don't you see that all these cool people in tech are jumping off the bridge, aren't you curious to jump off too? you aren't? what's wrong with you https://registerspill.thorstenball.com/p/they-all-use-it
@cwebber a few weeks ago, a coworker asked me if I had time to help him debug an issue with a script. He was using a function that I had never seen before to set up the file structure, and it was telling him "function not found". He asked me if it might be an issue with the library installation, where the interpreter couldn't find the function. I suggested that maybe he misspelled the function name, and asked him what did the docs say?
He told me "I dunno, this is just what ChatGPT told me to do".
...
...
...
...............
My brother in CHRIST you are a DEVELOPER of this library. WHY DIDN'T YOU JUST READ THE FUCKING DOCUMENTATION????
And this is why I will never use LLMs. I would like to avoid using the inefficient energy-wasting machine that spits out plausible-sounding but subtly incorrect answers, thanks. I prefer my code to actually be correct and work, thanks.
This has been my experience; I'll have an straightforward and obvious function that will nevertheless take me 20 minutes to write, so I'll think "well, let's see if ChatGPT can just knock it out for me". I then spend the next hour pointing out to the stupid thing the edge cases it doesn't handle and various other failure modes that a real programmer would never miss.; it replies "you're right" followed by a decently accurate restatement of the defect, then turns around and produces a new version that doesn't fix it, and/or introduces a new one. In the end i give up and write it myself.
