the halving of error rates for image inputs is pretty awesome, this makes it far more practical for issues where it isn't easy to input all the needed context. when I get lazy I'll just shift+win+s the problem and ask one of the chatbots to solve it.