We built Trott in public for Shipaton. Here’s what actually shipped.

We gave ourselves 60 days to ship Trott for the RevenueCat Shipaton. Two people, day jobs, one stubborn problem: you save hundreds of reels and never find the one you need. This is the honest version of how it went: what we shipped, what broke, and the bits we'd do differently.

The problem we couldn't let go of

It started with a Bali trip. Mrinmoyee had saved something like 200 reels over a year. Cafes, beaches, that one sunset spot. When it was finally time to plan, she couldn't find any of them. An hour of scrolling her own saved folder. Nothing.

That's the whole thing, really. Saving a reel feels productive. But a saved video is a sealed box: the address, the recipe, the name of the place. All of it locked inside footage you'd have to re-watch to use. We didn't want a prettier bookmarks folder. We wanted the box opened.

What we actually shipped in 60 days

By the end of the Shipaton, Trott was live on both the App Store and Google Play. Not a prototype. The real thing, with paying-capable infrastructure. Here's what made the cut:

• Share a video from Instagram, TikTok, or YouTube, straight from the system share sheet.
• The video gets read end to end: we look at the frames and the caption, then write down what's in it.
• Recipes come back as ingredients and steps. Travel reels become named places with a map. Products, workouts, tutorials, sorted by what they are.
• Natural-language search and chat across everything you've saved.
• Credits and purchases through RevenueCat, real-time status synced to the app.

How it's built, the short version

The app is Flutter, so one small team ships both the App Store and Google Play from a single codebase. Payments run on RevenueCat (the Shipaton sponsor), which saved us from building billing from scratch.

The hard part isn't the app. It's the pipeline that takes a video and turns it into something useful: reading what happens on screen, what's said, and what the caption claims, then writing it down in a way you can search by meaning instead of exact words, so "those quiet cafes" finds the right reel even if nobody typed the word "cafe." We'll spare you the plumbing.

What broke

Plenty. The honest list:

• iOS share extensions. Getting a shared link out of the extension and into the app reliably took far longer than the smart parts did. It's the kind of thing that looks like a checkbox and turns into a week.
• Reading from a dozen apps. Every app exposes its videos differently, and "reliably pull this in, every time" was its own quiet mountain of edge cases we didn't budget for.
• Doing the heavy work at the right time. The processing behind each save isn't cheap. Early on we did too much, too eagerly, and watched the bill. Credits and smarter queuing came out of that lesson.

What we'd do differently

Ship the ugly version sooner. We polished screens nobody had used yet. The features that mattered — share, extract, search — were the ones we tested last, and they're the ones that needed the most real-world bashing. Next time, the share button works on day three, not day thirty.

We'd also write down the unit economics earlier. "It works" and "it works without losing money on every video" are two different milestones, and we found the gap the hard way.

Where it is now

Trott is on the App Store and Google Play, free to start. We're still two people. If you try it and something's broken, tell us. That's the fastest way it gets better.

Share one reel to it. See if it finds what you needed. That's the only test that matters to us.