I’m not convinced tbh, a lot of my colleagues messages are “x broken” or “I do y” … no context whatsoever, and then I have to coax their context for making these decisions.
Since people started using Claude now I get full context of everything… might be too much sure, but to be honest over-communicating seems better than under-communicating - sure it’s boring and tedious but that shifts the blockage to me.
Otherwise each of these coaxing sessions is something I have to keep in my head until resolved, which is a load in and off itself.
And even better - I can point Claude to that message and gives me a summary. You might say this is silly because we are paying the LLM tax, but knowing what to share, and then verifying against my situation are two different things.
Sometimes they have checked the wrong thing, sometimes they need guidance, sometimes I need to investigate before I can answer. A full LLM message with context is like “summary of their working context that I can resume on my end” so I don’t waste cycles asking or rechecking etc.
And it is still possible to be a learning experience for both of us since the initial message is the start, we can then talk to each other like humans, both much more in sync than before…
It is kinda ironic but I think it does push things in a better direction. Would I have loved it if the message was human, direct and to the point - obviously, but we are all busy, we got shit to do, and this is a useful “resume” mechanic
1. I've experienced both (ai paste slop & ai providing helpful context that would otherwise have been absent) but the former has been a much bigger problem than the purported benefits of the latter
2. before the advent of LLMs, I had fairly strong contrarian opinions on the latter (clueless users asking questions with no detail) that I still hold to this day. A good proxy for my opinions on this is that I think https://nohello.net/ is a crutch for undersocialised (or overworked/near-burnout OR undiagnosed autistic) engineers & should not be normalised. The origin of this sentiment is users entering a quiet IRC room & saying "hello", then leaving & not being around when someone eventually replies: in that context nohello is pragmatic, but taking it beyond that to modern high-frequency realtime chat is an antisocial overstep. Leading with the entire context of your query may have some practical benefits but they're honestly overstated in exchange for killing natural social tendencies towards human dialogue. Similarly, a non-technical user giving you zero info about a problem is an opportunity for conversation & mutual education. It's relationship-building & helps build a foundation for further interactions with this person. It was never a problem in the first place; "solving" it is just another milestone on the road toward de-socialising humanity.
You are right - and this has been the case for me before, but unfortunately human resource constraints are not something most orgs optimise for, including my current one.
I would love to spend more time and talk to the people with their issues so we can come to a common understanding… but neither have the capacity for this, and now things are getting worse I think.
Maybe someday agents are good enough that we just discuss direction and implementation details, and every “developer” becomes more of an executive in a company, free to chat about more than just work, focusing on building human relationships and leaving “the tech stuff” to their agent teams…
And old geezers like us who actually know how to write ourself out of a for loop would be the premium - wizards holding forgotten knowledge, incomprehensible to mere uninitiated mortals. I wonder what distinctive dress and hair styles will we develop :-D
If you're absolving the other person from any understanding of the issue and the alternative is "stuff broken" then yes what you say makes sense. But that's an incredibly low bar and one would have to say - that's not a colleague, that's an end user type relationship.
> Since people started using Claude now I get full context of everything…
And I think this is where one key argument on the page itself I think fails:
> The person on the other side has the same tools you do.
Yes, the person on the other side has Claude; but in many cases, their Claude doesn't have the same context as my Claude. Saying, "This thing happened in an area of the code I don't have any expertise in. Below is Claude's analysis, which I'm not qualified to assess" is I think a valid time to copy and paste LLM output. [1]
The other thing to say is that it can be a "LMGTFY" style response; i.e., "Why did you ask me that, when you could have gotten a perfectly good answer from ChatGPT?"
[1] In the particular case I'm thinking about, it was a bug in an ocaml component; I don't know ocaml or need to know it; handing Claude's analysis 'caveat emptor' to the maintianers who do know ocaml I think is more helpful than not.
> Below is Claude's analysis, which I'm not qualified to assess
I’m reading a bunch of game modding discords, and this exactly what happens, and what’s hated. Because Claude is also not qualified to assess it. Not by default. The modder’s Claude might be because of the extra context and steering, but the user’s produces usless slop, instead the actual log, and context what the user did is what’s helpful.
I’d say this (when you don’t understand it yourself) is pretty much the worst time to send AI output.
Quick edit: Yeah, I get it in situations like GP where the user is unhelpful, doesn’t want to learn, doesn’t care, and you also can’t tell them to fuck off. But just because the user gives Claude the context they should have given you in the first place, doesn’t make it good.
Hah, I guess we are not far away from the “Avogadro Corp” future - and that book was written just _before_ LLM, chatgpt and the attention is all you need paper :) That author had quite the insight
> sure it’s boring and tedious but that shifts the blockage to me
I think that's the key. Generally people would rather not receive slop because the receiver has no idea how much thought the sender invested in the communication.
In your example, the more context the better. I think there's an interesting nuance there.
The premise of the article is that everyone has access to the “same tools” but that’s glossing over the specialization inherent in a specific expert’s harness config (memories, skills, rules, repos, etc) that produce actually helpful responses (eg engineer’s view vs product manager’s view).
Heh. Just got done writing (by hand!) a Principles of AI Use document for my (ironically) AI enablement firm, the first of which is:
Write as yourself. You’re being paid for your expertise and insights. Communicate them directly to us. Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else, and worse, risks skipping that step for yourself. Even if you’re fundamentally using Claude to gather your thoughts or help you prepare a response, you need to be writing it yourself, in your own voice. Not having Claude ape your voice, or “make it sound less like AI”. You, directly. Doing this will further reinforce your own understanding of the state of things, the same way teaching someone is the best way to learn. As a guideline: for Slack and email comms, this should be near-universally written as you. For deliverables that are longer form and follow a template like proposals, roadmap/discovery work, etc., use of agents is expected but, see Principle #2. (Own the Output.)
Just want to say, "Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else" is something that I don't think I really internalized until you wrote it in that way. Which I think is evidence of what you are talking about.
But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is...
As such, isn't the question really about
1) whether it was communicated clearly/effectively
2) whether it should have been communicated to others at all (rather than something that you just dealt with yourself)
As for the writer being paid to use their expertise, learn more things etc: absolutely. But they could spend whatever time is required to think about the issue and then still have Ai write it at their direction. And, regardless of who wrote it, it could still be poor, irrelevant or misleading.
I'm not defending AI over-usage, but you folks evidently haven't thought about any of this nearly sufficiently enough... and the policy doesn't seem to be addressing what it ought to be: whether everyone is actually thinking, learning, discerning, deciding etc...
Ps I passed your comments to chatgpt and asked what it thought about it all, and it missed all of this. But then agreed once I brought it all up.
>But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is...
Not the same burden.
The value of the person is digesting and giving the materially important stuff in the answer (including their expertise and opinion). Not the AI diarrhea.
> But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is...
No. People earn a reputation over time. An AI does not. If you copy paste an AI response verbatim, i need to read everything in your message with a big grain of salt. Imagine doing this all the time as a reader? Exhausting and a waste of others time. If i want to read raw AI output i’ll ask the AI directly, thanks.
If you write something in your own words, or at least synthesize an AI response before sharing it with me, then i know i’m reading something that has at least passed the smell test.
It’s the difference between reviewing a pull request from an intern and a principal software engineer.
Nice. This is exactly how I mostly use AI. Doesn't matter if it's code or an e-mail. I have the AI do the tedious work of ... well, uh, "working", and then ideally learn from it and then transform it into my own work, as I would have done it without AI, but just being much faster now. It also establishes a natural threshold of time/work ratio for me that keeps me from using AI for every little thing. Also it feels like I am learning something new with every prompt/response.
Seconding this request... Our team is going through AI adoption without a clear direction and we're getting to the point we need to sit and come up with some guidelines. It would be awesome to have some human-made examples as inspiration
yes, its all ai writing. A bunch of the metaphors here dont make sense.
"read it all, yes even the bullets. especially the bullets"
if you read one thing from a large wall of text, its gonna be the bullet points, right? thats a big part of what they're for?
"It is the conversational equivalent of forwarding the email."
this is a terrible metaphor. it seems to rely on the idea you would be mad if you sent an email to someone and they forwarded to someone else? but thats useful, theyre literally sending it to someone who is presumably better placed to help you, or raising it to someone more important because they think what youre saying is important
Interesting to note the page also ends with "written by a human, on purpose"
Actually tbh AI outputs can be pretty good and I have been surprised to see a lack of copy / pastes of them in the format of:
Unique human prompt + AI output which is useful and does a good job expanding on the prompt topic and which when pasted saves people the time of having to go prompt AI on the same topic (fully reviewed by a person before sharing it, to ensure the content has relevance)
Unless I'm your research assistant, why would I be asking AI questions for you when you could do it yourself? If you need an answer I already have I will gladly share it with you. If you need my help to find an answer, then tell me that and I gladly will help.
The article says they could have asked an AI themselves, but that is missing that I may have an agentic context with exactly the right information to answer the question well, which they do not have. For example because I worked on something closely related. Such an agent output is very useful, and taking the time to rewrite that to a short personal answer is not always worth it.
Agree that it isn’t nice to just copy & paste, but the assumption that your AI generated answers will the same of someone’s else is wrong. Everyone using AI for a while built a different set of preferences, memories and context in general which influence the answers.
Would be interesting if someone said “I spent $2.50 using llm-council with the three top SOTA models and think this reply makes sense: {paste}”
(But better “not sure, though after spending a few bucks on model APIs, perhaps {human-written analysis} is in the right direction”)
Also like you, I did think on the angry version “they have the same AI you do” isn’t likely to be right in the consumer space since their AI is ChatGPT (free version?!) and yours is Claude or something.
You're giving too much credit to people. The folks who run an LLM council aren't the ones blindly copy pasting things.
There's responsible AI use (which I think you - and I, and many people here) do, and the "I don't care, I've outsourced my brain to OpenAI" type of people.
Even with the same model results are different. I work in a startup where they went bonks with AI ~1 year ago. Every employee got a Claude account that I don’t even know what tier is (we just have what feels like unlimited tokens), they implemented infamous AI usage leaderboards, people were getting “you’re not using AI enough” feedback and all of that…
Over time, it was impressive to see how much the output of the same prompt would differ between us, and it seems to be heavily influenced by the role of the user. I.e. answers coming from the agent of a Data Engineer, a Backend Engineer and a Domain Researcher would not be the same at all. Some of us use harnessing tools, some not, but regardless it we saw differences.
So, for example, when a backend engineer would send a question to a data engineer, there was *some* added value even when it was a lazy copy & paste.
Similar thoughts here. My gut tells me "if I wanted AI responses I would just ask AI", but in truth LLMs are not like Google Search. LLMs are more like a slot machine where you put your prompt in and out comes a combination of insights that are personalized to the inputs.
Disagree - most people aren't bothered to ask AI and have a conversation with it, they will go to you to get an answer so they have someone to blame when it's wrong.
Ask the ai to point out the egregious mistakes only.
Especially today, I will prefer your slightly funny English to something that feels like AI.
And personally I do Support for people around the world and I often ask people to write in their native language.
Then I can translate myself and if the default ai translation gets something wrong I can add necessary context to improve it - which I can't really do if the only thing I get is the bad ai translation.
How did people do it before AI? Spellchecks have existed forever. Also, people have the ability to improve, so if you offload all your work elsewhere, you will never improve.
det sker nogle gange! når man er ny til et sprog, det kan være forskrakkelig at skrive i et professionelt område. Men det er bare en del af dine udvikling - du kan sagtens skrive på hvilken sprog når som helst, og de kan bruge deres værktøj til at forstå det, du siger.
Der skal kom et tid hvor du skal tale på en sprog på dit kontor. Uden hjælp. Du har bruge for at øve dig, og folk er alligevel pænere end du tror.
(this being written in a foreign language I am learning is part of the point)
This does feel like responding to rudeness with more rudeness. My personal guess would be that people who do this (pasting AI responses) know what they are doing, not genuinely trying to help. It's the equivalent of sending LMGTFY.
I was recently listening to a podcast about mentalization theory therapy methods for BPD .. for a friend .. Mentalization being the activity of understanding whats going on in other peoples minds, or in ones own. And they idea of pseudomentalizing or "pretend mode" was put forward, where a person looks or sounds like they are .. engaged in mentalizing communication, but are actually checked out. Then I started to think about all the AI slop going around .. and the phenomenon, this meme site is responding to, or AI posting responses.. and I wondered, is this a symptom of.. a global society wide phycological functional deficit, where A) we're none of us communicating as if there is people on the other end.. and B) we're not thinking about our own minds, and taking time to understand things..
[edit] to extend the connections
The charm of a messaging board like this is bringing "join attention" to a topic, not just sharing separate monologs.
And also, always wondering if something was written by a human or statistically generated could be seen as Epistemic Hypervigilance, leading to stress, and loss of epistemic trust. This could reduces opportunities for taking on new information of feedback, and cause rigidity and isolation.
I can really only imagine the social context you would want to use this response in would be inappropriate / anti-social. If you had a well-meaning work colleague send you something AI, sending this back is kind of not a good look. If it's meant to be an education piece we got the memo...
Pasting a blurb of AI is sometimes a form of "Let Me Google That For You" LMGTFY. If someone asks me a dumb question, I will often resort to giving them an AI answer, that they could have easily looked up themselves.
... the page never argues that AI is bad or not useful broadly. Not sure if this getting commented again and again is because people are just that bad at understanding that context matters. Not everything has to be evil / slop or pure / virtuous, regardless of context.
Since people started using Claude now I get full context of everything… might be too much sure, but to be honest over-communicating seems better than under-communicating - sure it’s boring and tedious but that shifts the blockage to me.
Otherwise each of these coaxing sessions is something I have to keep in my head until resolved, which is a load in and off itself.
And even better - I can point Claude to that message and gives me a summary. You might say this is silly because we are paying the LLM tax, but knowing what to share, and then verifying against my situation are two different things.
Sometimes they have checked the wrong thing, sometimes they need guidance, sometimes I need to investigate before I can answer. A full LLM message with context is like “summary of their working context that I can resume on my end” so I don’t waste cycles asking or rechecking etc.
And it is still possible to be a learning experience for both of us since the initial message is the start, we can then talk to each other like humans, both much more in sync than before…
It is kinda ironic but I think it does push things in a better direction. Would I have loved it if the message was human, direct and to the point - obviously, but we are all busy, we got shit to do, and this is a useful “resume” mechanic
1. I've experienced both (ai paste slop & ai providing helpful context that would otherwise have been absent) but the former has been a much bigger problem than the purported benefits of the latter
2. before the advent of LLMs, I had fairly strong contrarian opinions on the latter (clueless users asking questions with no detail) that I still hold to this day. A good proxy for my opinions on this is that I think https://nohello.net/ is a crutch for undersocialised (or overworked/near-burnout OR undiagnosed autistic) engineers & should not be normalised. The origin of this sentiment is users entering a quiet IRC room & saying "hello", then leaving & not being around when someone eventually replies: in that context nohello is pragmatic, but taking it beyond that to modern high-frequency realtime chat is an antisocial overstep. Leading with the entire context of your query may have some practical benefits but they're honestly overstated in exchange for killing natural social tendencies towards human dialogue. Similarly, a non-technical user giving you zero info about a problem is an opportunity for conversation & mutual education. It's relationship-building & helps build a foundation for further interactions with this person. It was never a problem in the first place; "solving" it is just another milestone on the road toward de-socialising humanity.
I would love to spend more time and talk to the people with their issues so we can come to a common understanding… but neither have the capacity for this, and now things are getting worse I think.
Maybe someday agents are good enough that we just discuss direction and implementation details, and every “developer” becomes more of an executive in a company, free to chat about more than just work, focusing on building human relationships and leaving “the tech stuff” to their agent teams…
And old geezers like us who actually know how to write ourself out of a for loop would be the premium - wizards holding forgotten knowledge, incomprehensible to mere uninitiated mortals. I wonder what distinctive dress and hair styles will we develop :-D
And I think this is where one key argument on the page itself I think fails:
> The person on the other side has the same tools you do.
Yes, the person on the other side has Claude; but in many cases, their Claude doesn't have the same context as my Claude. Saying, "This thing happened in an area of the code I don't have any expertise in. Below is Claude's analysis, which I'm not qualified to assess" is I think a valid time to copy and paste LLM output. [1]
The other thing to say is that it can be a "LMGTFY" style response; i.e., "Why did you ask me that, when you could have gotten a perfectly good answer from ChatGPT?"
[1] In the particular case I'm thinking about, it was a bug in an ocaml component; I don't know ocaml or need to know it; handing Claude's analysis 'caveat emptor' to the maintianers who do know ocaml I think is more helpful than not.
I’m reading a bunch of game modding discords, and this exactly what happens, and what’s hated. Because Claude is also not qualified to assess it. Not by default. The modder’s Claude might be because of the extra context and steering, but the user’s produces usless slop, instead the actual log, and context what the user did is what’s helpful.
I’d say this (when you don’t understand it yourself) is pretty much the worst time to send AI output.
Quick edit: Yeah, I get it in situations like GP where the user is unhelpful, doesn’t want to learn, doesn’t care, and you also can’t tell them to fuck off. But just because the user gives Claude the context they should have given you in the first place, doesn’t make it good.
Context is everything.
Btw, Claude code sent my first accidental AI email yesterday. I guess permissions changed? Before it could only draft. Uhoh, waiting for the onslop.
I think that's the key. Generally people would rather not receive slop because the receiver has no idea how much thought the sender invested in the communication.
In your example, the more context the better. I think there's an interesting nuance there.
Really? How does Claude gather that context?
Write as yourself. You’re being paid for your expertise and insights. Communicate them directly to us. Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else, and worse, risks skipping that step for yourself. Even if you’re fundamentally using Claude to gather your thoughts or help you prepare a response, you need to be writing it yourself, in your own voice. Not having Claude ape your voice, or “make it sound less like AI”. You, directly. Doing this will further reinforce your own understanding of the state of things, the same way teaching someone is the best way to learn. As a guideline: for Slack and email comms, this should be near-universally written as you. For deliverables that are longer form and follow a template like proposals, roadmap/discovery work, etc., use of agents is expected but, see Principle #2. (Own the Output.)
As such, isn't the question really about
1) whether it was communicated clearly/effectively 2) whether it should have been communicated to others at all (rather than something that you just dealt with yourself)
As for the writer being paid to use their expertise, learn more things etc: absolutely. But they could spend whatever time is required to think about the issue and then still have Ai write it at their direction. And, regardless of who wrote it, it could still be poor, irrelevant or misleading.
I'm not defending AI over-usage, but you folks evidently haven't thought about any of this nearly sufficiently enough... and the policy doesn't seem to be addressing what it ought to be: whether everyone is actually thinking, learning, discerning, deciding etc...
Ps I passed your comments to chatgpt and asked what it thought about it all, and it missed all of this. But then agreed once I brought it all up.
Not the same burden.
The value of the person is digesting and giving the materially important stuff in the answer (including their expertise and opinion). Not the AI diarrhea.
And because most ai's are not concise we blame ai's for faults we don't like in humans either: being longwinding
No. People earn a reputation over time. An AI does not. If you copy paste an AI response verbatim, i need to read everything in your message with a big grain of salt. Imagine doing this all the time as a reader? Exhausting and a waste of others time. If i want to read raw AI output i’ll ask the AI directly, thanks.
If you write something in your own words, or at least synthesize an AI response before sharing it with me, then i know i’m reading something that has at least passed the smell test.
It’s the difference between reviewing a pull request from an intern and a principal software engineer.
For purposes of training the AI (enablement firm)?
"read it all, yes even the bullets. especially the bullets"
if you read one thing from a large wall of text, its gonna be the bullet points, right? thats a big part of what they're for?
"It is the conversational equivalent of forwarding the email."
this is a terrible metaphor. it seems to rely on the idea you would be mad if you sent an email to someone and they forwarded to someone else? but thats useful, theyre literally sending it to someone who is presumably better placed to help you, or raising it to someone more important because they think what youre saying is important
Interesting to note the page also ends with "written by a human, on purpose"
Hypocritical to say the least, given the message.
DS9 Garak "especially the lies" reference maybe?
Unique human prompt + AI output which is useful and does a good job expanding on the prompt topic and which when pasted saves people the time of having to go prompt AI on the same topic (fully reviewed by a person before sharing it, to ensure the content has relevance)
Would be interesting if someone said “I spent $2.50 using llm-council with the three top SOTA models and think this reply makes sense: {paste}”
(But better “not sure, though after spending a few bucks on model APIs, perhaps {human-written analysis} is in the right direction”)
Also like you, I did think on the angry version “they have the same AI you do” isn’t likely to be right in the consumer space since their AI is ChatGPT (free version?!) and yours is Claude or something.
There's responsible AI use (which I think you - and I, and many people here) do, and the "I don't care, I've outsourced my brain to OpenAI" type of people.
Over time, it was impressive to see how much the output of the same prompt would differ between us, and it seems to be heavily influenced by the role of the user. I.e. answers coming from the agent of a Data Engineer, a Backend Engineer and a Domain Researcher would not be the same at all. Some of us use harnessing tools, some not, but regardless it we saw differences.
So, for example, when a backend engineer would send a question to a data engineer, there was *some* added value even when it was a lazy copy & paste.
Dont be a meat proxy! simple
idk about that
Especially today, I will prefer your slightly funny English to something that feels like AI.
And personally I do Support for people around the world and I often ask people to write in their native language.
Then I can translate myself and if the default ai translation gets something wrong I can add necessary context to improve it - which I can't really do if the only thing I get is the bad ai translation.
Der skal kom et tid hvor du skal tale på en sprog på dit kontor. Uden hjælp. Du har bruge for at øve dig, og folk er alligevel pænere end du tror.
(this being written in a foreign language I am learning is part of the point)
And in my experience, us Brits, and USA folk, we tend to be quite understanding of broken English
[edit] to extend the connections The charm of a messaging board like this is bringing "join attention" to a topic, not just sharing separate monologs.
And also, always wondering if something was written by a human or statistically generated could be seen as Epistemic Hypervigilance, leading to stress, and loss of epistemic trust. This could reduces opportunities for taking on new information of feedback, and cause rigidity and isolation.
Almost certainly written by AI
It also ends with "written by a human, on purpose"
Disagreed. Bad writing from a chatbot beats worse writing from its user every time - regardless of length.
One reason we have so much bot slop is the attractiveness of bad-writing bots to the many people whose writing is even worse.