The pitch meeting shows you a founder at their most rehearsed — every answer pre-run past advisors, every number rounded in the friendly direction. Reference calls show you their curated fan club. What you actually want is the version between those two: how this person talks when the stakes are lower and the interviewer is just curious.

That version exists in volume. Most founders raising anything have a public trail — podcast episodes, demo days, panel Q&As — stretching back years. The trail is where character shows as an arc: how they described the same pivot in year one versus year three, whether the churn question gets a number or a story, whether competitors are dismissed or understood. One interview is a performance. Three years of them is a pattern you can diligence.

deep_search assembles the founder's public record: interviews, podcasts and panels found at the transcript level, sortable oldest-to-newest so the arc is readable in order. Evidence timestamps mark where the topics you care about — churn, competition, the failed first product — actually come up in each hour-long episode.

deep_query interrogates the record like a diligence associate with unlimited patience: "how has this founder's story about their market changed?", "how do they respond when pressed on retention?" Answers come back cited to the second, so what lands in your IC memo is quoted behavior, not vibes from one Zoom call. Consistency across years is a signal money can't buy — and inconsistency is one you can't afford to miss.