[txutxu]

Not sure about creativity. But.

Previously, I had to deal with "Junior python" or "Junior bash" crap at $work.

Finding the dangerous bugs, was measured in seconds. Helping the person to be better in the future, used to work.

Now, all the company is requesting code to ChatGPT. Hey, please, use my script.

I have to deal with "looks fine at first/quick view" code, that needs deep analysis to understand what it's trying to do, why, and where are the (100% sure) hidden "break production" kind of failures.

It's more like "were is waldo"... you know that there will be at least one or two things really really wrong, always, always one or two (or more) catastrophic details, but that they are hidden below something "apparently nice".

And what is worse, all effort to point and fix issues, is lost. Or repeated again and again.

I apologize, but as a senior sysadmin/oncall model, I cannot run your chatgpt code, until you understand how things work.

[brabel]

While I sympathize with your point, I found this passage funny:

> code, that needs deep analysis to understand what it's trying to do

Because if there's on thing AI can already do really well, it's explaining what some code is doing.

[the_other]

> Because if there's on thing AI can already do really well, it's explaining what some code is doing.

But it just makes up a series of words which sound good. It's accidentally correct 75% of the time and just wrong the other 25%. (and IMO, just wrong in some way almost every time I use it).