[whimsicalism]

It’s funny reading this parallel world that some portion of people have constructed for themselves.

It has been three years and these tools can do a considerable portion of my day to day work. Salvage the wreckage? Unfortunately I think that many people’s jobs are essentially in the “Coyote running off a cliff but not realizing it yet” phase or soon to be.

[kuerbel]

I think this comment is reacting to a different argument than the one the article is actually making.

The piece isn’t claiming that AI tools are useless or that they don’t materially improve day-to-day work. In fact, it more or less assumes the opposite. The critique is about the economic and organizational story being told around AI, not about whether an individual developer can ship faster today.

Saying “these tools now do a considerable portion of my work” operates on the micro level of personal productivity. Doctorow is operating on the macro level: how firms reframe human labor as “automation,” push humans into oversight and liability roles, and use exaggerated autonomy claims to justify valuations, layoffs, and cost-cutting.

Ironically, the “Wile E. Coyote running off a cliff” metaphor aligns more with the article than against it. The whole “reverse centaur” idea is that jobs don’t disappear instantly; they degrade first. People keep running because the system still sort of works, until the ground is gone and the responsibility snaps back onto humans.

So there’s no contradiction between “this saves me hours a day” and “this is being oversold in ways that will destabilize jobs and business models.” Those two things can be true at the same time. The comment seems to rebut “AI doesn’t work,” which isn’t really the claim being made.

[whimsicalism]

You can read my reply to another comment making a similar point. In short, I think you are giving Doctorow far too much credit - the assumption that these tools are fundamentally incapable is woven throughout the essay, the risk always comes from the fact that managers might think these tools (which are obviously inferior) can do your job. The notion that they can actually do your job is treated as invariable absurd, pie-in-the-sky, bubble thinking, or unmentionable.

My point is I don’t think a technology that went from chatgpt (cool, useless) to opus-4.5+ in 3 years is obviously being oversold when it says that it can do your entire job beyond being just a useful tool.

[happy_dog1]

I think we have to be careful when assuming that model capabilities will continue to grow at the same rate they have grown in recent years. It is very well-documented their growth in recent years has been accompanied by an exponential increase in the cost of building these models, see for example (of many examples) [1]. These costs include not just the cost of GPUs but also the cost for reinforcement learning from human feedback (RLHF), which is not cheap either -- there is a reason that SurgeAI has over $1 billion in annual revenue (and ScaleAI was doing quite well before they were purchased by Meta) [2].

Maybe model capabilities WILL continue to improve rapidly for years to come, in which case, yes, at some point it will be possible to replace most or all white collar workers. In that case you are probably correct.

The other possibility is that capabilities will plateau at or not far above current levels because squeezing out further performance improvements simply becomes too expensive. In that case Cory Doctorow's argument seems sound. Currently all of these tools need human oversight to work well, and if a human is being paid to review everything generated by the AI, as Doctorow points out, they are effectively functioning as an accountability sink (we blame you when the AI screws up, have fun.)

I think it's worth bearing in mind that Geoffrey Hinton (infamously) predicted ten years ago that radiologists would all be out of a job in five years, when in fact demand for radiology has increased. He probably based this on some simple extrapolation from the rapid progress in image classification in the early 2010s. If image classification capabilities had continued to improve at that rate, he would probably have been correct.

[1] https://arxiv.org/html/2405.21015v1 [2] https://en.wikipedia.org/wiki/Surge_AI

[lostmsu]

No, models significantly improved at the same cost. Last year's Claude 3.7 has since been beaten by GPT-OSS 120B that you can run locally and is much cheaper to train.

[judahmeek]

And GPT-OSS's architecture improvements aren't already incorporated in SotA models?

[lostmsu]

The point is, that contradicts the claim that lately the progress is only made by throwing more compute.

[roxolotl]

But Corey isn’t saying it’s oversold he’s saying the value capture by a few companies enabled by AI is dangerous to society.

[whimsicalism]

I do not agree with your reading of the article. The premise - both implicit and stated explicitly throughout the article - is that companies are hyping this up because they want to be seen as growing, that this technology cannot do your job, that these are statistical tools foolishly being used to replace real workers. Look at the bits I quote in my other comment.

I would have been much more interested in reading the article you’re suggesting.

[johnnyanmac]

You need to read the article again with a more charitable lens. He starts with

>What I do not do is predict the future. No one can predict the future, which is a good thing, since if the future were predictable, that would mean we couldn’t change it.

It feels like shoving words on Cory's mouth to make statements like "he's saying it can't replace us". That is the exact point he avoids in the article to focus on the human. Not the tech and capabilities.

[artninja1988]

> The piece isn’t claiming that AI tools are useless or that they don’t materially improve day-to-day work

Would you call something that could replace your labor "spicy auto complete"? He also evokes nfts and blockchain, for some reason. To me this phrasing makes it sound like he thinks they are damn near useless.

[johnnyanmac]

What cory thinks now doesn't affect the piece that focuses on the future and how humanity shapes around it. It feels like missing the forest for the trees to focus on what he feels about AI in c.2025 when the piece is talking about ramifications.

[sodapopcan]

> I think this comment is reacting to a different argument than the one the article is actually making.

The headline.

[bee_rider]

We’re supposed to pretend people read articles instead of just the headline (it is in the posting guidelines). To play along with that rule, people will write as if the poster they are responding to missed some nuance of the article.

[whimsicalism]

I like that you all are having your own little side conversation making fun of me without engaging at all on the substance.

[sodapopcan]

This isn't a conversation, lol.

I don't have much to offer here (and yes, sorry, after I made my snarky remark I realize you had indeed read the article). I recognize AI's capabilities but mainly don't use it primarily for political reasons but also because I just enjoy writing code. I'll sometimes use up the chatgpt free limit using it as a somewhat better search engine (and it's not always better) but there's no way I'm paying for agents, which is everything to do with where the money is going, not the money itself. Of course there are other reasons outside of how AI is used by programmers that would derail the general theme of these threads.

I'm just drawn to these threads for the drama and sometimes it triggers me and I write a snarky throwaway comment. If the discussions, and particularly the companies themselves, could shift to actual societal good it can do and how it is concretely getting there, that would hold my attention. Instead we get Sona etc.

[bee_rider]

That’s fair.

I was accepting sodapopcan’s premise while responding to them. My joke was intended to be aimed at the posting guides and these little Hackernews traditions. But, it was a bit dismissive toward you, which is a little rude. Sorry.

[johnnyanmac]

If you're going to dismiss conversation, why bother with the fascimile?

[zdragnar]

Ah yes, the notoriously accurate headline.

[Sharlin]

I think you’ll find the essay much more nuanced than that. It only incidentally discusses what you’re thinking about.

> Think of AI software generation: there are plenty of coders who love using AI. Using AI for simple tasks can genuinely make them more efficient and give them more time to do the fun part of coding, namely, solving really gnarly, abstract puzzles. But when you listen to business leaders talk about their AI plans for coders, it’s clear they are not hoping to make some centaurs.

[kalkin]

The article does a pretty lazy* job of defending its assumption that "solving really gnarly, abstract puzzles" is going to remain beyond AI capabilities indefinitely, but that is a load-bearing part of the argument and Doctorow does try to substantiate it by dismissing LLMs as next-word predictors. This is a description which is roughly accurate at some level of reduction but has not helped anyone predict the last three years of advances and so seems pretty unlikely to be a helpful guide to the next three years.

The other argument Doctorow gives for the limits of LLMs is the example of typo-squatting. This isn't an attack that's new to LLMs and, while I don't know if anyone has done a study, I suspect it's already the case in January 2026 that a frontier model is no more susceptible to this than the median human, or perhaps less; certainly in general Claude is less likely to make a typo than I am. There are categories of mistakes it's still more likely to make than me, but the example here is already looking out of date, which isn't promising for the wider argument.

*to be fair, it's clearly not aimed at a technical audience.

[whimsicalism]

I disagree. The article leads with the sentiment that I mention and has it woven throughout. The theme is AI is obviously not capable of doing your job, the problem is that the stupid managerial class will get convinced it is and make things shitty.

> This is another key to understanding – and thus deflating – the AI bubble. The AI can’t do your job, but an AI salesman can convince your boss to fire you and replace you with an AI that can’t do your job.

> Now, AI is a statistical inference engine. All it can do is predict what word will come next based on all the words that have been typed in the past. That means that it will “hallucinate” a library called lib.pdf.text.parsing,

I think it is a convenient, palatable, and obviously comforting lie that lots of people right now are telling themselves.

To me, all the ‘nuance’ in this article is just because the coyote in Doctorow has begun looking down but still cannot quite believe it. He is still leaning on the same tropes of statistical autocomplete that have been a mainstay of the fingers-in-ears gang for the last 3 years.

[alpha_squared]

You're in half the comment replies with a confrontational tone and, at times, quite aggressively. It does not feel as though you're sincerely engaging, but instead have an ideological world view that makes it difficult to reconcile different perspectives.

I'm working directly with these tools and have several colleagues who do as well. Our collective anecdotal experience keeps coming back to the conclusion that the tech just isn't where the marketing is on its capabilities. There's probably some value in the tech here, which leads others like yourself to be so completely sold on it, but it's just not materializing that much in my day-to-day outside of creating the most basic code/scaffolding where I then have to go back and fix/correct because there are subtle errors. It's actually hard to tell if my productivity is better because I have to spend time fixing the generated output.

Maybe it would help to recognize that your experience is not the norm. And if the tech were there, where are the actual profits from selling it? It's increasingly more common for it to be "under development" for selling to consumers or only deployed as a chatbot in scenarios where it's acceptable to be wrong and warnings to verify output yourself.

[whimsicalism]

I’m replying to the people replying to me, which is hopefully permissible? I will respond aggressively to people who say that my work must not be very important or hard if I feel that AI can do a considerable portion of my day to day because I feel like that is initiating rudeness and find that the HN tendency to talk down to people expressing this opinion is chilling important conversations.

If my other replies come off as aggro, I apologize - I definitely can struggle with moderating tone in comments to reflect how I actually feel.

> Our collective anecdotal experience keeps coming back to the conclusion that the tech just isn't where the marketing is on its capabilities. There's probably some value in the tech here, which leads others like yourself to be so completely sold on it

Let me be clear - I am not so completely sold on the current iteration. But I think there has been a significant improvement even since the midpoint of last year, the number of diffs I am returning mostly unedited is sharply increasing, and many people I am talking to are privately telling me they are no longer authoring any code themselves except for minor edits in diffs. Given that this has only been 3 years since chatgpt, really I am just looking at the curve and saying ‘woah.’

[kalkin]

I don't think the commenter to whom you're replying is any more aggressive than, e.g., this one: https://news.ycombinator.com/item?id=46668988

It's unfortunately the case that even understanding what AI can and cannot do has become a matter of, as you say, "ideological world view". Ideally we'd be able to discuss what's factually true of AI at the beginning of 2026, and what's likely to be true within the next few years, independently of whether the trends are good for most humans or what we ought to do about them. In practice that's become pretty difficult, and the article to which we're all responding does not contribute positively.

[johnnyanmac]

>any more aggressive than, e.g., this one

Frequency is important too.

>independently of whether the trends are good for most humans or what we ought to do about them.

This whole article is about the trends and of they are good for humans. I was pleasantly surprised that this was not yet another argument of "AI is (not) good enough" since people at this point have their fences set on that. I don't think it's too late to talk about how we as humans can manage pandora's box before it opens.

Responses like this dismissing the human element seem to want to isolate themselves from societal effects for some reason. The box will affect you.

[kalkin]

Neither in my previous comment nor in my actual views do I dismiss the human element or expect to isolate myself from societal effects.

> I was pleasantly surprised that this was not yet another argument of "AI is (not) good enough"

The article does assert that, and that's important for its argument that ordinary workers just need to convince decisionmakers that things will go poorly if they replace us.

"Now, if AI could do your job, this would still be a problem... But AI can’t do your job."

This isn't ambiguous.

[meroes]

What masterful framing AI leaders developed that people say, “AI is doing my work”.

Did other technologies get phrased this way? The accounting software is doing my work? The locomotive is doing my work?

[TiredOfLife]

> The accounting software is doing my work?

That is literally what accounting software did. It made the work of 10 people into work of 3 people with software

[whimsicalism]

yeah it’s all branding. there is totally nothing different about this technology that might cause someone to say that…

e:tone

[zdragnar]

The technology is different, but we've been told that we'll be replaced by foreign teams for years. It's had some impact, but in all but two of the cases I've personally seen it has been significantly detrimental. In the majority of cases, the offshore team was either replaced or relegated to pointless busy work.

[whimsicalism]

i don’t find that at all a compelling argument, but maybe others will

[ctoth]

Do you not remember the early days of Covid?

People are really, really, really good at not seeing what they don't want to see.

[DetectDefect]

> It has been three years and these tools can do a considerable portion of my day to day work.

Is this really something you want to have proudly said? Because it makes it sound like your "work" is not very important.

[whimsicalism]

My work involves reading papers, doing high level math, coding it, reasoning about the business environment, etc. etc. My work is important and I find the classic HN impulse to stigmatize people for saying this to be silly.

It is you who is the fool if you haven’t managed to use these things to massively accelerate what you can do and if you cannot see the clear trend. Again, it has been three years since chatgpt came out.

[DetectDefect]

So the work is important and I am the fool because ... you think so? That is not a very intellectually-defensible position to hold. One could even reason the argument is so flawed because LLMs degrade thinking capacity.

[whimsicalism]

You’re arguing in a tautological fashion - anyone who suggests that AI can do a lot of their job must not have been doing important work in the first place. It’s a convenient psychological self defense mechanism that I see often here. Take care.

[fzeroracer]

> My work involves reading papers, doing high level math, coding it, reasoning about the business environment, etc. etc. My work is important and I find the classic HN impulse to stigmatize people for saying this to be silly.

This is what every person who's been laid off by AI says. Every single time. People really like to assume that the work they do is important, except companies don't care about important they care about pushing shit out the door, faster and cheaper. Your high level math and business reasoning do not matter when they can just let someone cheaper go wild and deliver faster with no guard rails.

[whimsicalism]

My job is essentially delivering faster with no guard rails. I know it is a common HN sentiment that nobody can do my job better than me because I am the most thoughtful person to ever do it.

This is explicitly not what I am saying given that I am leading with AI getting close to being able to do much of what is currently my job. I find it hard to imagine a world where we stagnate right where we are and it takes a decade to do anything more aka I cannot imagine a world where a considerable portion of jobs are not automatable soon - and I do not even think it will be shittier.

[sodapopcan]

What will not be shittier about it? What do you envision these jobs are that are going to replace the current ones?

[pixl97]

I mean this is what technology has been doing for a long time... Luddites and all.

HN just has a strong streak with programmers saying they'll never be replaced by the auto-loom.

[hamdingers]

> My work involves reading papers

And yet you did not read this essay, or at least did not understand it.

Whatever LLM you used to summarize it has let you down. I wonder how often that is happening in your day to day work, perhaps that's why you feel your job is at risk.

[whimsicalism]

I am commenting on something in the background space of assumptions of the essay. Just because it is not the main thrust doesn’t mean I didnt read the article or used an LLM to summarize it for me. Take care

[1vuio0pswjnm7]

"It's funny reading this parallel world that some portion of people have constructed for themselves."

Funnily enough, there was a story today in the WSJ about "a parallel world some portion of people have constructed for themselves"

Why the Tech World Thinks the American Dream Is Dying

https://www.wsj.com/tech/ai/why-the-tech-world-thinks-the-am...

[eudamoniac]

I mostly do normal React crap, ostensibly the easiest thing for these tools to do, and these tools cannot do a considerable portion of my work. Yes I've used the latest model. Yes I've used the latest agentic IDE. Yes I've tweaked my prompts and added repository rule files. Yes I've done this approximately every three months for the last two years. This shit does not work. Nobody ever posts proof of it working well in any <great project>.

I am at the point where if I read something from a software developer like, "these tools can do a considerable portion of my day to day work", I have to just assume that person's day to day work was garbage. And this is not terribly surprising, because a lot of software developers I have personally worked with did produce mostly garbage. Some amount of those people are surely using AI and posting about it, and that explains what we continually see online. Sorry to any offended.

[an0malous]

What is your day to day work?

[widowlark]

Nothing AI can do replaces the important part of my work.

[widowlark]

If your job can be done with AI, why should your employer still pay you? Let us know what the title is so we can work to outmode your career

[ninkendo]

> It has been three years and these tools can do a considerable portion of my day to day work.

Agreed.

> Unfortunately I think that many people’s jobs are essentially in the “Coyote running off a cliff but not realizing it yet” phase or soon to be.

Eh… some people maybe. But history shows nearly every time a tool makes people more efficient, we get more jobs, not less. Jevon’s paradox and all that: https://en.wikipedia.org/wiki/Jevons_paradox

[pixl97]

Right, cars lead to more horses... except it didn't.

[ninkendo]

Of course cars wouldn't lead to more horses, because horses were the thing being replaced. But cars sure as hell lead to a lot more drivers, which is more akin to the analogy.

To take a software engineer for an example, Jevon's paradox would say that since software engineering is now so much easier due to LLM's, the demand will increase due to the reduced cost, which will lead to more software needing created, which paradoxically leads to more software engineers. There's no equivalent of the "horse" in the analogy, because the same people who were coding before ("driving" the horse) will be aided by LLM's in the future ("driving" a car.)

[sodapopcan]

> But history shows nearly every time a tool makes people more efficient, we get more jobs, not less.

I hope so, but you have any ideas what they could be? This time feels different, especially because all the ultra-pro-AI people keep saying that "this time it's different" from a technological revolution. This is aiming to replaces people across many industries whereas historically it has been in smaller increments as new inventions are (more slowly) rolled out.

[kalkin]

> AI is a statistical inference engine. All it can do is predict what word will come next based on all the words that have been typed in the past.

If we keep saying this hard enough over and over, maybe model capabilities will stop advancing.

Hey, there's even a causal story here! A million variations of this cope enter the pretraining data, the model decides the assistant character it's supposed to be playing really is dumb, human triumph follows. It's not _crazier_ than Roko's Basilisk.

[prmoustache]

> AI is a statistical inference engine. All it can do is predict what word will come next based on all the words that have been typed in the past.

Ironically, that is also how humans "think" 99.9% of the time.