How the VP0 algorithm works
Most platforms treat their feed as a black box. Ours is documented here, completely — because a feed you understand is a feed you can trust. Every card also explains itself: tap ⋯ → “Why am I seeing this?”.
What we measure
Only your behavior on VP0, tied to a random anonymous id (or your account once you sign in): which designs you see, hover, open, run in the simulator, copy, save, skip, or mark “show fewer like this” — and how long you actively spend with them. No third-party data, no cross-site tracking, no contacts, no location.
How ranking works
Candidates come from parallel channels — your interest tags, designs co-engaged by people with taste like yours, creators you follow, this week’s trending outliers, brand-new designs getting their fair chance, and things you loved that deserve a revisit.
Each candidate is scored by a blend of signals: interest match, co-engagement similarity, proven quality, freshness, weekly attention, creator affinity and creator quality — minus fatigue for things you keep scrolling past. As our data grows, a learned model predicts how likely you are to open, run, or copy each design and ranks by that expected value, with the values themselves derived from what actually makes visitors come back — not from what merely gets clicks (clickbait with high opens but no real attention is demoted automatically).
The page is then re-ordered for variety: similar designs are spread out, no creator gets adjacent slots or more than a few per page, and one slot per stretch is deliberately reserved for something outside your comfort zone.
How new designs get their chance
Every newly approved design climbs a ladder of guaranteed exposure batches. Perform with the audience at each rung and it climbs to a bigger one; underperform and it parks — still searchable, just not pushed. No follower counts, no pay-to-play: the audience decides.
When you say no
“Show fewer like this” doesn’t just hide one card. It suppresses the design hard, cools everything that feels like it, and rests that creator in your feed for two weeks. Skipping several similar designs in one session cools that theme for the rest of it.
How we keep ourselves honest
Ranking changes must pass a fixed “golden” ordering test in our build pipeline and prove themselves on replayed real traffic before they ship. Live experiments run against the previous ranker as a permanent control, and a guardrail automatically shuts off any variant that measurably makes the feed worse.
Your controls
On every card (⋯): Why am I seeing this? and Show fewer like this. Anonymous browsing is always available — clear your browser storage and the profile is gone. More controls (muting creators and topics, an exploration dial) are on the roadmap and will be documented here when they ship.