A lot of these are just sticking a ui in front of someone else's AI and silently feeding it extra prompts.

This adds a tiny amount of value, sure, but enough to gamble millions on? Obviously not.

No wonder they failed

Did delivering quality really get easier? I'm certainly not seeing it in the software I use. Delivering scale doesn't mean the mission was executed well.

Just like how more RAM and compute in the cloud made us worse engineers (no need to turn your brain on to tune performance).

When the brain is off for one thing, it's off for the rest as well. There is a lot of talk about "we don't have to think about code so we can think of ideas", but that's not how it works. We just don't think.