Every deck has one: the market stat everyone repeats and nobody has sourced. "The market's growing 40% a year." "80% of teams plan to adopt this." It arrived via a competitor's deck via a blog via a webinar via — somewhere. You're about to say it to your board, and one director has a habit of asking, mildly, "where's that from?"

Claims like these are load-bearing. Strategy leans on them, hiring plans assume them, valuations quote them. And the chain of custody usually ends at a slide that cited another slide. Before it goes in your deck, it deserves ten minutes of adversarial sourcing: who actually said it first, in what context, with what basis — and does anyone credible dispute it out loud?

deep_search hunts the primary utterance: search the claim's substance with the factual quality verdict doing the filtering — surfacing content that cites specific data, named sources and first-hand evidence over content that just repeats the number confidently. Evidence timestamps land you on the exact moment the claim gets made (or made-up).

deep_query runs the adversarial pass: "what's the actual evidence for and against <claim>?" It reads across the record and returns both sides cited to the second — who asserts it and on what basis, who disputes it and why. Ten minutes later the stat in your deck either has a named, checkable source and a confidence level… or it's quietly gone, which is the cheaper way to learn it was folklore.