I feel the opposite. Interacting with humans, I definitely pay a social tax - I have to negotiate the feelings of the people involved. With LLMs there is no social tax, I can be as blunt as I like. But there are other taxes to pay with LLMs; they don't learn, they BS relentlessly, there is less fun and camaraderie.
Agreed, I haven't yet internalized them not remembering feedback from a couple days ago, which humans typically would. The process of remembering to update the context isn't entirely natural to me to this day.
There are ways to put a superficial personality on the LLM which can break up some of the mannerisms that get old. Give it a new accent, allow it to keep a memory of things that nudge it into one personality or another, make it behave differently on weekdays vs weekends. Its still superficial, but if you work with it any reasonable amount of time, I think having the ability to change some of the characteristics can make it more..."fun" I guess.
The first-order social consequences aren't truly there but i think the format in which you interface with the tool can trick you into behaving as if they were, which is taxing it itself.
One giant issue with say Claude Code is the speed, say you ask it to do something and you want to review the response, then I want to ask questions , I want to chat about the changes , but the slow speed of responses it just breaks the flow.
Unfortunetly now vibe coding is demanded to be used, to produce 10x more code or else some other developer will take your place.
> With LLMs there is no social tax, I can be as blunt as I like. But there are other taxes to pay with LLMs; they don't learn, they BS relentlessly, there is less fun and camaraderie.
My experience is the opposite. They bullshit a lot, derail the conversation/coding session, get defensive, and gaslight. It is not “social tax”, but they mirror sometimes the worst human behaviour.
There was no system prompt, I asked claude thru the CLI to implement something. It did not agree that I need that feature and offered something else. I started debating with it, why did I need it, and it started derailing the conversation, and explaining why I was wrong. And this is not the first and only case.
Man, I don't know. LLMs (with the good and bad) feel like the first time a tool has genuinely been an extension of me. And quite the opposite. I'm a quite introverted person. Spending time in a meeting or talking with other humans I find quite exhausting. I don't really get that at all with talking with LLMs.
That is interesting, but also I do wonder, what in life doesn't have a variable reward? You could say that any routine activities like maybe cleaning, etc, is all the same, but most non routine things in life seem like they would have variable reward?
E.g. talking to people definitely yields in very variable rewards, if you do non routine work, there's constant variable rewards, etc.
Happiness is derived from a balance of a meaningful life, and pleasant experiences we may choose.
Also, one may have no pain or concerns, and still existentially despair over a meaningless life built on intelligence campaigns exploiting millions of people.
Only psychopaths find short lived joy in harming others, and only make up around 1% of general populations. Have a wonderful day =3
I think happiness is very complicated. Would you think there's a perfect formula to follow that will yield you the "happiness"? And how does this relate to different brain wirings of different people? Would you say per brain wiring there's a different specific ideal formula one would have to follow? E.g. there's some base things like getting proper sleep, exercise, diet, but then there's some more specific things for that brain wiring?
Indeed, indifferentism must recognize both positions with similar regard. Then address ambiguity with logic and wisdom. For example, Bayes' theorem projects a reasonable expectation given available data. =3
Wouldn't Bayes and logic lead you to immediately believe that some claims, ideas, and beliefs may be more logically sound or matter more than others, steering you away from indifferentism, but the reason why I thought you might be stuck is because you were indifferent to moving away from it.
On occasion, the wisdom of rational self-interest resolves a precedence of each area of metaphysical, imaginary, and scientific subject matter. Often, conflating each area of study leads to nonsensical contradictions.
For example, Bayes theorem does not guarantee the best outcome, but does offer the best unbiased choice at that moment. =3
And on the other hand talking to people, but I guess most extroverts in a way are "addicted" to talking to people as well, so it's a fair comparison. From my personal experience talking to people can end in wildly differing and unexpected results. In fact, I think rewards from LLMs are way more static as opposed to rewards from talking to people.
How does that answer the question or relate to variable reward and addiction mechanisms? This just seems another unrelated negative sentiment argument towards LLMs?
LLM are great for context search, but were never real "AI" in any sense except in media/PR sensationalism. Scientific hubris and unethical use of LLM is interesting as exposes the primitive impulsive nature of many. Best of luck =3
This is my read on how people react to LLM responses, especially about personal/social issues too - that many are fooled by the LLM. On occasion, someone has shown me LLM output, believing it to be a profound and illuminating response to their prompt.
Even being able to see the original prompt, the response just seemed kind of empty (blocking out the advertisingesque magnificent language leaves little substance behind), but somehow they were convinced that it was a coherent answer to their prompt and even peeved with me for not sharing their astonishment at how well it addressed their concerns.
It's a kind of symbiosis between the LLM and the prompter, where the prompter's own cognitive biases fill in what the LLM can not. Since I don't share their same biases, the response is nearly meaningless.
Very uncomfortable experience having to pretend to agree with how profound it is to not make their prompter upset.
Would a person with such biases, seeking validation from an LLM having been anything else without an LLM?
If everyone in real life are trying to avoid upsetting such person, the humans in their real life would offer the same experience, no? Or alternatively said person would go into conflict with anyone pointing out issues in their views, ending up only surrounding themselves with people who will agree with them in order to not upset them.
I guess my concern is why did you agree with them instead of using the opportunity to showcase another point of view?
They used to call it schizophrenia, and put people in padded rooms for such delusions. Now, "AI" ego manipulation is a mental dildo for people that should know better philosophically. However, we shouldn't kink shame, so give this post a thumbs-down if you agree. lol =3
> When you use an LLM, you don’t get the tool magic: (almost) nobody will claim that Claude or Cursor feel like an extension of their body - they are not consistent or fast enough to trick the brain like a keyboard or a car can.
These all seem the same to me? None of them are an "extension of my body"; they're tools I use.
> With LLMs, you mostly just get more of the same: more code, more tests, more excuses.
You get more specs, more plans, more code, more tests. If you're getting excuses, something's wrong.
> Is it worth the social brainwork?
There isn't any social brainwork. I'm using natural language to build things, not engaging in social discourse.
> LLMs ask us to talk to them, but rarely reward that effort in kind.
Nor should they! They're not people, but they're designed to reach goals. If you set a goal (explicitly or implicitly) that you want a social conversation, they'll try to satisfy (and do poorly).
With a car, when I turn the steering wheel a quarter turn, the car immediately starts to turn at a known rate (I can't quantify it, but my muscle memory knows it). With an LLM, if I ask it to "turn the car", my "steering wheel" is natural language. I have to spend effort to figure out how to convey the rate at which I want the car to turn, and because the LLM has a noticeable delay, I also need to preemptively include some language anticipating how the expected rate should feel and what to do if it doesn't seem to feel right. And then the next time I say the same thing, the LLM decides to steer the car a bit differently anyways.
I feel on the other side of this. Just yesterday I was reviewing some code output from Claude and I realized a change that I had asked for in a previous review step wasn't what I wanted. I had a moment of social anxiety, like I didn't want to bother a coworker with my indecision. But I have to remember, the LLM doesn't care. It doesn't have an ego. It doesn't get annoyed at being asked to redo work.
I still say "please" and "thank you" frequently, but I'm starting to embrace the fact that the LLM doesn't care about grunt work, doesn't care about rework, doesn't care about nitpicking, doesn't have a preference in general. It needs very little more than for me to be completely clear in my instructions.
Sorry to hear about the moment of social anxiety (I assume it happens with humans occasionally). But can't wonder don't you have also moments of joy because of job well done, an appreciative colleague or something similar?
I don't like either the "negative" part, but I find it necessary to have both negatives and positives in life to create bonds, meaning and more simply, not to get bored. I would be worried that if I just talk with a machine (no feelings involved) I will get depressed and demotivated.
I'm pointing out how I noticed a particular emotional response when working with LLMs.
I've been an engineering manager in the past and I have tried my best to keep the needs of my team in mind when I am delegating work. I try to consider the person, their goals, motivations, preferences, frustrations. I consider before interrupting them if the minor issue I am bringing up is worth the distraction it might cause them, since switching tasks is a mental load.
But with LLMs, almost none of that matters. They don't have goals, motivations or preferences in the same way people do. I can interrupt it all day and it won't get frustrated or lose motivation.
I think anxiety is a harsher word than I mean, but it is close to the feeling I have when I'm about to deliver bad news to someone. When I'm about to say "you know all that work I asked you to do, I need you to throw it away and restart". And I model in my mind the frustration and demotivation this can cause a person. And then I feel anxious about causing them this frustration.
I have to train myself out of that when instructing LLMs. It doesn't mean I have to avoid moments of joy or appreciation. It means I have to understand LLMs have different needs than people, and I have to work towards those needs.
It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear! And it absolutely will not stop, ever, until … the project is complete?
Most of my workflows have slowly moved away from the chat interface with llms. Instead they look more like traditional Unix pipelines that just happen to call Unix tools that interact with llms.
This allows me to make more repeatable processes, not be tied down to vendor implementations of workflows and mix and match models for cost and efficacy.
There is nothing that ties you to talking with the text generator black box, and for most of my use cases it’s a negative.
Would you provide some examples? It sounds like you're feeding results from one LLM into another LLM invocation. Kind of like the loop thing everyone's talking about, but more like a workflow. Or programming with LLMs
In the simplest incarnation I’m just using the built into the agent cli parm’s that trigger non-tui behavior (for instance calling codex exec instead of just codex).
I’ve been experiencing similar feelings. Working with LLMs often takes almost as much mental energy as working with people, but the payoff does not always scale in the same way.
I think we are still on the early days of LLMs. Right now, using them productively requires deliberate thought and an acute knowledge of their limitations. As the author says, it’s easy to get angry at a model, or to foolishly let it nudge you towards more code and more tests — even when that is suboptimal.
To a certain extent, models keep getting better and better at discerning our intentions and providing value. Yet I am not sure whether we will reach a point where using them successfully no longer causes the kind of fatigue that it does today.
I talk almost exclusively with AI these days. There's no one around me who knows programming, I get tired of reading code for projects I'm not interested in, and the projects I am interested in are too difficult, so I just talk with AI, organize my thoughts, and read books.
There's no one around me who does programming. There are hardly any programmers in my town.
The upside is that most programming-related tasks in my town end up going through me. The downside is that there's not much work to begin with, and I can't talk about the things I'm actually interested in.
I'd like to stay in touch with friends who are interested in programming or academia, but since I didn't go to a good university, it seems like I haven't had much of a connection with them
You're right. Because conversations with AI don't force me to deal with boring code, don't scratch my ego, and they flatter me.
On top of that, talking to smart developers in real life is exhausting. Putting aside whether they share my interests, there are too many arrogant people. There's also the embarrassment of being asked, 'You don't even know this?' when they have knowledge I lack. The problem is that while that embarrassment could help me grow, it also leaves scars.
And on top of that, the Korean internet is more toxic than you'd think. Most of the male-dominated communities in my interests are filled with misogyny and derogatory remarks. (You can think of Korean internet communities as having 4chan as their baseline.)
So maybe I just chose AI to stay in a greenhouse.
So I'm not sure. Whether I lack the courage to leave the greenhouse, or whether I'm just genuinely exhausted.
The Korean programming communities just spam programming memes, and most of those are factually wrong. I don't want to bother fighting over them.
Even though I've successfully delivered to 40 different companies, my opinions are always seen as 'unsubstantiated personal views,' while those who come from prestigious companies have their opinions treated as 'insights born from experience.' So it feels like there's no one I can have an equal conversation with.
Ultimately, most relationships seem to require the other person to have something to give me, but I don't have anything to give them in return.
I get along quite well with people in real life. But I have no conversations with them about the things I actually care about. And that's lonely. They all say I'm kind and diligent, but I don't have anyone I can truly open up to.
I didn't expect LLMs to change their output so much depending on how I talk to them. I've done my own controlled tests and verified tone impacts quality even when content is held constant. This effect is called "Linguistic Convergence" or "Conversational Mirroring" and it's been studied extensively. The effect is minimal in coding contexts but becomes much more pronounced in collaborative contexts like creative brainstorming and concept development.
I noticed this because I use LLMs quite often as a note taker, research assistant and reference collector when I'm doing ideation, domain mapping and knowledge acquisition. In those kinds of sessions, if I issue purely directive instructions, the LLM's output quality will begin to drop quite quickly. And if adopt a tone of conversational engagement, the LLMs output quality remains high. As the article states, this can be a burdensome distraction which creates additional cognitive load. It's basically a non-economic cost to using LLMs in these contexts.
Reading research on this, it's fundamental to the nature of LLMs and can't simply be prompted away or easily fixed in fine-tuning. It's an artifact of attention dilution and contextual satiation. If I include semantic richness, structural variety, domain-specific terminology and explicit reasoning steps, it provides higher-entropy tokens to the model's attention mechanism which shifts the weight calculations toward richer areas of the model's latent space. By ingesting a composite of human's collective linguistic structures, it seems like models inherited some of our quirks and sensitivities too.
Could you imagine how destroyed people like this would be if they faced actual adversity in life? Like if they had to live through the Spanish Flu, or the Great Depression, or hell even lose a loved one to something as now trivial to treat as tetanus?
Oh, the horror, having to type to a system that will do your job for you while you sit in an air conditioned office in a comfortable chair listening to a podcast while you work.
> When you use an LLM, [...] you get to pay the social tax: you converse and negotiate and convince and sometimes even get angry1 at the so-called tool.
Might be a subjective opinion, but this is how writing code always felt to me, even pre-LLMs. An ongoing inner conversation where I try to convince the text on the screen to match the text in my head. It never really felt like tool use in the sense of manual labor.
The way I have worked so far is to look for ways I can influence the model's "thinking" and then add that to my main AGENTS.md. I try to steer it towards a thought process that mirrors or exceeds my own. I find it a fun challenge. I think this stuff becomes less necessary in a year or so as these sorts of tweaks become part of the shipped product from the model makers.
I think it depends. If talking to a tool solves a bigger exhaustion (a bug that has been bugging you for a while) then the exhaustion from talking to the tool becomes normalized. I guess this is true for any tool really. Using a hammer causes a bit of strain in your muscles but if it solves a bigger issue of nailing something in the wall then that strain in the muscle is fine. It will take some time to develop that muscle working with AI too.
As an aside, it's nice to see that Lobsters has remained a quiet success. As much as I love HN and the work Dan's done to keep it how it is, I welcome to variety. There are vanishingly few places for polite and earnest discussion online these days.
I quickly lost any interest in that site once I realized its dependence on social verification. That "bar" turns the site into an elitist social club. No thanks.
I feel this. These tools are viscerally unpleasant. Meetings used to be the thing I didn't look forward to but chatting with an AI is the new low point. Reading AI generated text is the written word equivalent of nails on a chalkboard.
>> Is it worth the social brainwork? IDK, for some tasks maybe - there are things a single person can do now that would have been impossible a year ago. But for all tasks? And wouldn’t that social brainwork do more good if it was directed at the real people you are working with?
It's the opposite. Many people find it exhausting to interact with humans, and do so only because they are required to.
Humans often don't understand what you are saying or asking, and they may not know exactly what steps they need to take to find the answer. They get tired. They might get their pride hurt. They might get angry or frustrated. They might judge you because your question is silly or just wrong.
LLMs, for all their faults, have none of these issues. I'm not saying I'd rather talk to LLMs all day every day, but when trying to get shit done, they really can be the superior coworker, especially if you're an introvert and suffer from social-battery-drainage issues.
Fooling yourself into thinking the LLMs are "understanding" what you are saying or asking is a trap. The output you get may be useful, but it is not due to any sort of understanding.
The elements of human work you mentioned are why it can be both rewarding and painful to interact with humans, but at the end of the day it is important to keep trying to do that work/keep trying to interact with each other. I don't know if we want to go back to cubicles where we just talk to robots all day. Of course some work environments are just awful, and there is not much remedy for that.
It’s going to sound delusional, but my own understanding of my understanding is that there’s probably no such thing, only a feeling of familiarity and some heuristics.
All I do is invoke “pictures” of semantic blobs in my mind, and my unconscious parts pull up more “pictures” in connection to that. Which I verbalize in turn, apply logic checks, and thus get this feeling that I’m meaningfully thinking about something, a sensation of comprehension. I could be even blatantly wrong about something, misunderstanding doesn’t feel any different in the moment.
And reasoning models (try to) capture that. The underlying processes are certainly different, it’s a Chinese Room alright - but if my own understanding is an illusion, the fact that machine model of it isn’t how I do it myself (but only a statistical approximation) doesn’t seem to matter - at least not for inputs and outputs between two black boxes (me or model).
I just don’t believe in p-zombies any more than I believe in Santa, I guess.
>> Fooling yourself into thinking the LLMs are "understanding" what you are saying or asking is a trap. The output you get may be useful, but it is not due to any sort of understanding.
I hear this a lot but I think it's a matter of semantics and ultimately not very useful. I don't care whether the LLM understands me the way a human would. I use the LLM to get useful output. I want it to do something and it does that thing.
I feel the opposite. Interacting with humans, I definitely pay a social tax - I have to negotiate the feelings of the people involved. With LLMs there is no social tax, I can be as blunt as I like. But there are other taxes to pay with LLMs; they don't learn, they BS relentlessly, there is less fun and camaraderie.
But "social tax"? No, there is not a social tax.
Agreed, I haven't yet internalized them not remembering feedback from a couple days ago, which humans typically would. The process of remembering to update the context isn't entirely natural to me to this day.
There are ways to put a superficial personality on the LLM which can break up some of the mannerisms that get old. Give it a new accent, allow it to keep a memory of things that nudge it into one personality or another, make it behave differently on weekdays vs weekends. Its still superficial, but if you work with it any reasonable amount of time, I think having the ability to change some of the characteristics can make it more..."fun" I guess.
I agree. I’m a social person but I also like my purely functional conversations with LLMs, and I have a lot of them going at once!
Doesn’t replace conversations with other people but at the same time I feel no tax talking to the LLMs.
The first-order social consequences aren't truly there but i think the format in which you interface with the tool can trick you into behaving as if they were, which is taxing it itself.
One giant issue with say Claude Code is the speed, say you ask it to do something and you want to review the response, then I want to ask questions , I want to chat about the changes , but the slow speed of responses it just breaks the flow.
Unfortunetly now vibe coding is demanded to be used, to produce 10x more code or else some other developer will take your place.
> With LLMs there is no social tax, I can be as blunt as I like. But there are other taxes to pay with LLMs; they don't learn, they BS relentlessly, there is less fun and camaraderie.
My experience is the opposite. They bullshit a lot, derail the conversation/coding session, get defensive, and gaslight. It is not “social tax”, but they mirror sometimes the worst human behaviour.
They only respond to the inputs. Try adding some communication preferences to your system prompt.
There was no system prompt, I asked claude thru the CLI to implement something. It did not agree that I need that feature and offered something else. I started debating with it, why did I need it, and it started derailing the conversation, and explaining why I was wrong. And this is not the first and only case.
Man, I don't know. LLMs (with the good and bad) feel like the first time a tool has genuinely been an extension of me. And quite the opposite. I'm a quite introverted person. Spending time in a meeting or talking with other humans I find quite exhausting. I don't really get that at all with talking with LLMs.
It’s because large language models give a variable reward and it’s essentially an addiction that many people have not even recognized.
That is interesting, but also I do wonder, what in life doesn't have a variable reward? You could say that any routine activities like maybe cleaning, etc, is all the same, but most non routine things in life seem like they would have variable reward?
E.g. talking to people definitely yields in very variable rewards, if you do non routine work, there's constant variable rewards, etc.
Happiness is derived from a balance of a meaningful life, and pleasant experiences we may choose.
Also, one may have no pain or concerns, and still existentially despair over a meaningless life built on intelligence campaigns exploiting millions of people.
Only psychopaths find short lived joy in harming others, and only make up around 1% of general populations. Have a wonderful day =3
I think happiness is very complicated. Would you think there's a perfect formula to follow that will yield you the "happiness"? And how does this relate to different brain wirings of different people? Would you say per brain wiring there's a different specific ideal formula one would have to follow? E.g. there's some base things like getting proper sleep, exercise, diet, but then there's some more specific things for that brain wiring?
>Would you think there's a perfect formula
The advice comes from coursework most medical students must do in year 2. It is backed by research data on integrated healthcare programs.
Simply avoiding misery is not the whole equation. =3
https://www.youtube.com/watch?v=LO1mTELoj6o
Would you say you are happy as in you have solved happiness for yourself?
On occasion, as I often chose wisdom, logic, and indifferentism. =3
Do you mean those three things are what let you solve it, or that they're what kept you from solving it more often?
Philosophical indifferentism is part of it for sure. lol =3
And what is the reason for indifferentism as opposed to atheistic agnosticism?
Indeed, indifferentism must recognize both distinctive positions with similar regard. =3
I see, that totally makes sense. You got stuck there and there's just no way out.
Indeed, indifferentism must recognize both positions with similar regard. Then address ambiguity with logic and wisdom. For example, Bayes' theorem projects a reasonable expectation given available data. =3
https://www.youtube.com/watch?v=4q1dgn_C0AU
Wouldn't Bayes and logic lead you to immediately believe that some claims, ideas, and beliefs may be more logically sound or matter more than others, steering you away from indifferentism, but the reason why I thought you might be stuck is because you were indifferent to moving away from it.
On occasion, the wisdom of rational self-interest resolves a precedence of each area of metaphysical, imaginary, and scientific subject matter. Often, conflating each area of study leads to nonsensical contradictions.
For example, Bayes theorem does not guarantee the best outcome, but does offer the best unbiased choice at that moment. =3
https://en.wikipedia.org/wiki/Stoicism
[dead]
Yep, LLM chat is classic skinner box design. It’s effectively gambling.
> it’s essentially an addiction that many people have not even recognized
A little patronizing, no? Maybe some people actually like it, even if you or I don't.
It's a slot machine that plays you.
In what sense is this true that isn't true of a simple web search?
And on the other hand talking to people, but I guess most extroverts in a way are "addicted" to talking to people as well, so it's a fair comparison. From my personal experience talking to people can end in wildly differing and unexpected results. In fact, I think rewards from LLMs are way more static as opposed to rewards from talking to people.
A web search doesn't use isomorphic plagiarism of naive users work to sell to other users.
LLM are actually good at context search, but are mostly not being used as intended. The LLM hype bubble has to end sooner or later. =3
How does that answer the question or relate to variable reward and addiction mechanisms? This just seems another unrelated negative sentiment argument towards LLMs?
Not really, people are simply fooled by pernicious sycophantic idealism, and their own cognitive biases.
https://en.wikipedia.org/wiki/Mechanical_Turk
LLM are great for context search, but were never real "AI" in any sense except in media/PR sensationalism. Scientific hubris and unethical use of LLM is interesting as exposes the primitive impulsive nature of many. Best of luck =3
This is my read on how people react to LLM responses, especially about personal/social issues too - that many are fooled by the LLM. On occasion, someone has shown me LLM output, believing it to be a profound and illuminating response to their prompt.
Even being able to see the original prompt, the response just seemed kind of empty (blocking out the advertisingesque magnificent language leaves little substance behind), but somehow they were convinced that it was a coherent answer to their prompt and even peeved with me for not sharing their astonishment at how well it addressed their concerns.
It's a kind of symbiosis between the LLM and the prompter, where the prompter's own cognitive biases fill in what the LLM can not. Since I don't share their same biases, the response is nearly meaningless.
Very uncomfortable experience having to pretend to agree with how profound it is to not make their prompter upset.
Why did you pretend to agree with them?
Would a person with such biases, seeking validation from an LLM having been anything else without an LLM?
If everyone in real life are trying to avoid upsetting such person, the humans in their real life would offer the same experience, no? Or alternatively said person would go into conflict with anyone pointing out issues in their views, ending up only surrounding themselves with people who will agree with them in order to not upset them.
I guess my concern is why did you agree with them instead of using the opportunity to showcase another point of view?
They used to call it schizophrenia, and put people in padded rooms for such delusions. Now, "AI" ego manipulation is a mental dildo for people that should know better philosophically. However, we shouldn't kink shame, so give this post a thumbs-down if you agree. lol =3
> When you use an LLM, you don’t get the tool magic: (almost) nobody will claim that Claude or Cursor feel like an extension of their body - they are not consistent or fast enough to trick the brain like a keyboard or a car can.
These all seem the same to me? None of them are an "extension of my body"; they're tools I use.
> With LLMs, you mostly just get more of the same: more code, more tests, more excuses.
You get more specs, more plans, more code, more tests. If you're getting excuses, something's wrong.
> Is it worth the social brainwork?
There isn't any social brainwork. I'm using natural language to build things, not engaging in social discourse.
> LLMs ask us to talk to them, but rarely reward that effort in kind.
Nor should they! They're not people, but they're designed to reach goals. If you set a goal (explicitly or implicitly) that you want a social conversation, they'll try to satisfy (and do poorly).
With a car, when I turn the steering wheel a quarter turn, the car immediately starts to turn at a known rate (I can't quantify it, but my muscle memory knows it). With an LLM, if I ask it to "turn the car", my "steering wheel" is natural language. I have to spend effort to figure out how to convey the rate at which I want the car to turn, and because the LLM has a noticeable delay, I also need to preemptively include some language anticipating how the expected rate should feel and what to do if it doesn't seem to feel right. And then the next time I say the same thing, the LLM decides to steer the car a bit differently anyways.
I feel on the other side of this. Just yesterday I was reviewing some code output from Claude and I realized a change that I had asked for in a previous review step wasn't what I wanted. I had a moment of social anxiety, like I didn't want to bother a coworker with my indecision. But I have to remember, the LLM doesn't care. It doesn't have an ego. It doesn't get annoyed at being asked to redo work.
I still say "please" and "thank you" frequently, but I'm starting to embrace the fact that the LLM doesn't care about grunt work, doesn't care about rework, doesn't care about nitpicking, doesn't have a preference in general. It needs very little more than for me to be completely clear in my instructions.
Sorry to hear about the moment of social anxiety (I assume it happens with humans occasionally). But can't wonder don't you have also moments of joy because of job well done, an appreciative colleague or something similar?
I don't like either the "negative" part, but I find it necessary to have both negatives and positives in life to create bonds, meaning and more simply, not to get bored. I would be worried that if I just talk with a machine (no feelings involved) I will get depressed and demotivated.
I'm pointing out how I noticed a particular emotional response when working with LLMs.
I've been an engineering manager in the past and I have tried my best to keep the needs of my team in mind when I am delegating work. I try to consider the person, their goals, motivations, preferences, frustrations. I consider before interrupting them if the minor issue I am bringing up is worth the distraction it might cause them, since switching tasks is a mental load.
But with LLMs, almost none of that matters. They don't have goals, motivations or preferences in the same way people do. I can interrupt it all day and it won't get frustrated or lose motivation.
I think anxiety is a harsher word than I mean, but it is close to the feeling I have when I'm about to deliver bad news to someone. When I'm about to say "you know all that work I asked you to do, I need you to throw it away and restart". And I model in my mind the frustration and demotivation this can cause a person. And then I feel anxious about causing them this frustration.
I have to train myself out of that when instructing LLMs. It doesn't mean I have to avoid moments of joy or appreciation. It means I have to understand LLMs have different needs than people, and I have to work towards those needs.
It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear! And it absolutely will not stop, ever, until … the project is complete?
Most of my workflows have slowly moved away from the chat interface with llms. Instead they look more like traditional Unix pipelines that just happen to call Unix tools that interact with llms.
This allows me to make more repeatable processes, not be tied down to vendor implementations of workflows and mix and match models for cost and efficacy.
There is nothing that ties you to talking with the text generator black box, and for most of my use cases it’s a negative.
Would you provide some examples? It sounds like you're feeding results from one LLM into another LLM invocation. Kind of like the loop thing everyone's talking about, but more like a workflow. Or programming with LLMs
In the simplest incarnation I’m just using the built into the agent cli parm’s that trigger non-tui behavior (for instance calling codex exec instead of just codex).
I’ve been experiencing similar feelings. Working with LLMs often takes almost as much mental energy as working with people, but the payoff does not always scale in the same way.
I think we are still on the early days of LLMs. Right now, using them productively requires deliberate thought and an acute knowledge of their limitations. As the author says, it’s easy to get angry at a model, or to foolishly let it nudge you towards more code and more tests — even when that is suboptimal.
To a certain extent, models keep getting better and better at discerning our intentions and providing value. Yet I am not sure whether we will reach a point where using them successfully no longer causes the kind of fatigue that it does today.
I talk almost exclusively with AI these days. There's no one around me who knows programming, I get tired of reading code for projects I'm not interested in, and the projects I am interested in are too difficult, so I just talk with AI, organize my thoughts, and read books.
There's no one around me who does programming. There are hardly any programmers in my town.
The upside is that most programming-related tasks in my town end up going through me. The downside is that there's not much work to begin with, and I can't talk about the things I'm actually interested in.
I'd like to stay in touch with friends who are interested in programming or academia, but since I didn't go to a good university, it seems like I haven't had much of a connection with them
Why talk to AI though, there are plenty of humans you can reach online if you want to
Many evenings I spend on voice chat with friends around the world, these days
You're right. Because conversations with AI don't force me to deal with boring code, don't scratch my ego, and they flatter me.
On top of that, talking to smart developers in real life is exhausting. Putting aside whether they share my interests, there are too many arrogant people. There's also the embarrassment of being asked, 'You don't even know this?' when they have knowledge I lack. The problem is that while that embarrassment could help me grow, it also leaves scars.
And on top of that, the Korean internet is more toxic than you'd think. Most of the male-dominated communities in my interests are filled with misogyny and derogatory remarks. (You can think of Korean internet communities as having 4chan as their baseline.)
So maybe I just chose AI to stay in a greenhouse.
So I'm not sure. Whether I lack the courage to leave the greenhouse, or whether I'm just genuinely exhausted.
The Korean programming communities just spam programming memes, and most of those are factually wrong. I don't want to bother fighting over them.
Even though I've successfully delivered to 40 different companies, my opinions are always seen as 'unsubstantiated personal views,' while those who come from prestigious companies have their opinions treated as 'insights born from experience.' So it feels like there's no one I can have an equal conversation with.
Ultimately, most relationships seem to require the other person to have something to give me, but I don't have anything to give them in return.
I get along quite well with people in real life. But I have no conversations with them about the things I actually care about. And that's lonely. They all say I'm kind and diligent, but I don't have anyone I can truly open up to.
I just use it as Google, like 'ts Omit example' or something.
so I don't get exhausted. If you write complete sentences and say please to the clankers, you're definitely gonna waste energy.
I agree they are too slow though, especially when they "Think" for so long and then say "something went wrong" after 30s.
I didn't expect LLMs to change their output so much depending on how I talk to them. I've done my own controlled tests and verified tone impacts quality even when content is held constant. This effect is called "Linguistic Convergence" or "Conversational Mirroring" and it's been studied extensively. The effect is minimal in coding contexts but becomes much more pronounced in collaborative contexts like creative brainstorming and concept development.
I noticed this because I use LLMs quite often as a note taker, research assistant and reference collector when I'm doing ideation, domain mapping and knowledge acquisition. In those kinds of sessions, if I issue purely directive instructions, the LLM's output quality will begin to drop quite quickly. And if adopt a tone of conversational engagement, the LLMs output quality remains high. As the article states, this can be a burdensome distraction which creates additional cognitive load. It's basically a non-economic cost to using LLMs in these contexts.
Reading research on this, it's fundamental to the nature of LLMs and can't simply be prompted away or easily fixed in fine-tuning. It's an artifact of attention dilution and contextual satiation. If I include semantic richness, structural variety, domain-specific terminology and explicit reasoning steps, it provides higher-entropy tokens to the model's attention mechanism which shifts the weight calculations toward richer areas of the model's latent space. By ingesting a composite of human's collective linguistic structures, it seems like models inherited some of our quirks and sensitivities too.
The productivity gains are well worth any nits I have about responding to the LLM.
LLM: "I've just refactored your code base. Would you like me to also fully document it?"
op: "you're so needy!!!"
Could you imagine how destroyed people like this would be if they faced actual adversity in life? Like if they had to live through the Spanish Flu, or the Great Depression, or hell even lose a loved one to something as now trivial to treat as tetanus?
Oh, the horror, having to type to a system that will do your job for you while you sit in an air conditioned office in a comfortable chair listening to a podcast while you work.
> When you use an LLM, [...] you get to pay the social tax: you converse and negotiate and convince and sometimes even get angry1 at the so-called tool.
Might be a subjective opinion, but this is how writing code always felt to me, even pre-LLMs. An ongoing inner conversation where I try to convince the text on the screen to match the text in my head. It never really felt like tool use in the sense of manual labor.
Not my experience, but I'm still new at this.
The way I have worked so far is to look for ways I can influence the model's "thinking" and then add that to my main AGENTS.md. I try to steer it towards a thought process that mirrors or exceeds my own. I find it a fun challenge. I think this stuff becomes less necessary in a year or so as these sorts of tweaks become part of the shipped product from the model makers.
I think it depends. If talking to a tool solves a bigger exhaustion (a bug that has been bugging you for a while) then the exhaustion from talking to the tool becomes normalized. I guess this is true for any tool really. Using a hammer causes a bit of strain in your muscles but if it solves a bigger issue of nailing something in the wall then that strain in the muscle is fine. It will take some time to develop that muscle working with AI too.
My exhaustion comes from how long winded LLMs are. I ask simple questions and get an essay with bullet points.
i read this post while AI was busy managing jira tickets for me. Otherwise I wouldn't have had the time to be browsing HN right now
Looks like the author submitted it themselves to lobste.rs. Some nice discussion there, as well.
https://lobste.rs/s/csgzki/exhaustion_talking_tool
As an aside, it's nice to see that Lobsters has remained a quiet success. As much as I love HN and the work Dan's done to keep it how it is, I welcome to variety. There are vanishingly few places for polite and earnest discussion online these days.
I quickly lost any interest in that site once I realized its dependence on social verification. That "bar" turns the site into an elitist social club. No thanks.
I talk to my AI all day. Sometimes my voice goes hoarse
I feel this. These tools are viscerally unpleasant. Meetings used to be the thing I didn't look forward to but chatting with an AI is the new low point. Reading AI generated text is the written word equivalent of nails on a chalkboard.
>> Is it worth the social brainwork? IDK, for some tasks maybe - there are things a single person can do now that would have been impossible a year ago. But for all tasks? And wouldn’t that social brainwork do more good if it was directed at the real people you are working with?
It's the opposite. Many people find it exhausting to interact with humans, and do so only because they are required to.
Humans often don't understand what you are saying or asking, and they may not know exactly what steps they need to take to find the answer. They get tired. They might get their pride hurt. They might get angry or frustrated. They might judge you because your question is silly or just wrong.
LLMs, for all their faults, have none of these issues. I'm not saying I'd rather talk to LLMs all day every day, but when trying to get shit done, they really can be the superior coworker, especially if you're an introvert and suffer from social-battery-drainage issues.
Fooling yourself into thinking the LLMs are "understanding" what you are saying or asking is a trap. The output you get may be useful, but it is not due to any sort of understanding.
The elements of human work you mentioned are why it can be both rewarding and painful to interact with humans, but at the end of the day it is important to keep trying to do that work/keep trying to interact with each other. I don't know if we want to go back to cubicles where we just talk to robots all day. Of course some work environments are just awful, and there is not much remedy for that.
It’s going to sound delusional, but my own understanding of my understanding is that there’s probably no such thing, only a feeling of familiarity and some heuristics.
All I do is invoke “pictures” of semantic blobs in my mind, and my unconscious parts pull up more “pictures” in connection to that. Which I verbalize in turn, apply logic checks, and thus get this feeling that I’m meaningfully thinking about something, a sensation of comprehension. I could be even blatantly wrong about something, misunderstanding doesn’t feel any different in the moment.
And reasoning models (try to) capture that. The underlying processes are certainly different, it’s a Chinese Room alright - but if my own understanding is an illusion, the fact that machine model of it isn’t how I do it myself (but only a statistical approximation) doesn’t seem to matter - at least not for inputs and outputs between two black boxes (me or model).
I just don’t believe in p-zombies any more than I believe in Santa, I guess.
>> Fooling yourself into thinking the LLMs are "understanding" what you are saying or asking is a trap. The output you get may be useful, but it is not due to any sort of understanding.
I hear this a lot but I think it's a matter of semantics and ultimately not very useful. I don't care whether the LLM understands me the way a human would. I use the LLM to get useful output. I want it to do something and it does that thing.
AI is the new dope for developers.
Introverts are first line of serious addicts.
ADHD developers are next.
Procrastinators are after that.
[flagged]
[flagged]
[flagged]