Skip to main content
The recipes form one workflow: build a verified view of an AI-market question, then keep it current. Each one takes the previous step’s output, so they read in order. A signal is a source: an essay, podcast episode, or newsworthy event. Gildea decomposes each into verified units: its individual sentences and claims. Themes tag the signal; entities tag both the signal and its units. Discover works bottom-up: the data surfaces which themes and entities matter right now. Retrieve works top-down: you pull the verified units from the signals in your scope. Verify and Embed both act on those units, one (optionally) checks a unit against its source, the other saves it alongside your own data. Refresh pulls only what changed as new signals land. You don’t have to run all five. Start at Discover when you don’t yet know what to track; start at Retrieve when you already do.