Comment by maccard
15 hours ago
At my last job, a coworker did this to me. The first time it happened, I ignored it. The second time, I responded in public saying “thanks but I can ask Claude myself.” Nobody ever pasted me an LLM response again. YMMV with team size and seniority though
At my current job I just asked a coworker in chat if the reply was his or LLM. He quickly owned up and the incidence of brokered LLM messages in chat dropped signficantly.
Unfortunately people are being encouraged to use LLM in docs, emails, and presentations so it's probably a losing battle.
That's a better approach than what I did, and I'll use that going forwards.
We had an exec show how he was using AI to write interview feedback from his very sparse notes, in front of a few thousand people.
We have a LLM bot in chat that analyzes bugs when.you drop in a link and anything else if you ask it directly. It's actually quite effective: all the slop is now localized and can be ignored/blocked easily, and people no longer feel the need to copy and paste slop themselves.
I've been thinking about these kind of interactions a lot lately. I don't know what it says about human communication patterns but I'm not super excited about it.
I've been guilty of acting as a middleman for review comments from difficult coworkers. I farm out my interaction to avoid the confrontation, mental exhaustion, and to avoid negative feelings I got from those reviews. I don't mean 'my work is being challenged', I mean people who: delay, nitpick, blow up architecture after not participating in design, assert false info as truth and require you prove them wrong, require 10x evidence of everyone else. Not people who are like 'Hey, I don't understand this code can you add some comments or lets hop in a room for review'
I've also been on the receiving end. I like to think I'm NOT one of these toxic reviewers.
Now, we have Business Owners and Managers who are now even further away from engineers. Important meetings have 0 engineers in them ( a couple people who think they're engineers because they were for 5 years in the 80s ). Output is an AI generated transcript that gets copy/pasted into copilot to generate jira tasks by the Product Owner.
It seems like we (humans) really don't like working with other people. Or, maybe we like working with people who can advance our careers and want to automate interaction with people who cant?
I think this effort by Business Owners to set up an AI wall may be the closest I've seen to 'digital classism' (maybe there is a better term for it?).
>Or, maybe we like working with people who can
https://croissanthology.com/earring
We like whispering earrings. We'll follow a low friction happy path even if it leads us off a cliff.
I did a similar thing and I got reported to HR.
What was the outcome of the report to HR?
I think if you work somewhere that this is the result, you're unlikely to have much progress challenging anything directly. In order to actually have change you need to be able to challenge and stamp out these sorts of behaviours.
> What was the outcome of the report to HR?
"Do not make negative assumptions" basically. But also a drop in how often and quickly I did reviews.
Would you be OK with someone looking up the answer on the internet?
You mean like when juniors would copy paste Stack Overflow without verifying or even understanding it?
No. We hated that too.
Actual research is, of course, fine.
I see a possible misunderstanding on my part here.
My experience has typically been that engineers respond, "Claude said..." and then say what they think about it. If people are really receiving just a straight dump of AI response from their colleagues, then yeah, I would be disappointed too.
But, I have no problem receiving from a colleague, "Claude says ..., but I don't know what to make of it." The embedded space can be really hard, so I don't fault someone for feeling clueless about something. I feel that way every time I get handed something new to work on which I have zero experience with.
1 reply →
[flagged]
This smells an awful like a post-hoc justification of why you can’t answer it yourself.
If you _could_ do that and hand the information to Claude, you could also just…give the person that information directly, and avoid wasting their time and the tokens.
Uh? I don't understand, are you saying that instead of asking Claude myself I should write a question for other people to do it? Which is by the way only the initial part of the work I described?
2 replies →
You can also respond ‘I don’t know’ instead of doing the equivalent of responding with the first page of an ‘I feel lucky’ google search
This is a very good analogy. You have to do way more effort to make sure it's correct than if you have asked it yourself with a controlled context. I generally don't want to feel lucky when I ask questions to coworkers.
> Then do it and don't ask questions.
I'm not sure if you're trying to be funny, but this is such an unintellectual take that it veers into caricature. Props if that was your purpose.
No, it was a suitably nasty reply to someone who boasted of being nasty to colleagues trying to help.
If your question can be answered by Claude, and you didn't ask Claude yourself, and when other people do it for you, you complain that you could have done it yourself, then my honest reply is that you should just shut up and learn to use Claude.
5 replies →
If I've asked *you* a question, it's because I want to know what *you* think.
I don't want to know what Claude thinks.
If you don't know the answer, say that. "I don't know" is a perfectly good answer, albeit one which I might follow up with something like "Well why don't you know? This is literally your job to know", which is kind of what 90% of my emails at work for the past three weeks have been.
Out of interest, do you pay people extremely well to work with you?
"I don't know" is an extraordinarily powerful answer, but maybe it's getting ruined. I'm not sure it can be entirely ruined.
In a human interaction, when used confidently, it means 'I'm so authoritative I don't even have to front. I don't have to do the slightest thing to impress you. My foundation is so secure it doesn't need defending or justification at all". Also, 'that turns out to be outside the massive quantity of stuff I do know, how interesting'. Think Holmes, perhaps erroneously saying 'Now that I know that, I shall do my best to forget it' on the grounds that his work wasn't helped by the fact.
It could get ruined by an AI normalization of the idea that 'everybody should know literally everything, because asking an AI is the same as knowing and then you know the thing, having been told it'. Plenty of money is now marketing this position to the world, and so it's possible 'I don't know' will become ruined, a shameful response.
I feel like that may not happen in practice, though, because that pro-AI position is what used to be called 'pants-on-head R-slur' and I feel like the people who buy into this extreme position will make SUCH a disaster of things that it will tend to make that position, instead, the shameful response… and 'I don't know' will remain a valid signaling mechanism to convey 'I'm so authoritative I don't even have to front', all the more if it's more of a shocking response to get.
3 replies →
Most people who post LLM output don't do what you do though.
If someone asked you a question, it's most likely because they value you. Whether it's your skills, your domain knowledge, or something else, people ask you questions because they know you can provide some value. When you give them a Claude response, not only are you being rude, but you're devaluing yourself in the eyes of those around you. They learn that you can't be relied upon any more, and so they stop asking you questions.
Asking the LLM a question _is_ a skill, just like googling. The problem is not asking a LLM, but pasting the 300-chars response without any context.
3 replies →
If I asked a question to Claude, it's because I couldn't answer it myself, but also, because I used my skills and domain expertise to explain clearly the problem to Claude.
And if I paste back to you an answer from Claude, it means that I have read and evaluated what Claude wrote back, and I have extracted the part of it that most clearly contains the answer as well as the diagnosis- if I gor anything relevant.
Of course I could instead pretend I came up with the answer myself in 5 minutes thanks to my superior analytical skills, and write down two short lines that point to the solution without any explanation. Would you find it better?
4 replies →
You took all this time to reformulate the question, add the relevant implicit context, and do back and forth BUT you did not have time to reformulate the answer to something human readable that YOU own and can take responsibility for? This does not make sense to me. Nobody said not to consult an llm while formulating an answer. But if you give an answer, own it. Don't throw slop grenades.
…yes?
The amount of text isn’t equal and the benefit is primarily that I can push the LLM into a basin you don’t know the keywords for — something that takes a few minutes to type a couple sentences and me scan the results to make sure they’re roughly on track. And much, much less time than me trying to (re)write a treatise on a topic.
Dropping you an LLM reply is the equivalent of dropping you a link and telling you to learn about the topic — and appropriate in roughly the same circumstances.
Why would I reformulate what Claude already explained in a way that I find satisfactory? Only to give you the impression that I came up with it myself?
> if you give an answer, own it. Don't throw slop grenades.
If I send you some text copied and pasted from Claude, it means that I have read it and I think it's worth for you reading it. In this sense I'm already owning it. The problem is not the pasted answer, it's your assumption that a Claude-generated text must be "slop" that I haven't read.
1 reply →
I do not think people should stop asking questions. What is the point of teams / departments / orgs ?
[dead]
devils advocate; why didnt you ask claude then? stop wasting people's time when you could get your own answers.
Because I wanted to ask a human who’s job it is to know, instead of an AI that might make things up and waste all our time.
Why is it their job to know, but it's not your job to know?
7 replies →