AppScreenshotKit

AppScreenshotKit was another one from the sprint with Nick. He’s an expert at building iOS apps, and App Store screenshots are a real, boring pain point — most of them look bad, and making good ones is fiddly. So we built a tool to make pixel-perfect ones fast.

The Thesis

Two bets. First, that we could build and ship something genuinely useful in a day. Second — the one I cared about most — that organic distribution alone could get a brand-new tool in front of the right people, with zero ad spend, if you knew where to show up.

What I Built (and How Fast)

The product went live in about 24 hours. Within the first week it had real users — and, more importantly, actual purchases — almost entirely from Reddit posts I wrote and from AI tool directories that picked it up. I’d optimized the listing and positioning so it got found. A meaningful share of all our traffic came straight from AI-discovery and directory sources.

That was the bet paying off: directory-and-community distribution is a real, scalable early channel. I now treat “one-day build + ninety-day organic distribution” as a repeatable way to test a concept.

What Happened

Then we started using our own product — and it had real problems. It had been built fast, and it showed; it didn’t work as well as it needed to. There wasn’t much appetite to keep polishing it, and underneath that was a deeper disagreement.

In the AI era, building the software is no longer the hard part — anyone with some technical skill can ship quickly now, and I was scaling up my own building fast too. The hard, undervalued part is marketing and sales: generating quality demand. That’s exactly the part a lot of technical founders discount, because it looks easy from the outside. That misalignment — on what was actually valuable — is why the partnership on this one didn’t continue.

What I Learned

I proved distribution wasn’t the bottleneck; I did it in a weekend. The real disciplines were the ones around it: shipping something you’d actually stand behind, and being honest enough to stop selling a tool once you no longer believe in it. In a world where anyone can build, the moat is distribution and judgment — not the code.