While I was building RoleFinder, I was also taking an AI bootcamp. A good one — mostly self-paced, well-structured curriculum. It gave me a higher-level product perspective through an AI lens. But the real deep diving came from building around a mission, a goal, an idea — not just hypotheticals and anecdotes, no matter how interesting or relevant they were. I had to solve a personal pain point.

Not because the bootcamp isn’t valuable. It is. But every time I sat down to work through a module, I kept thinking about the next piece of RoleFinder I wanted to build. The workshop taught concepts. RoleFinder needed solutions. The pull of a real problem beat the push of structured learning every time.

Here’s what I’ve come to believe: workshops give you frameworks. Building gives you judgment.

Frameworks are useful. They give you vocabulary, patterns, a sense of what’s possible. But judgment — knowing when to apply which pattern, when to break the rules, when to stop researching and start shipping — that only comes from doing the thing.

I learned more about n8n’s quirks in three days of debugging the heisenbug than I would have in a week of tutorials. I learned more about AI collaboration by iterating with Claude Code on a GitHub repo than I did reading about prompt engineering.

And here’s the part I didn’t expect: the skills don’t stay contained.

A few weeks after RoleFinder was running, I needed to find local tax accountants for a side project. Old me would have Googled, scrolled, copied names into a spreadsheet. New me spun up an Apify actor against Google Business listings, piped it into n8n, landed it in a database table. Ten minutes. No friction.

I didn’t learn "how to build a job monitoring system." I learned a new way of working. That’s the difference.

This has been the most fun I’ve had in nine months of job hunting — not because the job hunt got easier (it’s still hard), but because I stopped just searching and started building something.

Next up: the invitation.