Viewer Pulse

What viewers are actually arguing about, taken from the volume and shape of public comment rather than from a handful of quotes. We classify comments mechanically, publish the classifier’s known failure modes alongside the result, and never reproduce a comment author’s name, avatar or channel.

Aggregate counts here are floors, not exact values: the classifier reads text patterns and misses anything phrased unusually, and the categories can overlap, so they must never be added together. Every piece in this section repeats those limits next to the numbers rather than burying them at the end.

This is the smallest section on the site, and it will stay that way for a while. A piece here needs a corpus big enough that the classifier output means something — we hold 12,211 comments across 297 videos, which is enough for a handful of questions and nowhere near enough for most of the ones people ask. When a question cannot be answered from what we actually collected, we leave it alone rather than write around the gap.

How we source and check everything on this site is in our editorial policy.