Jan Lehnardt 
First time working with someone who habitually uses LLMs. After respectfully requesting to not get pasted pages of output in chat conversation, they started copy-pasting selected passages in response to the things I hand-write. I notice increasingly that the human is just copy-pasting LLM output back and forth instead of thinking for themselves and maybe noticing that half the time the content is obvious and the other half irrelevant. Making progress this way is like walking through molasses.
May 11 at 5:22pmWeb
Show threadPositivDenken 🤯May 11
@janl we used to compress content over the wire, now we inflate it.
Show threadmaxine 🇵🇸May 11
@janl This is genuinely the worst thing I keep repeatedly encountering, but I keep wondering - surely neither one of us can be happy in this arrangement? Like I'd rather the LLM interface just go and do whatever non-work hobbies they have and let me do whatever I need to in peace than waste my time with a conversation that I'll still need to supplant with personal research into whatever topic/work history they were supposed to inform me on.
Show threadWolfram wants peaceMay 11
@janl and it guides your thinking instead of you guiding the process or the thinking, and it is easy to get misguided and totally loose control
#2cents
Show threadJan Lehnardt May 11
@wolframkriesing I see no guidance whatsoever
Show threadKaterina MarchánMay 11
@janl wow this is so fucking disrespectful
Show threadBen RamseyMay 11
@janl This seems to happen a lot, and it wouldn’t be so bad if they prompted the LLM to keep its answers short to a few sentences at most.
Show threadtanteMay 11
@janl wow. What a shitty behavior
Show threadDarachMay 11

@janl It will absolutely get far worse. I'm using LLMs / agentic at work and the spectrum of variants of good verses bad use is far more extreme than with manual methods.

The key is level of active thinking and understanding which I don't think has changed much. Rather, consequences are more extreme and accelerated...

Show threadseism0saurusMay 11

@janl

I offen get request from a coworker who obviously let a llm write for him. I just read what he wants and ignore everything else. When I put a lot of smileys and useless Unicode chars in my answer. Often it takes more time to do that then the actual work. But a little bit if sabotage is required when handling people with ai. Sometimes in weiten dear Claude instead his name. He never complained.

Show threadGavin LoganMay 11
@janl I have a colleague who does this, I told the team it will cost them £1 every time they want me to respond
Show threadMikhail 💛💙May 11
@janl People copy paste my PR reviews into LLM and paste back answers :(
Show threaddr 🛠️🛰️📡🎧May 12
@janl Someone I otherwise respect a lot has started producing AI code and quoting AI on powerpoint charts. It's very sad.