A famous investor speaks for ninety minutes; the internet keeps eleven words. By the time the take reaches you — via a headline, a tweet, a summary of a summary — "we're cautious on duration risk in one specific scenario" has become "LEGENDARY INVESTOR PREDICTS CRASH." The paraphrase cycle doesn't just compress; it systematically distorts toward whatever gets clicked.

If your work quotes these people — in notes, memos, client letters — the distortion is your liability. "Did they actually say that?" has exactly one honest answer: go to the tape. The tape is public. What's missing is a fast way from the viral fragment back to the full sentence, in context, with a timestamp.

deep_search finds the source appearance: search the investor and topic at the transcript level, and evidence timestamps land you on the exact passage — not the video, the moment. The interview a quote came from stops being a needle in a media-tour haystack.

library_import_youtube_video pins the source appearance into your library, and get_youtube_transcript hands you the full text around it — pulls are free once the video is imported. Now you see what the eleven viral words were extracted from: the qualifier that preceded them, the scenario they were conditional on, the follow-up where the interviewer pushed and the answer got nuanced. Your note quotes the actual sentence with a timestamp — and when someone forwards you the headline version, you're the one who can reply with what was really said.