does anyone else feel like a single model just agrees with whatever you already think?

been using gpt pretty heavily for decisions and i keep noticing it kind of mirrors me. if i frame a question like im leaning towards option a, itll gently back option a. reframe it leaning b and suddenly b is the smart move. good writer, but a bit of a yes-man when it comes to actual judgement calls.

what ive started doing is running the same question past a few different models and reading where they disagree — thats usually the part worth thinking about. the disagreement surfaces the stuff one model alone glosses over.

got kind of obsessed with this and ended up building a little iphone app called war table that does it for me (5 models argue, then one verdict) — wartable.co if youre curious, but thats not really the point of the post.

real question for you all: do you trust one model for the hard calls, or cross-check across a few? and if you cross-check, whats your setup?