> it's also possible that you are indeed deciding (skewing) most of your interviews in the first several seconds.

Not the original commenter, but I've had interviews where I was feeling pretty negative about a candidate in the first 5-10 minutes, but ultimately recommended them (sometimes even quite enthusiastically).

Given that empirical evidence to the contrary, it seems like a pretty incredible claim to insist that I'm deciding primarily based on the first few seconds. Do you actually have anything to back up your claim? That would suggest that, for instance, performance on a coding exercise has absolutely no bearing on my recommendation, since it usually takes a few minutes to get the coding rolling.