The moat wasn't the AI model. It was refusing to ship a single-signal product.

Disclosure up front since this is the self-promotion forum and I'm the one building it: I'm one of the people behind Clad9, an AI wardrobe app. Posting here on purpose, not sneaking it into a thread — figured the build story might be more useful to this crowd than a launch announcement.

Every "AI closet" pitch sounds identical for the first thirty seconds: photograph your clothes, get outfit suggestions. The differentiation always turns out to live one layer down, in which single signal the product picked to be good at — and that's also where the category quietly fails.

Look at what's actually out there. One competitor nails color harmony — genuinely good color theory, matches pieces that visually work together. Another does a chat-style interview to build a style profile. A third tracks wear history and layers in weather data so you stop repeating the same five outfits. A fourth proved the whole concept works, but shipped iOS-only. Each of these is a real, defensible product. None of them constrain on more than one axis at a time.

The reason that matters more than it sounds: color, body, and calendar aren't independent variables you can bolt together later — they interact. A color pairing that's textbook-correct (complementary hues, right off the color wheel) can still clash because it ignores undertone: burgundy is a cool red, rust is a warm red, and pairing either with the wrong-temperature green looks subtly off even though the hue math checks out. Body proportion changes which of those "correct" pairings actually reads well on a given frame. And the calendar changes the formality floor underneath both — an outfit that's perfectly color-matched and well-proportioned can still be the wrong choice if the day calls for an interview instead of a coffee run. Ship a product that only solves one of these and you've built a genuinely good tool that still gives wrong answers some mornings, because the other two axes are quietly overriding it.

So the actual build decision wasn't "which signal should we be best at" — it was refusing to pick one. Clad9 runs personal color analysis (undertone, not just "season") from a single photo, filters by body proportion, and reads the calendar for what the day actually calls for, then returns exactly one complete outfit built only from clothes the person already owns. Not a feed of options to choose from — one recommendation, the same way a person who actually knows your wardrobe would just tell you what to wear.

The other unglamorous problem in this category is onboarding. Most "digitize your closet" flows die at item 14 of 200, because photographing a wardrobe one garment at a time is exactly the kind of chore people start and abandon. We built the intake around a single video walkthrough of the closet instead — one pass with a phone camera, and the backend handles segmentation, background removal, and tagging from that. It's not zero-effort, but it's a five-minute task instead of a forty-minute one, and that's the difference between people finishing setup and people quitting at item 14.

None of this is live behind a paywall, either — the color/body/calendar engine is free to use right now, which was a deliberate call: this category has enough single-purpose free tools that a genuinely integrated one needed to be free too, or the comparison would never happen. The feature breakdown covers the mechanics in more depth, the color methodology page is the fullest writeup of the undertone approach above, and the comparison page is an honest rundown against the color-only, interview-based, and wear-history competitors mentioned above — written to be fair to what they're each actually good at, since none of them are bad products, just narrower ones.

Curious how this lands with people who've shipped in adjacent spaces (fashion, personalization, recommendation systems generally) — particularly whether "constrain on N signals simultaneously instead of picking one to be best at" holds up as a moat elsewhere, or whether wardrobe/styling is unusual in how much the signals interact with each other.