Launch week generates a feedback firehose with a deadline attached: the first-look and hands-on videos land within days, they shape the launch's public narrative, and they contain your earliest unfiltered read on what shipped well — but only if someone processes them while the launch team can still respond. By week two, the videos are watched or wasted; the iteration window they informed has moved.

The PM reality is crueler: launch week is exactly when nobody has three hours for YouTube. So the earliest, most consequential external feedback arrives precisely when the team is least able to consume it — unless consumption stops requiring the hours.

deep_search catches the wave daily: uploads about your launch, filtered to today/this_week, newest first — first-looks, hands-ons and early reviews as they land, with evidence snippets showing what each covers.

run_youtube_tool compresses each into triage form: summarizer for the take, key-quotes for the verbatim reactions — what delighted, what confused, what got compared to rivals, in the creator's own words. By end of each launch-week day, the team reads a feedback sheet: reaction themes, quotes with sources, confusion recurring across creators (that's your docs hotfix), and the early narrative forming (that's comms' brief). The window stays open because the processing stopped costing hours.