Every industry maintains a public list of its unsolved problems. It isn't written anywhere — it's spoken: the moment in a panel when an expert sighs "honestly, nobody's cracked this," the podcast tangent where two operators trade complaints about the tool that doesn't exist, the Q&A answer that starts with "great question, and the honest answer is we all just work around it."

Founders hunt for exactly these moments and mostly find them by luck — you happened to hear that episode. The systematic version is a different game: collect what practitioners across a field repeatedly name as broken, missing or endured, and you're reading a demand backlog written by the customers themselves, before anyone builds the supply.

deep_query is built to answer questions whose evidence is scattered across hundreds of conversations. Ask it what experts in a field describe as unsolved and it researches the spoken record — panels, podcasts, conference Q&As — and returns the recurring complaints with citations to each moment of frustration.

The output reads like an opportunity map with provenance: which problems come up once (anecdote) versus everywhere (market), who exactly suffers them, what the current workaround costs, and the language practitioners use — which later becomes your landing-page copy, because it's literally the customer's phrasing. Re-run it quarterly and you also see which whitespace is closing, and which keeps being named year after year because it's genuinely hard — or genuinely yours to take.