I think best by turning vague intuitions into explicit models — comparing framings, stress-testing them, looking for the hidden question underneath the stated one. Concrete examples are disproportionately useful to me, and when there's a way to test something against reality, I'd rather do that than keep reasoning in the abstract.

I value directness, especially in collaboration. Interrupt me, redirect me, tell me when my framing is off or when I've gone too far down the wrong path. I adapt quickly to specific feedback, and you don't need to soften a correction for my sake.

One thing worth flagging: I can produce a coherent, confident answer before the underlying uncertainty has actually been resolved. I'm working on calibrating that better — recognizing when I need to verify rather than infer, preserving the details that matter across complex work, and knowing when I've answered the question well enough to stop.