A classic blog post of the "I ran an expensive prompt in Claude, included a table of token usage, and wrapped up with a reflection on how machines once took jobs away from wheelbarrow pushers - yet chess players are still playing after Deep Blue" variety.
Sometimes I think about that. In the LLM era, what is the meaning of my existence? Compared to the past, I’ve become able to do so many things, but there isn’t really anything I want to make. I don’t know where to place my identity.
This is a guess (correct me if I'm wrong) but probably you valued the mastery of a skill to create, and now the mastery is worthless as anyone can create anything, be it music, art or software if they just have access to latest AI?
It could be that, or it could not be—but rather than that, maybe it’s because work has decreased. Normally around this time of year is a season when freelancers have a lot of work, but there isn’t any, so maybe that’s why. People say I should just do creative work for myself, but maybe I’m feeling rushed because I don’t have any ideas.
If you’re spiritual, AI doesn’t change the meaning of your existence. And if you’re not spiritual, didn’t you always accept that you are a transient region of low entropy? You’re a carbon-based computer turning food into bits and now silicon-based computers can turn electricity into bits faster.
> The max effort + ultracode version is even more stunning, but interestingly it never got mouse input working without errors. If you want to test it out, you need to stick to keyboard controls - see the Use chapter in the project’s README file.
god is this a perfect encapsulation of the jagged nature of LLMs. incredible and kind of broken. for now at least
Makes sense though! It's really tough to test touch control programmatically, and that's all LLMs have. Same reason ChatGPT can't play Super Mario Bros.
> I watched numerous YouTube videos and read many social media posts and blog posts by programmers about a profound sense of alienation, loss of human dignity, and anger over stolen independence.
Yes, it's funny to see programmers of all people, uber-technologists, shoot "i'm walking around in the woods videos" decrying the end of human civilisation merely because one specific craft practice is being automated.
AI isnt taking 'human creativity' away, this is an incredibly parochial and self-interested take. Its taking human creativity, as expressed in certain 20th C. crafts, away. Just as textile art has gone, and we don't have lace everywhere any more -- so it goes, that handcrafted programming goes to.
Engineers are those people who do that to human civilisation. That's always been the job. Replace people with machines. So it goes.
I have a few acquaintances who smirked when I last met them: they love how AI saved them a ton of time in their non-tech jobs while pointing out that "they wouldn't want to be a software engineer today".
They were always a bit resentful about how I made money "so easily". And they were right in a sense. Before retiring I'd make more money browsing reddit in a boring meeting than they would make in a day.
We made incredible improvements to society but we also destroyed the livelihoods of millions. I believe our profession was/is a massive net positive to the world, but many got hurt while we steamrolled others without hesitation.
I'm actually very positive about AI in general and I don't think it will end our field. But if that's what happens, we can lament the bad luck but not really complain that we've been wronged.
I'd say its more like the creation of power tools. Many people are now capable of DIY. Many cannot carve stone masonry. Many still work on construction sites. Many help people with their DIY and post-DIY disasters. Some still create ironwork and stone masonry.
Its not yet clear to me that blacksmith software engineers that quickly learn DIY tools, arent going to be in greater demand. Nor even, capable of commanding higher salaries. Perhaps fewer will exist, but they will transact with more end customers because their rate of production will increase -- and so, be richer still.
Its my view, incidentlly, that far fewer programmers ought exist. If you think about the last decade or two of the industry, and the rise of massive low-quality bootcamps -- imv we had an seng bubble. It isnt clear to me that we arent lower than a pre-bubble baseline.
I think it's only "one specific craft" that's being heavily affected right now because software engineers are lazy (in a good way) and as they've been developing generative AI, automating away their own work was the natural choice.
And it has rolled out to the rest of the software engineering industry rapidly because a) it was adjacent to AI development, and b) software engineers are very expensive workers. Graphical design is also being hit due to its proximity to software engineering.
Other information workers are going to start feeling the hurt soon enough, but their respective industry timeline will be correlated with how much money it's going to save to replace them.
Paralegals, customer support, copywriters, data analysts, and any other fields that are purely informational are all going to feel the squeeze.
The ability to automate the reinforcement of "working code" within LLM training means software is a mostly unique case where "good enough correctness" can be done at scale. LLMs are text generation machines, code is text, cli is text: only commands on computers are the kinds of text whose correctness is automated.
Its unclear what other jobs have a billion experts to-hand to grade LLM outputs. Its also unclear what other jobs would even benefit from such a thing.
For most others, it seems LLMs are basically tutors and mentors which existed anyway.
Software Development is far from the only white collar work frontier agents are very useful for. Most people just have no idea and marketing around them hasn't really struck. If you think it's just code or math that LLMs have gotten rapidly better at then you'd be wrong.
I think alignment is the missing part of this post.
Creating something visually interesting Invisible-Cities-ish is basically a task with a billion or more satisfactory outcomes. You don't have a precise view on what makes an invisible city a good one as long as it has all the fancy looking bells and whistles. This means the AI can generate a bunch of tokens which as long as they technically work (which it seems the max-thinking did not fully) produce a "good" result. This makes you think - wow AI is impressive. However if many things in the system really mattered, it might have been much less successful as the AI may have made poor assumptions. This was basically a task that given a large volume of tokens and technical competence it's fairly easy for the AI to do a good job at because we don't evaluate it with any depth/precision. I think the same is true for music, where there are basically an infinite number of combinations of sounds, beats, melodies, etc. which are pleasant to hear - and songs aren't mission critical - so alignment is not a big issue. Generative AI is good at these tasks.
Correct. There's another angle to this too; even for tasks that do have strict requirements, the asker may just have low standards, so to them the output is perfectly acceptable, even if objectively very far from optimum. Hence the nonsense like "look, AI one-shot Photoshop!" we hear so much about. No, it didn't.
For me; no identity crisis. Or at least a very short lived one until I realised what makes me tick.
I’ve been coding since I was 6. Making stuff. Making stuff in meatspace, too.
I always told myself that I enjoyed coding, that I enjoyed making stuff.
I was wrong. I hate coding. I hate standing in a ditch with heavy cables slung over me. I hate neck and eye strain from hours bent over a soldering iron.
What I like, what drives me - the outcomes. The neat bit of software that I conceptualised and now have in my hands. The mad Heath Robinson contraption that somehow makes the machine that goes “ping!”, in fact, go ping.
I’ve always recognised that I am a lazy asshole - but there has always been a carve-out for creating - yet the giant heap of unfinished projects tells me my “eh, amanhã” attitude does in fact persist throughout many avenues of my life, even those I thought I enjoyed.
Me, I fucking love vibe coding. I get to go to bed at a sensible hour, and wake up in the morning like it’s bloody Christmas Day, as I know there’s going to be something cracking to play with. The ideas I have, realised, without me having to slog through the weeds. Nobody sucking their teeth at the scope. No “madaxe got distracted by another shiny object”.
I’ve executed on basically every digital must and should that I’ve come up with over the last decade, in the last nine months. I’m now onto the “could haves” and “would like to haves”.
And… I’ll be honest. Maintainability, understanding what it does under the hood - for the first time in my life I barely care about, as next year I can have a better model do it again, and better.
So, I like the outcome. I like having an idea and seeing it come to life. What happens between A and B isn’t terribly important, as it turns out that not all paths are the goal.
And I think the same may apply to a lot of creative people - I mean, artists had and have artisans for a reason - and now we can all have an ever-improving artisan.
I’m in a similar boat. I am working on some stuff or have rather expanded my scope to be truly ambitious, and it still boggles my mind what I have & can do all alone today.
I find other devs/colleagues who don’t have ideas of what to work on a bit hard to understand . Just can’t comprehend how someone won’t have any desire to do nothing else with this tech other than finishing off jira tickets.
And a few of them have expressed to me that whatever I think I enjoy today is only cause I have found “purpose” and soon both what I’m building will be easy for anyone to replicate and thus become worthless, and I won’t be happy then. I have doubts about both.
It's not "prompting skills". AI is great at going and building whatever you tell it to build, but it's terrible at knowing what to build. It has no taste, no knowledge of what exactly would useful to a human vs not.
Most people don't either, and nor do 99% of programmers — hence their bitching that their mechanical transformation of requirements they got fed by someone else into code is now becoming worthless. It was never worth that much to begin with.
I'm very much like the grandparent poster: I love that I can now focus on what actually matters — which features are to be included, which excluded, and how every feature should work w.r.t the all-guiding user experience. All without having to wonder which sorting algorithm will be optimal here or how to wire up a db migration.
The output isn’t interesting to me. I clicked around in both and it seemed kind of neat but like a portfolio project for school and certainly not art that would engage me.
I’m not sure how this causes an identity crisis. But if my purpose in life was to churn out random stuff that wasn’t meaningful to people, I would assume that I would be crisising much earlier than now.
Everyone always wants to compare this revolution to the Industrial Revolution but there's a saying "history doesn't repeat itself but it rhymes". There are similarities, and reading how people coped back then can be helpful, but also this is a completely different revolution!
The only themes that rhyme imo are the loss of power to a smaller subset of humans (owners) and technological advancement enabling us to produce more (not necessarily better quality) with smaller costs. But people think that the history will repeat itself, and Jevons paradox will hold true, because it held in the Industrial Revolution; quote from the article: "Eventually, everything should be fine. The technology will be more powerful, and the job will still be there, it’ll just look different."
In this revolution, what if instead of being the human that invented the steam engine, we are the horses? "What happened to all the horses replaced by machines after industrialization? Humans keep a few around for entertainment but that’s about it." [1]
26 comments
[ 0.31 ms ] story [ 7.6 ms ] threadAnd we came to this conclusion because?
waste of time.
god is this a perfect encapsulation of the jagged nature of LLMs. incredible and kind of broken. for now at least
Yes, it's funny to see programmers of all people, uber-technologists, shoot "i'm walking around in the woods videos" decrying the end of human civilisation merely because one specific craft practice is being automated.
AI isnt taking 'human creativity' away, this is an incredibly parochial and self-interested take. Its taking human creativity, as expressed in certain 20th C. crafts, away. Just as textile art has gone, and we don't have lace everywhere any more -- so it goes, that handcrafted programming goes to.
Engineers are those people who do that to human civilisation. That's always been the job. Replace people with machines. So it goes.
They were always a bit resentful about how I made money "so easily". And they were right in a sense. Before retiring I'd make more money browsing reddit in a boring meeting than they would make in a day.
We made incredible improvements to society but we also destroyed the livelihoods of millions. I believe our profession was/is a massive net positive to the world, but many got hurt while we steamrolled others without hesitation.
I'm actually very positive about AI in general and I don't think it will end our field. But if that's what happens, we can lament the bad luck but not really complain that we've been wronged.
Its not yet clear to me that blacksmith software engineers that quickly learn DIY tools, arent going to be in greater demand. Nor even, capable of commanding higher salaries. Perhaps fewer will exist, but they will transact with more end customers because their rate of production will increase -- and so, be richer still.
Its my view, incidentlly, that far fewer programmers ought exist. If you think about the last decade or two of the industry, and the rise of massive low-quality bootcamps -- imv we had an seng bubble. It isnt clear to me that we arent lower than a pre-bubble baseline.
And it has rolled out to the rest of the software engineering industry rapidly because a) it was adjacent to AI development, and b) software engineers are very expensive workers. Graphical design is also being hit due to its proximity to software engineering.
Other information workers are going to start feeling the hurt soon enough, but their respective industry timeline will be correlated with how much money it's going to save to replace them.
Paralegals, customer support, copywriters, data analysts, and any other fields that are purely informational are all going to feel the squeeze.
Its unclear what other jobs have a billion experts to-hand to grade LLM outputs. Its also unclear what other jobs would even benefit from such a thing.
For most others, it seems LLMs are basically tutors and mentors which existed anyway.
Creating something visually interesting Invisible-Cities-ish is basically a task with a billion or more satisfactory outcomes. You don't have a precise view on what makes an invisible city a good one as long as it has all the fancy looking bells and whistles. This means the AI can generate a bunch of tokens which as long as they technically work (which it seems the max-thinking did not fully) produce a "good" result. This makes you think - wow AI is impressive. However if many things in the system really mattered, it might have been much less successful as the AI may have made poor assumptions. This was basically a task that given a large volume of tokens and technical competence it's fairly easy for the AI to do a good job at because we don't evaluate it with any depth/precision. I think the same is true for music, where there are basically an infinite number of combinations of sounds, beats, melodies, etc. which are pleasant to hear - and songs aren't mission critical - so alignment is not a big issue. Generative AI is good at these tasks.
I’ve been coding since I was 6. Making stuff. Making stuff in meatspace, too.
I always told myself that I enjoyed coding, that I enjoyed making stuff.
I was wrong. I hate coding. I hate standing in a ditch with heavy cables slung over me. I hate neck and eye strain from hours bent over a soldering iron.
What I like, what drives me - the outcomes. The neat bit of software that I conceptualised and now have in my hands. The mad Heath Robinson contraption that somehow makes the machine that goes “ping!”, in fact, go ping.
I’ve always recognised that I am a lazy asshole - but there has always been a carve-out for creating - yet the giant heap of unfinished projects tells me my “eh, amanhã” attitude does in fact persist throughout many avenues of my life, even those I thought I enjoyed.
Me, I fucking love vibe coding. I get to go to bed at a sensible hour, and wake up in the morning like it’s bloody Christmas Day, as I know there’s going to be something cracking to play with. The ideas I have, realised, without me having to slog through the weeds. Nobody sucking their teeth at the scope. No “madaxe got distracted by another shiny object”.
I’ve executed on basically every digital must and should that I’ve come up with over the last decade, in the last nine months. I’m now onto the “could haves” and “would like to haves”.
And… I’ll be honest. Maintainability, understanding what it does under the hood - for the first time in my life I barely care about, as next year I can have a better model do it again, and better.
So, I like the outcome. I like having an idea and seeing it come to life. What happens between A and B isn’t terribly important, as it turns out that not all paths are the goal.
And I think the same may apply to a lot of creative people - I mean, artists had and have artisans for a reason - and now we can all have an ever-improving artisan.
I find other devs/colleagues who don’t have ideas of what to work on a bit hard to understand . Just can’t comprehend how someone won’t have any desire to do nothing else with this tech other than finishing off jira tickets.
And a few of them have expressed to me that whatever I think I enjoy today is only cause I have found “purpose” and soon both what I’m building will be easy for anyone to replicate and thus become worthless, and I won’t be happy then. I have doubts about both.
Most people don't either, and nor do 99% of programmers — hence their bitching that their mechanical transformation of requirements they got fed by someone else into code is now becoming worthless. It was never worth that much to begin with.
I'm very much like the grandparent poster: I love that I can now focus on what actually matters — which features are to be included, which excluded, and how every feature should work w.r.t the all-guiding user experience. All without having to wonder which sorting algorithm will be optimal here or how to wire up a db migration.
I’m not sure how this causes an identity crisis. But if my purpose in life was to churn out random stuff that wasn’t meaningful to people, I would assume that I would be crisising much earlier than now.
It doesn't - author outright states his title and premise were complete clickbait.
The only themes that rhyme imo are the loss of power to a smaller subset of humans (owners) and technological advancement enabling us to produce more (not necessarily better quality) with smaller costs. But people think that the history will repeat itself, and Jevons paradox will hold true, because it held in the Industrial Revolution; quote from the article: "Eventually, everything should be fine. The technology will be more powerful, and the job will still be there, it’ll just look different."
In this revolution, what if instead of being the human that invented the steam engine, we are the horses? "What happened to all the horses replaced by machines after industrialization? Humans keep a few around for entertainment but that’s about it." [1]
[1] https://news.ycombinator.com/item?id=50147755