UI
Two screen variants
Render two versions of the same page. See which one users land on without bouncing. 200 humans, ranked by morning.
←Datapoint
For developers
An API call away from human feedback
Drop a comparison into your code. A few minutes later, get a ranked answer back from real users in the segment you specified. Python, TypeScript, or curl.
How it works
Treat human feedback like any other API call.
One-line SDK
Python and TypeScript. Pass two things to compare(), get back which one humans prefer. The full API surface is in the docs; curl works fine too.
Compare almost anything
Two screenshots. Three CTAs. Five icons. Audio clips. Video cuts. If you can render it, you can rank it.
Real users, not panelists
We recruit through ads, so respondents come from the audience you'd actually ship to instead of a pool of professional researchers.
Pay per response
No seats, no contracts. Free credits to start. Built for "I'll just try it" engineering instead of procurement cycles.
What developers test
Skip the Discord poll. Ask real users.
COPY
CTA wording
"Get started". "Try it free". "Start building." Find the one your users click instead of the one your team argues about.
NAME
Project name
Five candidate names against five logo concepts. A ranked winner before you pay for the .com.
Frequently
Asked Questions
datapoint.compare(["Option A", "Option B"], audience="US adults 18-35") returns a ranked result. Python and TypeScript. The full SDK lives in our docs.
No. It's an API for human preference. You hit it like any other model: pass two things, get back which one humans prefer and why. The same primitive used to train preference models, exposed as a tool.
Yes. Pass a URL, file, or base64 blob. Mix media types in one comparison if you want.
Minutes and cents. Most comparisons return in minutes. Pay per response, no seats. Free credits to start.
Drop a comparison into your next deploy.
Free credits to start. Python and TypeScript SDKs in the docs.