Every team has these arguments: "the customer explicitly said latency was the blocker" / "no, they said pricing." "The keynote promised that integration" / "it promised a beta." Two confident memories of the same recorded event, escalating in Slack, resolved by whoever remembers most confidently — which correlates with seniority, not accuracy.

The event was recorded. The argument is optional. What's been missing is retrieval cheap enough to use mid-thread: nobody scrubs an hour of video to win a Slack argument, so memory stays the arbiter and the confident stay undefeated.

deep_context is the arbiter, given an imported archive (your all-hands, customer sessions, webinars — see the institutional-memory use case): ask what was actually said about the disputed point and the answer comes back cited to the exact chunk — video, minute, wording.

get_youtube_transcript supplies the verbatim layer when the exact phrasing is the dispute: the full passage around the moment, quotable into the thread. The Slack argument ends with a link and a timestamp — "23:41, here's the sentence" — and over time the team's epistemics quietly upgrade: people check before claiming, because checking became cheaper than being wrong.