On every contested question, the internet serves you one expert at a time — each confident, each persuasive, each contradicting the last one you heard. Learn from any single voice and you inherit their blind spots along with their insights; the YouTube algorithm makes it worse by feeding you more of whoever you clicked first, until one camp's take feels like consensus.

Real understanding of a contested question isn't knowing the right answer — it's knowing the shape of the disagreement: what each camp claims, what evidence they lean on, where they talk past each other, and what they quietly agree on. That shape normally takes months of wide listening to perceive. It's also exactly the kind of thing you can just ask for.

deep_query maps disagreements natively: ask the contested question and request the split, and it researches across the camps' actual statements — returning who argues what, on what evidence, cited to the moments they said it. Not a both-sides summary; a structured map of the actual dispute, drawn from primary voices.

Follow-ups sharpen the map: where do the camps agree without admitting it? which claims are empirical (checkable) versus definitional (people meaning different things by X)? who has changed their position, and what moved them? — that last one being the highest-signal question on any dispute. You end up with the thing months of listening builds: the ability to locate any new take you hear on the existing map, instantly.