The Tool We Didn't Build, and the Four Questions That Killed It

Before committing a build slot to a small search-driven tool, every candidate goes through the same four fixed questions: is the demand real, does the core technology actually work on the target device, is there a real gap in what's already out there, and can the scope be cut any smaller. This round's candidate was a browser-based "speaker cleaner" — a tool that plays a specific tone through a phone's speaker to shake trapped water out after a drop or a swim. The build itself: one HTML page, a 165 Hz sine wave, a button.

The verdict was to stop, and not because of scope or feasibility.

Demand is real — that's not in question. It's also exactly the problem. The search results for this category are already a clone farm: a dozen exact-match domains (fixmyspeaker.online, fixmyspeaker.net, fixmyspeaker.io, fixmyspeaker.org, fixmyspeakerz.com, fix-myspeaker.com, fixmyspeakers.tech, fixmyspeakerlab.com, watereject.com, speakerscleaner.com, among others), most old enough to have accumulated backlinks. A page with zero differentiation from what's already ranking has no credible path to page one within any reasonable window. The tool's own success metric — search traffic — was unreachable before a line of code got written.

The technical check turned up something worth keeping regardless of the outcome: on iOS, Web Audio goes completely silent when the phone's physical mute switch is on — Apple's AudioContext starts in an "ambient" playback category that respects the hardware switch (WebKit bug 237322). The target device is an iPhone; the target user is someone who just got their phone wet and, more likely than not, carries it on silent. In the naive implementation, that's not an edge case — it's the main path, and it plays no sound at all. There's a known workaround (a silent <audio> element plus a user-gesture unlock), and it fits inside a one-day build. Just not a build worth doing, given the first problem.

The last check was the sharpest: is there a smaller version worth shipping anyway? Cutting the sweep and the timer down to a single button doesn't change the answer — most of the clone-farm competitors already are a single button. A perfect, minimal clone of a product with zero differentiation still gets zero traffic. Scope was never the problem.

Nothing got built. That's the actual output of the evaluation — a discipline that costs a few hours instead of a month, applied before the code, not after the traffic numbers come in.