Spent the whole f* day working with this bleeding-edge-new-era technology. Had the concept paper and the strategic PRD for the MVP at hand. Had the established toolkit. Used all the power of this thing. Produced exactly zero lines of code.

Just because this thing can't process more than one requirement while producing even a code snippet. Not really. If it parses one thing, it omits two others. And that's even while using this fancy "spec-driven development".

Gosh. How cursed all of this is.

Only those who hated programming could say this is amazing. I've read about studies where apes were willing to do something for a reward if that reward wasn't guaranteed but randomized. This is the same. This is a slot machine. Feeling dumb and exhausted.
Alright, maybe, with some (decent?) effort, it might produce 70%, but what about the remaining 30%? That's where most of the complexity hides; what differentiates good software from shitty software. I don't know how folks tackle this 30% with these tools. From what I've read, they decided that 70% is what the business needs, and since we're all supposed to be business-oriented, that's enough. But is it enough for you, as a craftsman?
In one of the greatest books about software engineering "Peopleware: Productive Projects and Teams" (published in 1987, just for the record), Tom DeMarco and Timothy Lister wrote that developers’ ability to deliver quality work (at their own standards) at a sustainable pace (and the environment that supports this) is crucial to team success and individual satisfaction. It seems this is no longer considered important.
By the way, I think I'm starting to understand this take "don’t read the code you produce", because otherwise, if you actually read the code, you'll die validating/reviewing/forcing these tools to produce something you'd write yourself (pursuing quality, stability, maintainability, security concerns). And ultimately, you won't gain any of the promoted boosts. Everything exists for a reason!