Hyperblues

For You Study

Reference implementation

A live personalized feed by @spacecowboy17.bsky.social. Public metadata needs no API key; personalized output requires a signed-in Bluesky identity.

Generatorforyou.club
StatusChecking…
Feed likesChecking…

Published model

Published facts are separated from hypotheses we still need to test.

1 · Viewer taste

Uses the viewer’s likes as positive preference signals.

2 · Taste neighbors

Finds people who liked the same posts as the viewer.

3 · Candidate transfer

Shows other posts those similar people liked recently.

Likely strength

Inference: discovery beyond the follow graph with a personal relevance anchor.

Likely risk

Inference: popularity, taste-cluster lock-in, and repeated authors may compound without diversity controls.

Hyperblues test

Compare taste-neighbor candidates with One Per Person; vary only author caps, recency, and exploration share.

Inspect your live result

The sample is fetched through your linked Bluesky account and summarized locally in this browser.

Ready. Sign in and link Bluesky under Account if needed.
No sample loaded.