I mean, if we are vibe coding the future let's go all the way.

This is surprisingly similar to a book my son got from the library about how to code for 5 year olds or something along those lines. The example was how do you tell a robot to tie their shoes? With plenty of examples of the robot tying shoes not on its feet, or placed on its head, etc.

The underlying point: computers are stupid, but iteratively less so if you course-correct them a little bit each time. There is absolutely no reason we should need to bother with human readable computer languages within 10 years - straight to Assembly! Blazingly fast and self-healing and inscrutable.

Now we will still need "computer operators" to define what we want with as much fidelity as we need, but ask yourself this - if we could guarantee the program wpuld be correct within the boundaries we define, aren't we in the exact same position as today? It comes down tp our fallible, human ability to precisely tell the computer what we want. It's just a change in the I/O, and potentially a big change for the better.

I need a simpler explanation of this.

No special subagents, Ralph just forks himself, right?

But I think I missed a lot.

"You'll get 90% done with it", but also it won't be maintainable, but will maintain itself and fix all the errors it causes. Except permanently 10% broken?

> If you wanted to be cheeky, you could probably find the codebase for CURSED on GitHub. I ask that you don't share it on socials, because it's not ready for launch.

I'll respect the authors wishes not to link it. But I have to say, while my expectations for code quality were low, what I found was far below them.

From the github UI:

> Sorry, we had to truncate this directory to 1,000 files. 631 entries were omitted from the list.

"These packages taste like burning."