It's quite simple.

I perfer building and using software that is robust, heavily tested and thoroughly reviewed by highly experienced software engineers who understand the code, can detect bugs and can explain what each line of code they write does.

Today, we are now in the phase where embracing mediocre LLM generated code over heavily tested / scrutinized code is now encoraged in this industry - because of the hype of 'vibe coding'.

If you can't even begin to explain the code or point out any bugs generated by LLMs or even off-load architectural decisions to them, you're going to have a big problem in explaining that in code review situations or even in a professional pair-programming scenario.

LOL.. I was going to say after working in the tech industry.. half the time it is a rats nest in there.

There are excellent engineers.. but their are also many not so great engineers and once the sausage is made it usually isn't a pretty picture inside.

Usually only small young projects or maybe a beautiful component or two. Almost never an entire system/application.

exactly, as smart engineers end up having to work with midwits and it's not going to be pretty.

Unfortunately, all of modern software depends on some random obscure dependency that is not properly reviewed https://xkcd.com/2347/

It's a funny comic, but can you actually give an example of what it's talking about? "Properly reviewed" can be construed as "has been working for a long time for a lot of people", which definitely can't be said about any AI process or any AI generated code. At the very least, 1 human person actually sat down and wrote the tools the comic is poking fun at. But with AI, we are currently producing code that was neither peer reviewed nor written (a process which includes revision) -- it was instead "generated". So it's still a step backwards.

I guess you need an active developer license to write blog posts