The filter benched my fun ideas. I built the one it kept.

Sixth one.

Sparkfile is built around a filter that tells you which of your ideas is worth building and kills the rest. I’ve been running my own ideas through it. The last time, it benched three I liked and kept one I didn’t want to build. This is about the one it kept.

It started as a worry, not an idea. I just finished a new website for Dan Bailey’s, an outdoor company here in Montana, and their team is actively adding products and pages to it. Every change is a chance to break accessibility, and lawsuits against Shopify stores over it are real and getting more common. I’d found an accessibility scanner on GitHub, and I wanted to run it against their store to see what our launch and their edits had left behind. Running it is what did it. Somewhere in there I stopped thinking about Dan Bailey’s and started thinking, this should be an app, because every Shopify store has this exact problem.

So on July 10 I ran it through the filter, next to three ideas I was more excited about: Tone Farm iOS, a guitar-tone app; a tool to get stores surfaced inside AI shopping agents; and a self-serve Shopify store builder. The filter benched all three, each for the same reason, no defensible wedge and a platform that would eventually build it themselves. The accessibility scanner cleared. Boring category, non-discretionary, real legal pressure behind it, and a wedge I actually have, because I build accessible Shopify themes for a living. On paper it was the obvious one.

I didn’t want it to be. I want to build the fun apps, the guitar-tone stuff, the things I’d use myself. An accessibility scanner is not the app I daydream about. But I need real customers and real revenue, not another passion project, and I built this whole thing to test one belief: that a filter should beat my gut about what to build. This was the test. If I’m going to tell people to trust the filter over their own excitement, I had to do it first. So I built the one it kept.

I started that same afternoon, a few hours after the verdict. It’s called Kōkua. It scans a Shopify store, finds the accessibility problems, and hands you the code-level fixes, instead of dropping an overlay widget on top, which is the approach that actually gets stores sued. I built it in Claude Code with the Chrome extension driving the Shopify admin. Seventy-five commits over three days. The one real wall was the scanner engine fighting the build tool, and once that was sorted the rest moved fast.

The hardest part wasn’t the code, it was the App Store submission. All the listing copy and details and screenshots that I would have either agonized over for a day or gotten wrong. Having Claude draft the first version of it and then correcting it turned a day of second-guessing into an hour. I also formed the company behind it, Kahakai LLC, and filed it with the state. I’ve started companies before, more than a few, mostly things I cared about that never had a clear reason to exist. This one feels different. Less like a dream, more like something with an actual job to do. It might be the first one I’ve formed that sticks.

Here is where it actually stands: submitted, not approved. It’s in Shopify’s review queue right now, and it could get rejected. It has zero customers. None of this is a win yet. It’s a real thing that exists, built from the idea a filter told me to build instead of the ones I wanted, and now I wait.

The point isn’t that I built an app in three days. It’s that the filter turned “which of my ideas is real” from a slow, expensive guess into a morning, and then made me build the answer instead of the fantasy. That is the whole reason Sparkfile exists. I built a filter to keep people from pouring months into the wrong idea, and the first real test of it was whether I would take its verdict over my own excitement. I did. Now we see if Shopify agrees.