not on my most optimistic dreams I expected someone to "get it" as much as Giulio does and yes that's the point today models are already incredibly capable, and they're smart enough to build infinitely complex software, but mistakes accumulate, and it reaches a level of tech debt where progress stalls. if there only was some magic way to tell the agent "hey, this is wrong, try again" until no bugs are left, then, there would be no limits to the complexity of software that AI can build that's what Bend does!! sure, it takes more time (and tokens) to solve the same problem, but that's exactly because writing correct code takes time! Bend refuses any incorrect output, and that's what you WANT to happen. merging bugs give you an illusion of progress, but it bites later I for one am not using any other language anymore. if Bend is an absolute failure, at least it is the language I'll be using for my own projects moving forward (and Bend in Bend is already in progress)