> During Friday's AMA, Altman admitted the routing system that automatically selected which AI model to use had malfunctioned on launch day. "Yesterday, the autoswitcher broke and was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber," he wrote.

I mentioned it here previously, but I am inclined to distrust the veracity of this defense and view it as a good scapegoat for why the default performance sucked. I'd be more likely to believe that it was intentionally neutered (purposely using weaker models more often, heavily quantized, whatever) to reduce costs or whatever other reason you can think of and this "the autorouter was completely broken and we totally didn't notice for weeks on end" is just damage control (I don't think sama has earned any level of "just trust me, bro" credit).