This explains the success of StackOverflow: don't have to wait on your colleagues or even your own investigative work, the answer is usually a Google query away.
This also explains the success of ChatGPT in dev work: StackOverflow is still too much bullshit to deal with. LLMs strip that all away. It's such a big difference it's worth risking being led astray by hallucinations.
For coding, if the LLM hallucinates, you feed the error message into the LLM and it apologises and corrects itself. For bonus points, you can tell it to run the code until it works.
This also explains the success of ChatGPT in dev work: StackOverflow is still too much bullshit to deal with. LLMs strip that all away. It's such a big difference it's worth risking being led astray by hallucinations.