before the next.

  4. 4

    Put it in front of a real person

    My first Common Read tester was my dad. Real use beats my guesses.

  5. 5

    Ship small, then improve

    Fewer features, done well. The next version comes from what people actually do.

What it runs on

Nothing exotic: four tools, each doing one job.

Claude Code — writes the codeCloudflare Pages — hostingElevenLabs — lesson voicesGumroad — checkout

Where I still get stuck

[One honest mistake you made while building, and what you changed because of it.]

If you can map a process, you can probably build a tool. Start with the problem.

Get the next note.

One short note a month, at most. What shipped, what's coming.