We need to separate two concepts with the AI conversation. Is it a real technology? Yes, it is. And it works, often. It isn't the scifi world ending technology they want you to believe in, but damn can it make an android app quickly. And get most of the buttons right, if you test it enough and ask nicely.

False, it's able to fool c-suite execs and gullible rubes into thinking they can build an app. In reality, these vibe coding projects all fall apart. Zitron relies on Nik Suresh's arguments, which I agree with, that software development: shipping, running, and supporting software, is a hell of a lot more than just code.

The bad quality of the code and resulting app is really a secondary discussion to whether this can really count as software development. One of my favorite software development tools of the last decade is actually a Microsoft standard called LSP, which allows text editors to talk to compilers to get code suggestions and linting as you type. You used to have to use an 'integrated' text editor, or IDE, that usually only worked with one or a handful of languages. Now, everyone can use their editor of choice.

The ai craze, to me, is much lower in productivity gains than LSP for software development, it might be negative because you are handing off the one part of software development that ties all of its subdisciplines together. It's crazy how fast the industry standard went from 'clean code' and concerns about whitespace, furiously tracking issue lifetimes, to just throwing up their hands, actively telling people not to care. I hate both paradigms, to be clear, but at least they were trying to measuring things in the former.