Test whether ASK actually helps

Compare two ways of selecting news, then judge for yourself. This is a blind comparison on clearly labeled synthetic test stories, not a live provider test or proof of commercial success.

Nothing here trains an AI model, contacts a news provider, or sends your choices to ASK. These votes last only while this page is open.

Choose a test to begin.

What counts as learning?

A ranking that gets better after explicit feedback is personalization. An AI model does not magically learn just because ASK saves a memory. We require measurable gains on held-out examples and must reject changes that worsen source traceability or privacy.

Track four outcomes for each new build: relevant results in the first three positions, links with identifiable sources, account/privacy regressions, and real user preference in blind tests. Keep the old release as the baseline.