What is your motivation for using AI despite the tension you feel? Pressure at work? Feeling the utility is just too good to say no to? Do you feel you're getting what you want out of your effort still?
It is just so easy, I know exactly what I want, I know how to make it appear, even though I know how to write it myself and I know it is good for me to do so, as I start typing my brain just goes into 'why are you spending time doing this, you could be doing the next thing, lets get this quickly and move to the next thing' and all of a sudden there is nothing left, and I look back, and there was nothing to begin with.
"the price of getting what you want is getting what you once wanted" -- Neil Gaiman
If I sat down to read War and Peace and it slowly morphed into a film adaptation and eventually morphed into TikTok I'd be pissed. That is often how I feel when I try to use AI to learn or create.
If I sat down to read War and Peace and it slowly morphed into a fim adaptation than a TikTok version of a film adaptation, it might still be an improvement over «War and Peace».
It takes effort to be that bad. (And people who did not get enough from reading the release version, say that surviving drafts are nicer to read than the final version)
as I start typing my brain just goes into 'why are you spending time doing this, ...
I think this is what they call "cognitive offloading" which leads to "cognitive surrender." Doing things that require effort seems to be something that you use or lose. As you say, it's addictive to want to avoid effort if you can, and it seems to show up almost immediately when you start using LLMs. These are the real effects that worry me. I choose not to subject myself to this in my own time since there is absolutely no reason to be hurrying so much.
I think its interesting that a research professor can stay sharp when they delegate mundane, but necessary, labor to interns and graduate students, but delegating similar kinda of work to an LLM automatically makes you dull.
I think it depends on how you engage with the process. People can use AI to turn of their brains, but it isn't the only way to engage with the technology.
I can't tell in which direction you are being tongue-in-cheek but the difference is probably a direct consequence of the changes you made for the purpose of the analogy.
Hm. Then I have answered you already. Maybe a radical stance but at this point I think anyone who earnestly compares LLMs with humans should be sent to bed early without dinner
I'm not a professor but I imagine delegation like this to be similar to when I delegate work to a junior colleague that makes for a good learning task. I still have to fully understand it and will likely be putting in more work to supervise than I would doing it myself in many cases. But the investment in learning is worth it - the colleague won't need as much help next time.
If I was delegating because I thought "why am I spending my time doing this" I think that would make a world of difference.
I am not even sure addictive is the right word to use. Addictive how exactly? Addictive would mean you want to be in some state more and more, but there is no state to be in.
Cognitive offloading/surrender is when you ask the LLM 'tell me what to do', in contrast when you tell it 'make me 6502 emulator to run apple2 basic from this rom'. The problem is, the road from 'make me 6502 emulator' leads to 'please tell me what to do' very quickly.
But something is malignantly addictive if (1) it causes real problems for the addict, and (2) it offers itself as relief from the very problems it causes. A malignant addiction is also distinguished for spreading the problems of the addiction out and in in interference patterns, creating difficulties for relationships, communities, and the addict’s very sense of self and soul.
He's talking about television in the essay, but I think it also applies to agentic coding. The real problem it causes is a shrinking tolerance for thinking deeply about details, and the relief it provides is that it often, with enough tokens, quickly obviates having to do such thinking. Maybe the latter qualifications also apply.
That said, I willingly use agents for coding and other cognitive work all the time. Like most useful technologies there's a bit of a (faustian?) bargain that's made with AI. My hope is that on balance I'm better off, like the way my GPS lets me more confidently explore new cities and trails even though I struggle with basic directions outside of my neighborhood (RIP hippocampus).
It's anecdotal, but I have a friend that (in my opinion) is very over-reliant on GPS, to the point where the "best route" it picked one time was a one-lane, barely paved rural road without any form of lighting and lots of "Deer Crossing" signs (which can ruin your night very quickly if you fail to spot a crossing deer). It was not a fun ride.
I've also come across plenty of articles over the years about people killing themselves by driving off cliffs or into bodies of waters by blindly following GPS directions (and even more sadly more than one involved an ambulance).
For consistency's sake, I avoid GPS as much as I do LLMs which is probably a much less popular stance. As a result I have a good sense of direction, which is something you can develop, not some innate thing you have to be born with. I do occasionally go the wrong way and have to check a map, but it's not often. Other people I know who rely on GPS completely can't find basic locations even in their own neighborhoods.
I don't have any moral issue with GPS but if you're going to spend a lot of your life doing something, in my opinion, you might as well get good at it.
I guess my use of LLMs and map apps aligns: it'd better not be tied to an internet connection, and as long as it gets enough boring work right, I expect to correct things here and there…
wow this is exactly how i feel every time i sit down to code again. i miss it so much but within 4 minutes i have the same internal monologue and dont miss it any more, for a time. im starting to accept it will never be the same. ive seen too much and now i get more out of seeing my ideas be built in real time than i do from managing context switching cost of using ai for things i could do but dont want to and writing by hand the things i “ought” to
I was not commenting on your AI stance, I was commenting on mcDonalds. The way the sentence reads, it sounds like you are saying, "Guys, we may not like it but we need to consume it. Just like McDonalds!".
that reads funny because no one needs to (or should imo) eat McDonalds.
yea maybe I should rephrase that, by McDonalds I meant the alienation between me and my food. When I was 12 years old I was a shepherd for a bit, and some times I had to eat animals I was friends with, the relationship I had with my food was so different, and now it is just calories I consume.
Honestly, I think it made me appreciate it more, although I think I had less of an emotional attachment to our cows, chickens, and turkeys than OP had with their flock. It’s been 25 years or so since that’s been a part of my life, but my wife and I source all of our beef for the year from a local farmer/family friend. We have met our food, we know it was well taken care of, and we are thankful for it. And to OP’s point, we’ll still buy chicken fingers or pork chops at the grocery store sometimes. We’re not dogmatic about it but we appreciate it.
I've thought chorded keyboards were interesting ever since I read a magazine article about one in the 1980s. It was about a palm-sized word processor (roughly the shape of a large computer mouse), with a one-line LCD. The intention was that you could use it to input raw text, to be transferred and edited on another computer. One button for each finger, except two for the thumb. I don't remember what it was called.
But I've never tried a chorded keyboard myself.
The fastest I've even typed on a non-traditional keyboard is somewhere in the 50wpm range on my Samsung Sidekick 4G Android phone (from 15 years ago). Just using my two thumbs, but it was a strain to hit that speed.
bmo | 22 hours ago
What is your motivation for using AI despite the tension you feel? Pressure at work? Feeling the utility is just too good to say no to? Do you feel you're getting what you want out of your effort still?
[OP] jackdoe | 22 hours ago
It is just so easy, I know exactly what I want, I know how to make it appear, even though I know how to write it myself and I know it is good for me to do so, as I start typing my brain just goes into 'why are you spending time doing this, you could be doing the next thing, lets get this quickly and move to the next thing' and all of a sudden there is nothing left, and I look back, and there was nothing to begin with.
"the price of getting what you want is getting what you once wanted" -- Neil Gaiman
wmurra | 22 hours ago
If I sat down to read War and Peace and it slowly morphed into a film adaptation and eventually morphed into TikTok I'd be pissed. That is often how I feel when I try to use AI to learn or create.
k749gtnc9l3w | 19 hours ago
I cannot resist the tempation, sorry.
If I sat down to read War and Peace and it slowly morphed into a fim adaptation than a TikTok version of a film adaptation, it might still be an improvement over «War and Peace».
(Yes, I have read it.)
wmurra | 18 hours ago
yeah I wouldnt be able to resist the temptation either, thats why id be angry!
k749gtnc9l3w | 18 hours ago
Anything that reminds people how «War and Peace» is intentionally badly written is a good thing, though!
I guess on the software side, the analogy is that you need to be explicitly malicious to release an OS like Windows 11, LLMs or not are secondary.
tome | 5 hours ago
It's intentionally badly written?
k749gtnc9l3w | 4 hours ago
It takes effort to be that bad. (And people who did not get enough from reading the release version, say that surviving drafts are nicer to read than the final version)
tome | 4 hours ago
Any links you can share to those people's comments?
k749gtnc9l3w | 3 hours ago
I think the version I remember is from a video, which makes finding a reference complicated…
bmo | 21 hours ago
Thanks for sharing.
I think this is what they call "cognitive offloading" which leads to "cognitive surrender." Doing things that require effort seems to be something that you use or lose. As you say, it's addictive to want to avoid effort if you can, and it seems to show up almost immediately when you start using LLMs. These are the real effects that worry me. I choose not to subject myself to this in my own time since there is absolutely no reason to be hurrying so much.
xjix | 16 hours ago
I think its interesting that a research professor can stay sharp when they delegate mundane, but necessary, labor to interns and graduate students, but delegating similar kinda of work to an LLM automatically makes you dull.
I think it depends on how you engage with the process. People can use AI to turn of their brains, but it isn't the only way to engage with the technology.
tentacloids | 9 hours ago
I can't tell in which direction you are being tongue-in-cheek but the difference is probably a direct consequence of the changes you made for the purpose of the analogy.
xjix | 3 hours ago
I'm actually asking the question that I'm asking.
tentacloids | 2 hours ago
Hm. Then I have answered you already. Maybe a radical stance but at this point I think anyone who earnestly compares LLMs with humans should be sent to bed early without dinner
bmo | 2 hours ago
I'm not a professor but I imagine delegation like this to be similar to when I delegate work to a junior colleague that makes for a good learning task. I still have to fully understand it and will likely be putting in more work to supervise than I would doing it myself in many cases. But the investment in learning is worth it - the colleague won't need as much help next time.
If I was delegating because I thought "why am I spending my time doing this" I think that would make a world of difference.
[OP] jackdoe | 18 hours ago
I am not even sure addictive is the right word to use. Addictive how exactly? Addictive would mean you want to be in some state more and more, but there is no state to be in.
Cognitive offloading/surrender is when you ask the LLM 'tell me what to do', in contrast when you tell it 'make me 6502 emulator to run apple2 basic from this rom'. The problem is, the road from 'make me 6502 emulator' leads to 'please tell me what to do' very quickly.
mbil | 16 hours ago
In E Unibus Pluram, DFW says:
He's talking about television in the essay, but I think it also applies to agentic coding. The real problem it causes is a shrinking tolerance for thinking deeply about details, and the relief it provides is that it often, with enough tokens, quickly obviates having to do such thinking. Maybe the latter qualifications also apply.
That said, I willingly use agents for coding and other cognitive work all the time. Like most useful technologies there's a bit of a (faustian?) bargain that's made with AI. My hope is that on balance I'm better off, like the way my GPS lets me more confidently explore new cities and trails even though I struggle with basic directions outside of my neighborhood (RIP hippocampus).
spc476 | 15 hours ago
It's anecdotal, but I have a friend that (in my opinion) is very over-reliant on GPS, to the point where the "best route" it picked one time was a one-lane, barely paved rural road without any form of lighting and lots of "Deer Crossing" signs (which can ruin your night very quickly if you fail to spot a crossing deer). It was not a fun ride.
I've also come across plenty of articles over the years about people killing themselves by driving off cliffs or into bodies of waters by blindly following GPS directions (and even more sadly more than one involved an ambulance).
bmo | 15 hours ago
For consistency's sake, I avoid GPS as much as I do LLMs which is probably a much less popular stance. As a result I have a good sense of direction, which is something you can develop, not some innate thing you have to be born with. I do occasionally go the wrong way and have to check a map, but it's not often. Other people I know who rely on GPS completely can't find basic locations even in their own neighborhoods.
I don't have any moral issue with GPS but if you're going to spend a lot of your life doing something, in my opinion, you might as well get good at it.
k749gtnc9l3w | 7 hours ago
I guess my use of LLMs and map apps aligns: it'd better not be tied to an internet connection, and as long as it gets enough boring work right, I expect to correct things here and there…
anex9d | 17 hours ago
wow this is exactly how i feel every time i sit down to code again. i miss it so much but within 4 minutes i have the same internal monologue and dont miss it any more, for a time. im starting to accept it will never be the same. ive seen too much and now i get more out of seeing my ideas be built in real time than i do from managing context switching cost of using ai for things i could do but dont want to and writing by hand the things i “ought” to
gspr | 10 hours ago
You and I are so incredibly different. And that is fine.
wmurra | 23 hours ago
Funny sentence. It makes it sound like getting used to MacDonalds is a good thing.
[OP] jackdoe | 23 hours ago
Its just saying: it is what it is. I have been in this for a while and its time to move on.
I wrote https://punkx.org/jackdoe/misery.html May 2025 and finally I feel I am making some progress. Its time to let go.
wmurra | 22 hours ago
I was not commenting on your AI stance, I was commenting on mcDonalds. The way the sentence reads, it sounds like you are saying, "Guys, we may not like it but we need to consume it. Just like McDonalds!".
that reads funny because no one needs to (or should imo) eat McDonalds.
[OP] jackdoe | 22 hours ago
yea maybe I should rephrase that, by McDonalds I meant the alienation between me and my food. When I was 12 years old I was a shepherd for a bit, and some times I had to eat animals I was friends with, the relationship I had with my food was so different, and now it is just calories I consume.
wmurra | 22 hours ago
woah thats crazy. also a very interesting analogy. I have no idea what it might feel like to eat something I once loved.
tonyarkles | 16 hours ago
Honestly, I think it made me appreciate it more, although I think I had less of an emotional attachment to our cows, chickens, and turkeys than OP had with their flock. It’s been 25 years or so since that’s been a part of my life, but my wife and I source all of our beef for the year from a local farmer/family friend. We have met our food, we know it was well taken care of, and we are thankful for it. And to OP’s point, we’ll still buy chicken fingers or pork chops at the grocery store sometimes. We’re not dogmatic about it but we appreciate it.
ansible-rs | 2 hours ago
I've thought chorded keyboards were interesting ever since I read a magazine article about one in the 1980s. It was about a palm-sized word processor (roughly the shape of a large computer mouse), with a one-line LCD. The intention was that you could use it to input raw text, to be transferred and edited on another computer. One button for each finger, except two for the thumb. I don't remember what it was called.
But I've never tried a chorded keyboard myself.
The fastest I've even typed on a non-traditional keyboard is somewhere in the 50wpm range on my Samsung Sidekick 4G Android phone (from 15 years ago). Just using my two thumbs, but it was a strain to hit that speed.