Before designing the flow, every designer does the same furtive research: open five competitor products (the ones with free tiers), click around, screenshot, guess at the reasoning. The survey is limited to what you can access, biased toward what you already knew, and shallow on the why — you see the pattern but not the narration.

Product demos fix all three limits, unofficially: companies demo their best flows constantly — launch videos, webinars, conference walkthroughs — narrating the reasoning as they click. "We redesigned this so users wouldn't have to..." is design rationale, spoken aloud, about the exact pattern you're weighing. The whole market's design decisions, demonstrated and explained, sitting in public.

deep_search collects the pattern corpus: demos and walkthroughs across your space — including products you can't access, in markets you don't work in — found at the transcript level, so "demos showing how products handle <interaction>" returns those moments precisely.

library_import_youtube_video and deep_context turn it into a queryable pattern library: import the relevant demos and ask across them — "how do these products handle <flow>? what reasoning do presenters give? where do the approaches split?" — answered with citations to the demos. The design review stops being five screenshots and a hunch; it's the market's approaches, with their narrated rationale, compared in one pass.