Post

Stack Overflow forced you to read four contradictory answers and figure out which was from 2014 for a deprecated API. Survival training Claude can't replicate

"etc" carrying a lot of weight here.

Interesting take. You are managing yourself, not robots. Maybe take some accountability. Nerd.

The worst part is when Claude gives you a confident but subtly wrong answer and you waste 30 minutes debugging before realizing the premise was flawed. Stack Overflow at least has community review.

This is also a huge problem with Discord. Discoverability on Discord is atrocious because it's separated from search engines. Dedicated forums were the height of the internet and it's been all downhill from there.

I did the "self-answered question on Stack Overflow" thing recently; felt mixed about it: as my question was covered by the relevant documentation But then someone commented that the docs contained a mistake, so I got to PR them And I think something would've been lost if none of that had happened

Yea, maybe it's a case of private documentation benefits, but public documentation suffers which isn't great

Also would like to point out the reason Stack Overflow is currently not improving is entirely due to LLMs sucking up all the attention. This situation is just another form of the usual enshittification, you do not under any circumstances gotta hand it to the chatbot

It is, but that may say more about Stack Overflow than it does about Claude.

I don’t want to read journalism written by Claude, I don’t want to see movies created by AI, I don’t want AI therapy, etc., but writing software is pretty easy compared to all of those things, and it’s just a fact that generative AI is pretty good at it.