How does HackerTube select videos?

How does HackerTube select eligible videos?

The documented first pass scans archived comments for YouTube links and requires at least 100 characters of the commenter’s own text. It then applies the people-and-months threshold. At most twenty mentions are retained per video, favoring different commenters. That cap means the displayed mention count is not the total historical number of references.

How does HackerTube rank selected videos?

The export score counts distinct recommendation months, with extra weight for strong recommendations. The stance labels separate recommendations and strong recommendations from references, jokes and criticism. This is the exported ranking signal, not a complete specification of the live For You feed. Model labels can misread context, so a high score is not a factual accuracy certificate.

Why a mentioned video may not be recommended

Imagine a tutorial linked in three months by three people, each warning that it is misleading. It can satisfy a recurrence rule without receiving a positive stance. This hypothetical example explains why filtering by positive recommendations matters. The documentation reports that about a quarter of the exported videos have no positive recommender in the retained labels.

The threshold also favors material with time to accumulate mentions. Missing recent videos should not be interpreted as evidence that the community rejected them.