All my stackoverflow-style queries are now going to whatever AI chatbot is most accessible when I need my answer.

They tend to provide answers that are at least as correct as StackOverflow (i.e. not perfect but good enough to be useful in most cases), generally more specific (the first/only answer is the one I want, I don't have to find the right one first), and the examples are tailored to my use case to the point where even if I know the exact command/syntax, it's often easier to have one of the chatbots "refactor" it for me.

You still want to only use them when you can verify the answer and verifying won't take more time. I recently asked a bot to explain a rsync command line, then finding myself verifying the answers against the man page anyways (i.e. I could have used the manpage instead from the start) - and while the first half of the answer was spot on, the second contained complete hallucinations about what the arguments meant.

I am using free chatgpt and free deepseek mostly.

They are both terrible in terms of correctness compared to duckduckgo->stackoverflow.

As an example deepsek makes stuff up if I as for what syscall to use for deleting directories. And it really misleads me in a convincing way. If I search then I end up in the man page and I can exentually figure it out after 2-3 minutes

Also with AI, I get an answer instantly—no snark, no misunderstanding my question just to shut it down, and no being bounced around to some obscure corner of Stack Exchange.