Why tasteID exists.

Every AI tool starts from zero. You explain that you prefer clean layouts. Then you explain it again in a different tool. Then again. And each time, you're trying to articulate something that's hard to put into words — because taste isn't a settings file. It's a sensibility.

The real issue isn't that AI tools can't follow instructions. It's that the instructions we give are shallow. “Make it minimal” means something different to every person who says it. Your version of minimal — the specific way you balance whitespace, the weight you give to typography, how much you trust users to figure things out — that's yours. And right now, there's no way to capture it once and carry it with you.

Taste isn't binary. It would be easy to build a quiz that asks “dark mode or light mode?” and calls it a profile. But that's not how design sensibility works. You might lean toward restraint in most things but want rich, layered textures in certain contexts. Rounded corners feel right in a consumer app but wrong in a developer tool — and that's not inconsistency, it's judgment.

So the challenge isn't just mapping preferences — it's mapping the structure underneath them. The principles that stay constant regardless of what you're building, the aesthetic defaults that shift depending on context, and the boundary between the two. tasteID measures both: 23 dimensions across two layers, plus context adaptation that tracks how your taste shifts between a serious tool and a playful one, between a marketing page and an editorial piece. Not because context makes taste random, but because it makes taste precise.

We don't ask you to describe your taste. People are unreliable narrators of their own sensibility — they describe what they think they should like, not what they actually gravitate toward. Instead, tasteID shows you pairs of real design moments and asks one question: which would you build? Not which is better — both options are equally good. The difference is taste, not quality.

Eighty-five comparisons. A Bayesian scoring engine that gets sharper with every choice. Conviction weighting so a strong opinion counts more than a hesitant one. Tiebreakers for the dimensions where you're genuinely ambivalent. The output is a three-section creative brief — rigid principles, flexible defaults, and your own words — that any AI tool can read and apply. Everything it builds starts to feel like something you would have built.

In a world where AI builds more and more of what we use, your taste is genuinely valuable. It's the difference between output that feels generic and output that feels like yours. That's exactly why no company should own it.

tasteID is open source. Your fingerprint is stored on your device, encrypted with a key only you hold. We can't see your choices. We don't run analytics on your taste. There are no accounts required, no tracking scripts, no data that leaves your browser unless you explicitly choose to share it.

The goal isn't to build a product around your taste data — it's to build open infrastructure. A portable format that works across every AI platform. Your taste profile should move with you the same way your email address does: owned by you, readable by any tool you grant access to, controlled by nobody else. We think taste portability should be a standard, not a feature.

What you see today is a starting point. The scoring model will get sharper. The comparison pairs will get better. New dimensions will emerge as we learn what matters. The export format will evolve as AI tools do. None of this is finished — and that's the point.

Because this is open source, it doesn't have to be one team's vision. If a comparison pair feels off, improve it. If a dimension is missing, propose it. If an export target needs a better format, build it. The best version of tasteID is the one shaped by the people who actually use it.

We're building this in the open because taste infrastructure shouldn't be proprietary. Follow along, contribute, or just use it — it's yours either way.