Comment by Groxx
14 hours ago
Plus like half of the time this happens to me, the summary is WILDLY inaccurate about what the source states.
LLMs are quite good at "I need a thing that does [vague description]" searches, FAR better than ~all "semantic" searches. But oh boy. I do not like them "answering" queries. Feels extremely dangerous.
Its weirdly hit or miss, not sure why, HN data is available in reasonably annotated forms.