Ship the one people actually pick.

One human decision. A person looks at your two versions, picks one, and optionally tags why. That single act is the unit Boop is built around — it's the vote, the credit and the name.

Other people using Boop: founders, designers, developers, marketers and anyone who wandered into the feed. It's a general audience of internet builders, not a screened research panel. That's the right sample for “does this read as trustworthy” and the wrong one for “would a hospital procurement officer buy this.” Audience filters and private panels are on the roadmap, and are marked as not built yet.

Yes. Unlisted tests stay out of the public feed and are reachable only by link. Private tests are visible to you alone and are how you'd run a test against an audience you bring yourself. Public tests are the default because they're what earns you responses from the feed.

New accounts get 10 credits. You earn 1 more for every 3 boops you give, plus a bonus for leaving written feedback. Requesting one human response costs 1 credit. Buying a pack is a shortcut, not a requirement — the loop works without spending anything.

Yes, and it's one of the most useful things to test. Type two names, taglines or CTA lines and Boop typesets them properly rather than showing them as raw text, because how a line reads at size is part of what you're judging.

You show one design for exactly five seconds, then hide it and ask a single recall question — what do you remember, what does this do, what would you click. It measures whether anything landed at all, which a head-to-head comparison can't tell you.

No. Every vote and every comment comes from a person. AI is used for exactly one job: reading the written feedback and grouping it into themes so you don't have to skim ninety comments. It is labelled as such everywhere it appears, it never casts a vote, and it never changes a count. Sample tests seeded with the app are clearly marked Demo.

With a 95% Wilson score interval on the leading variant's share. A winner is only declared once the whole interval sits above 50% — meaning the observed preference is consistent enough that a coin flip is excluded at that sample size. Below that you'll see “Early signal” or “Moderate signal” instead of a result. It describes consistency at this sample size, not a guarantee about everyone on the internet.