Finding a good faith interlocutor for this will be tricky. You'd need someone who made a earnest effort to use AI for practical work, using current models, and didn't see benefit. The most engaged skeptics in my world don't meet this low bar.

It's a monster in my small benchmark of practical coding problems. Not just in problem solving, but it's also the only model out of 20+ that I test that made zero prompt adherence errors.

Solid suggestion, thanks. I'll add a screenshot before the next release.