The counterpoint about 'polished generic framings' is real, but I think there's a middle path: using the LLM as a sparring partner rather than an answer machine. The value isn't in accepting the first response - it's in the back-and-forth. Push back on the generic framing. Ask 'what's wrong with what you just said?' The struggle to articulate doesn't disappear; it just gets a more responsive surface to push against.