I’m Geoff, and I’m going to be building 42 AI prototypes in public over 2026 - 42-26. Each prototype will attempt to tackle a “hairy problem” - something that doesn’t have a solution just yet, and may benefit from the fresh models and approaches coming through the AI boom. This blog will be capturing what worked and what didn’t. It will also have shipped, accessible demo’s available for anybody to poke at.

Why 42?#

Partly because it’s the answer to life, the universe, and everything 🐬. Mostly because it’s an ambitious enough number to keep me honest. One prototype roughly every week-and-a-bit, for a full year.

Why build in public?#

I’ve spent years at Atlassian and Salesforce watching the gap between what AI could do and what businesses actually get out of it. The gap isn’t the technology — it’s the execution. The last mile. The part where someone has to make it stick inside a real workflow with real people.

So instead of writing blog posts about how AI is changing everything (you’ve read enough of those), I figured I’d just… build things. Explore what makes them sticky or not. And show my working.

Some will be useful. Some will be weird. A few will probably be outright failures - and those might be the most interesting ones to write about.

What to expect#

Each prototype will get a write-up covering:

  • What it does — the problem it’s solving and for who
  • How it works — the technical bits, kept readable
  • What I learned — the honest version, including what went wrong

The prototypes will lean toward the kinds of problems I see with clients at Moustache Metrics — small and mid-sized businesses trying to get AI working without a dedicated data team. Think automating the spreadsheet everyone hates, not building the next GPT.

Follow along#

If you’re interested, the best way to keep up is to check back here. I’ll also be pushing code to GitHub as I go.

First prototype incoming soon. What’s the worst that could happen?