Search
Search every extracted claim across the corpus — not just the concepts we've pre-tagged.
Keyword matches the literal words you type (D1 full-text search). Meaning finds claims discussing the same idea in different words — e.g. a "sugar cravings" search surfacing a claim about "comfort eating," with no words in common — via semantic embeddings (Vectorize). Keyword is exact and predictable; meaning is fuzzier and better for open-ended exploration ("what does the corpus say about X") than a precise lookup.