Hacker News Reader

What is it about streaming specifically that necessitates this? Am I missing something obvious?

Excuse is probably that classifiers for streaming are less robust.

It's easier to get a partial response for something like a CBRN topic