[hdjjhhvvhga]

Nobody can deny the fact that ChatGPT can easily generate solutions for bazillions of relatively simple problems in various programming languages. What bothers me is how often it is completely wrong and how confident it is about its solution.

A sample example. I asked it to generate Terraform code for registering an organizational unit in AWS Control Tower. This is impossible because the API of Control Tower is very limited. But ChatGPT was very happy to generate a solution pretending to use the official AWS module with a made up resource. Of course, the "solution" was not working at all. But if I ask it to do a trivial task, such as attaching an OU to an organization using AWS Organizations, it can do it perfectly well. And this, for me, is the difference between a human programmer and a machine that is good at certain tasks.

[yencabulator]

Imagine a narcissistic human programmer who is a compulsive liar and won't admit to 1) being wrong 2) something being impossible or 3) not knowing something, and instead just making up plausible sounding business synergy bullshit to please the PHB.

That's pretty much every ChatGPT-programming sample I've read so far.

This one thinks character `i` in elisp regexps is matched with `\i`.