I find it difficult to separate this piece’s tone from its content. The tone puts me off and makes it hard for me to judge it on its merits, despite some of the arguments seeming sound and well supported.
It's like someone arguing that cheese isn't real. Yes I can go to the grocery store and take a picture of cheese and show it, but what's the point? They can live in their own world. It doesn't change any of our lives. The world is what it is.
Zitron is begging for a collapse at this point. Yes, his macro analysis correctly identifies a massive financial risk but his incessant pessimism completely misses the incredible ground-level utility that many of us on HN celebrate every day through undeniable, massive productivity gains.
At this point I'm trying to believe there's a middle ground where the level of individual capability this unlocks, leads to major discoveries.
He has recently made the very good point that actually, the FAANG companies are struggling to put any ROI numbers on that incredible ground-level utility.
Uber, for example, is so unclear there is any ROI, they are cutting their exposure pretty radically.
He points out that one single Anthropic customer — a payments provider — accidentally had to pay Anthropic $500M for one month of token spend.
That is half what Apple is reportedly paying Google for the supply side of their entire consumer AI strategy.
Before you spend 20 minutes reading this article, it's worth understanding that the writer has been posting popular but consistently wrong takes for 2+ years (e.g. https://www.wheresyoured.at/peakai/ from March 2024) arguing that AI is failing, is a waste of money, is bad, will never work, etc.
When asked whether it will be possible to fix Sora's videos after they've been generated, Murati said "eventually," and then couched that by saying "that's what we're trying to figure out...how to use this technology as a tool that people can edit and create with." She promised that there would "eventually" be "more steerability, control and accuracy...and reflecting of intent of what you want."
You'll "eventually" be able to add audio to Sora videos, and when asked when Sora's generative videos will be available to the public, she once again said "eventually," and when pushed said that Sora's launch would "definitely be this year, but could be a few months."
Murati, living in a world of "eventuallies," provided no technical insights, no specifics, and very few details.
Some people seem to see the world only through bubbles. But if you look at human history, despite the ups and downs, we have a trajectory; generally speaking, human-created systems evolve toward ever-increasing complexity, impact, and efficiency.
The current wave of AI unlocked language - the tools are now speaking and understanding. This, on its own, is astonishing progress. Language is the foundation of our culture and society; it is the very technology that got us, as a species, to where we are today. To have tools that can understand, manipulate, and produce it is a massive leap forward.
Once you see things that way, it is clear that we are not in a bubble; we are in a transition. Yes, there is tons of hype and over-investment, but the demand is real, and so is the impact. Unless you are deep in the tech and have that structural depth, it is easy to dismiss. This is like the invention of the personal computer, but with 100x the impact and speed.
Ed Zitron speaks to a particular type of angry tech conservative. He’s not speaking truth or exposing anything. He’s the soothing voice the tech nerds of yesterday year are yearning for.
The angry polemic that goes on and on and on with cuss words used liberally is just meant to evoke emotion and cathartic resolution to the type of people mentioned above. Not truth.
The thing is, there are a lot of people that find comfort in what he’s writing - primarily because it’s a coping mechanism against how quickly things are moving and a way to deal with being left behind. When you spend time, years, building institutional knowledge and making a whole identity out of it, you obviously will feel bad with the threat of it being commoditised.
I would write against the content of the article but I find it easier and more illuminating to write what he has said before instead. Then it shows how incorrect the guy has been and with what confidence he keeps speaking with.
Buried lede (if the title is the actual promise), the sources don't seem to back the title either. Someone with more patience can correct me if I accidentally missed a bombshell anyway.
Edit:
> If you’re wondering what the story is, [...] I expect it to be out in the next two weeks [...] I can guarantee you it’ll be worth it, and you’ll be stunned by what I report.
Ok, this takes clickbait to new lows. The headline is trying to sell the teaser here, with very limited meat in the middle of the sandwich.
> the sources don't seem to back the title either.
This is nothing new for Zitron. The last article of his I saw:
> I also severely doubt that Anthropic managed to make the cost of running its services profitable in the space of six months.
> [Per The Information in January], Anthropic missed on its gross margin projections, saying that its inference costs were 23% higher than the company had anticipated.
> How did Anthropic, which faced a massive influx of new business to the point that Anthropic was forced to buy more compute from Elon Musk, magically become profitable? Other than that discount, of course.
If you follow the link to The Information, you’ll see that it’s a paid article with the headline “Anthropic Lowers Gross Margin Projection as Revenue Skyrockets”. But what happens when you actually read the article?
> Anthropic last month projected it would generate a 40% gross profit margin from selling AI to businesses and application developers in 2025, according to two people with knowledge of its financials. That margin was 10 percentage points lower than its earlier optimistic expectations, though it’s still a big improvement from the year before.
So, according to Zitron’s own source, Anthropic are actually earning 40% gross profit margin on inference, and that is a dramatic jump upwards! This totally contradicts his position that it’s an implausible “swindle” for Anthropic to claim profitability. He’s counting on the fact that most of his readers don’t subscribe to The Information and will only see the headline, or that they will just see a citation and trust that it backs him up without checking.
Whenever I read these kind of articles about AI financials, I'm reminded of identical screeds I read about Uber a few years ago. They were angrily insistent that Uber was a scam company run by criminals and charlatans and could never, ever become profitable or make money for its investors. It was a house of cards that would come crashing down sooner or later, and take everyone's money with it. Now it's 2026. Uber still exists, has revenues of $50bn and is apparently a highly profitable business. I don't know if the original investors have made their money back yet, but Uber certainly hasn't collapsed.
Maybe AI is different. Certainly, the level scale of investment is on a different order of magnitude. But I'm wary of believing anything about the financial impossibility of AI being sustainable when I've seen such similarly confident arguments proved wrong in the past.
Funny thing, the uber's investor results from last year only mentions "profit" once, in a motivating paragraph where they say they will be great.
But it's famous for having collapsed after their IPO. It took 4 years to get back at the same nominal valuation (not inflation corrected), and after all the 2020s inflation it is still at 2x the initial price.
Ed is an interesting character. His financial analysis of the AI industry makes logical sense to me (though I am not knowledgeable enough to actually know if it is correct.) However, he seems to be so angry at AI in general, that he misses the obvious areas where LLMs are actually changing the State of the Art.
Coding seems to be one of the core use-cases for LLMs (as Simon Willison pointed out recently) and even if that's the only real use-case for LLMs, they're wildly useful. I do understand that useful != profitable and that's where I think Ed has a real point: until inference becomes much cheaper these companies cannot be profitable. Some mega-players will pay the API token price, but most will not.
It seems that a certain kind of person cannot separate the following things:
1) I dislike AI as a technology
2) I dislike the people and companies that profit from AI
3) I think AI is useless
These are three completely separate positions to have. You can think AI is incredibly useful and also dislike it because it will, for example, reduce your relative status in society. You can love the tech but think that Sam Altman is a dishonest person, etc. But for some reason, most anti-AI commentators feel compelled to present all three arguments.
Which is even sillier when you think about it, because if it's useless, then you really shouldn't care: the markets will eventually find out that it's useless, and everything will go back to normal, and the people you don't like will have lost money, so there's no point in being outraged. Of course, I don't really believe that they think it's useless. I do think they're worried about what it'll do to their prestige, though, and they're just hoping beyond hope that somehow everyone will one day "wake up" and share their belief that LLMs are just "stochastic parrots" with no utility, despite the fact that people are using them every day and can watch in real time as they improve.
So here's the thing. I am not generally an angry person. But Ed's writing really resonates with me, because for the last four years these people have been making a strategy of scaring the shit out of us while trying to ruin something I genuinely love (coding), while simultaneously fucking up the economy and multiple industries and turning the internet into slop. I very badly want more people to call these guys "chucklefucks" or whatever innovative ways he comes up with to insult them because they deserve far more public ridicule and disdain than the (captured, useless) media is giving them.
So far the data for productivity in coding is.. sus. The productivity gains outside of toy projects are mostly anecdotal and it's hard to tell if those accounts are even real humans or just astroturfing and bots. Almost every programmer I know personally has a pretty measured opinion on where these things are useful and where they're not. The breathless hype seems mostly from non coders.
I do think Ed in intentionally ignorant of the capabilities of LLMs. But I also don't know that I would classify LLMs as 'wildly useful' for coding. Most productivity gains seem to be hallucinated, and while it's too early to make any claims on long term outcomes, there are plenty of studies indicating they might be even more negative.
There are definitely use cases for LLMs in coding. And at times, they can be wildly useful. But I feel like the industry atm wildly overestimates their broader/long term utility.
Anecdotally, I have not seen an explosion in quality/bespoke software since LLMs. In fact I've noticed the opposite to quite the extreme. Not only are new products worse in quality, but the quality of existing products is falling off a cliff.
> I do understand that useful != profitable and that's where I think Ed has a real point: until inference becomes much cheaper these companies cannot be profitable.
If inference becomes cheaper, it becomes cheaper for everyone.
I would say his overall negative outlook is a well needed counterbalance to the completely delusional hype one is exposed to on a daily basis. The truth will then probably land somewhere in the middle.
His rhetoric is a bit obsessive and frankly biased against AI.
That said, I think his voice is useful as a counter to the mainstream opinion.
Given the amount of investments, approaching AI from the angle of economics seems correct.
We all have some level of personal experience using AI/LLMs, both chatbots and coding tools, and I personally enjoy using them, but I am sure this experience is relevant in this discussion.
I also enjoy luxury hotels, gourmet food, jet skis and helicopters, but this is not something I indulge in often because of the cost-utility ratio.
The real cost of AI may or may not be lower than its utility. The bet is that utility is increasing while cost is falling.
Zitron is in the business of content creation and not successful predictions. It doesn't matter how many times he (and several others around) will say the end is here, they have to be right only once.
BTW, one thing for sure he is right about are the economics, as of today there is no way these massive investments are gone be paid.
> This is a hysterical era perpetuated by liars, cowards, imbeciles, craven boosters and the easily-fooled. Those excited about generative AI are either the victim or the perpetrator of a con centered around a technology to ingratiate at the highest cost possible.
Who writes like this? When you lead with "everyone who doesn't agree with me is a lying cheat coward imbecile" I think we should just turn the volume down on you to zero.
This is breakdown in dialog. If it leads like this then I I don't care how accurate the critical analysis to follow is. I didn't read the rest of the article and don't think anyone else should either out of sheer disdain for this argumentation style.
AI has been slowing down relatively, considering its trajectory over the past 20-30 years. For one, even if LLM may have plateaud in terms of intelligence-parameters ratio, research is on-going on new frontiers for ML, including (but not limited to) world models. Other research directions are studying backpropagation and its physical analogies, such as equilibrium of chaotic states.
In addition, there's a lot of research on the hardware angle and actual prototypes are already being built such as AI-on-chip Cerebra and Taalas for one.
>LLM-type AI exacts huge costs because it is terrible at reporting "I don't know". When it doesn't know, it generates noise and polishes it.
>If a "confidence too low for output" signal could be extracted, this whole technology would be a lot more useful
Anthropic's interpretability research explored this topic a bit in 2025. Apparently, the signal is extractable:
It turns out that, in Claude, refusal to answer is the default behavior: we find a circuit that is "on" by default and that causes the model to state that it has insufficient information to answer any given question. However, when the model is asked about something it knows well—say, the basketball player Michael Jordan—a competing feature representing "known entities" activates and inhibits this default circuit (see also this recent paper for related findings). This allows Claude to answer the question when it knows the answer. In contrast, when asked about an unknown entity ("Michael Batkin"), it declines to answer.
By intervening in the model and activating the "known answer" features (or inhibiting the "unknown name" or "can’t answer" features), we’re able to cause the model to hallucinate (quite consistently!) that Michael Batkin plays chess.
Sometimes, this sort of “misfire” of the “known answer” circuit happens naturally, without us intervening, resulting in a hallucination. In our paper, we show that such misfires can occur when Claude recognizes a name but doesn't know anything else about that person. In cases like this, the “known entity” feature might still activate, and then suppress the default "don't know" feature—in this case incorrectly. Once the model has decided that it needs to answer the question, it proceeds to confabulate: to generate a plausible—but unfortunately untrue—response.
That's real progress. The paper behind it is [1] They try to extract "attribution graphs" to understand why the LLM produced some result. It's encouraging to see more work on what's going on inside. They obtained insight into a specific kind of hallucination: not finding a specific fact. "We uncover circuit mechanisms that allow the model to distinguish between familiar and unfamiliar entities, which determine whether it elects to answer a factual question or profess ignorance. “Misfires” of this circuit can cause hallucinations." That should be tested on queries which resulted in making up legal citations.
I hadn't heard of the TMobile and Brex spend caps, only knew about Uber's because it went viral last week. I expect we'll see more of that now that everyone is paying per token, and it sort of feels like you cannot both have spending caps and require extensive AI usage for performance reviews -- I wonder that will shake out in the end?
Anecdotally, $dayJob consumes Anthropic models via Azure subscriptions which lend themselves pretty neatly to the spending dashboards Ed mentions are missing from Anthropic themselves, and finance seems ok with the current usage, but there's no real hard incentives internally for AI usage either.
I guess Q3-4 are going to be interesting to see where this all goes.
Ed's argument for why "AI is slowing down" rests on company spending caps, in particular the Uber $1,500/engineer/tool cap.
I interpret the exact same evidence in the opposite direction. A year ago the idea that a company would spend $1,500/month/employee on AI tooling felt absurd, what could people possible want to do with AI that would cost that much?
Then coding agents (and, increasingly, general purpose agents) happened and suddenly companies are having to set limits because otherwise the demand from their employees is too high.
The TAM of these AI companies just leapt up to $1,500/knowledge-worker/month, how is that "slowing down"?
I don't really understand how engineers at Uber are hitting $1500/month. Are they forced to pay API costs?
My company provides employees with API keys and soft limits, but as soon as you approach ~$400/month they ask that you get a Claude/Codex Max subscription instead. Curious if it's not the same case at Uber.
>but as soon as you approach ~$400/month they ask that you get a Claude/Codex Max subscription instead
While this seems to be allowed because the current ToS don't seem to explicitly forbid it, I'd be surprised if this loophole stayed open for long... Why would they even distinguish between business and (much cheaper) individual plans if companies can work around it by telling employees to just pay for the latter themselves?
Enterprise agreements are billed at the API rate (albeit sometimes with committed spend discounts). There is no equivalent of the Max subscription in this context.
Saying its going to be 1500 a month across the board is highly speculative. How many companies can even demonstrate that they're getting more than 18000 dollars a year in surplus value per employee by using this tech?
I don't know that Uber is generalizable. I also think specific company dynamics matter (are your execs encouraging tokenmaxxing, have you IPO'd or are you playing w/ VC money all of whom are encouraging tokenmaxxing, etc).
Another way you could take it is, avg Uber salary is what $300k/yr? Does Uber think LLMs make their engineers at most .5% more productive?
156 comments
[ 0.16 ms ] story [ 102 ms ] threadIt's like someone arguing that cheese isn't real. Yes I can go to the grocery store and take a picture of cheese and show it, but what's the point? They can live in their own world. It doesn't change any of our lives. The world is what it is.
At this point I'm trying to believe there's a middle ground where the level of individual capability this unlocks, leads to major discoveries.
Uber, for example, is so unclear there is any ROI, they are cutting their exposure pretty radically.
He points out that one single Anthropic customer — a payments provider — accidentally had to pay Anthropic $500M for one month of token spend.
That is half what Apple is reportedly paying Google for the supply side of their entire consumer AI strategy.
Tim Lee also pointed out that when Ed has posted details on some of his analysis, they have had some....oddities: https://x.com/binarybits/status/2034377838883700953
Please speculate on why OpenAI wouldn't just leave it up (whether or not they were able to improve it).
The current wave of AI unlocked language - the tools are now speaking and understanding. This, on its own, is astonishing progress. Language is the foundation of our culture and society; it is the very technology that got us, as a species, to where we are today. To have tools that can understand, manipulate, and produce it is a massive leap forward.
Once you see things that way, it is clear that we are not in a bubble; we are in a transition. Yes, there is tons of hype and over-investment, but the demand is real, and so is the impact. Unless you are deep in the tech and have that structural depth, it is easy to dismiss. This is like the invention of the personal computer, but with 100x the impact and speed.
The angry polemic that goes on and on and on with cuss words used liberally is just meant to evoke emotion and cathartic resolution to the type of people mentioned above. Not truth.
The thing is, there are a lot of people that find comfort in what he’s writing - primarily because it’s a coping mechanism against how quickly things are moving and a way to deal with being left behind. When you spend time, years, building institutional knowledge and making a whole identity out of it, you obviously will feel bad with the threat of it being commoditised.
I would write against the content of the article but I find it easier and more illuminating to write what he has said before instead. Then it shows how incorrect the guy has been and with what confidence he keeps speaking with.
Edit:
> If you’re wondering what the story is, [...] I expect it to be out in the next two weeks [...] I can guarantee you it’ll be worth it, and you’ll be stunned by what I report.
Ok, this takes clickbait to new lows. The headline is trying to sell the teaser here, with very limited meat in the middle of the sandwich.
This is nothing new for Zitron. The last article of his I saw:
> I also severely doubt that Anthropic managed to make the cost of running its services profitable in the space of six months.
> [Per The Information in January], Anthropic missed on its gross margin projections, saying that its inference costs were 23% higher than the company had anticipated.
> How did Anthropic, which faced a massive influx of new business to the point that Anthropic was forced to buy more compute from Elon Musk, magically become profitable? Other than that discount, of course.
If you follow the link to The Information, you’ll see that it’s a paid article with the headline “Anthropic Lowers Gross Margin Projection as Revenue Skyrockets”. But what happens when you actually read the article?
> Anthropic last month projected it would generate a 40% gross profit margin from selling AI to businesses and application developers in 2025, according to two people with knowledge of its financials. That margin was 10 percentage points lower than its earlier optimistic expectations, though it’s still a big improvement from the year before.
— https://archive.is/aKFYZ
So, according to Zitron’s own source, Anthropic are actually earning 40% gross profit margin on inference, and that is a dramatic jump upwards! This totally contradicts his position that it’s an implausible “swindle” for Anthropic to claim profitability. He’s counting on the fact that most of his readers don’t subscribe to The Information and will only see the headline, or that they will just see a citation and trust that it backs him up without checking.
Maybe AI is different. Certainly, the level scale of investment is on a different order of magnitude. But I'm wary of believing anything about the financial impossibility of AI being sustainable when I've seen such similarly confident arguments proved wrong in the past.
But it's famous for having collapsed after their IPO. It took 4 years to get back at the same nominal valuation (not inflation corrected), and after all the 2020s inflation it is still at 2x the initial price.
Coding seems to be one of the core use-cases for LLMs (as Simon Willison pointed out recently) and even if that's the only real use-case for LLMs, they're wildly useful. I do understand that useful != profitable and that's where I think Ed has a real point: until inference becomes much cheaper these companies cannot be profitable. Some mega-players will pay the API token price, but most will not.
These are three completely separate positions to have. You can think AI is incredibly useful and also dislike it because it will, for example, reduce your relative status in society. You can love the tech but think that Sam Altman is a dishonest person, etc. But for some reason, most anti-AI commentators feel compelled to present all three arguments.
Which is even sillier when you think about it, because if it's useless, then you really shouldn't care: the markets will eventually find out that it's useless, and everything will go back to normal, and the people you don't like will have lost money, so there's no point in being outraged. Of course, I don't really believe that they think it's useless. I do think they're worried about what it'll do to their prestige, though, and they're just hoping beyond hope that somehow everyone will one day "wake up" and share their belief that LLMs are just "stochastic parrots" with no utility, despite the fact that people are using them every day and can watch in real time as they improve.
So far the data for productivity in coding is.. sus. The productivity gains outside of toy projects are mostly anecdotal and it's hard to tell if those accounts are even real humans or just astroturfing and bots. Almost every programmer I know personally has a pretty measured opinion on where these things are useful and where they're not. The breathless hype seems mostly from non coders.
There are definitely use cases for LLMs in coding. And at times, they can be wildly useful. But I feel like the industry atm wildly overestimates their broader/long term utility.
Anecdotally, I have not seen an explosion in quality/bespoke software since LLMs. In fact I've noticed the opposite to quite the extreme. Not only are new products worse in quality, but the quality of existing products is falling off a cliff.
If inference becomes cheaper, it becomes cheaper for everyone.
That said, I think his voice is useful as a counter to the mainstream opinion.
Given the amount of investments, approaching AI from the angle of economics seems correct.
We all have some level of personal experience using AI/LLMs, both chatbots and coding tools, and I personally enjoy using them, but I am sure this experience is relevant in this discussion.
I also enjoy luxury hotels, gourmet food, jet skis and helicopters, but this is not something I indulge in often because of the cost-utility ratio.
The real cost of AI may or may not be lower than its utility. The bet is that utility is increasing while cost is falling.
BTW, one thing for sure he is right about are the economics, as of today there is no way these massive investments are gone be paid.
Who writes like this? When you lead with "everyone who doesn't agree with me is a lying cheat coward imbecile" I think we should just turn the volume down on you to zero.
This is breakdown in dialog. If it leads like this then I I don't care how accurate the critical analysis to follow is. I didn't read the rest of the article and don't think anyone else should either out of sheer disdain for this argumentation style.
In addition, there's a lot of research on the hardware angle and actual prototypes are already being built such as AI-on-chip Cerebra and Taalas for one.
>If a "confidence too low for output" signal could be extracted, this whole technology would be a lot more useful
Anthropic's interpretability research explored this topic a bit in 2025. Apparently, the signal is extractable:
https://www.anthropic.com/research/tracing-thoughts-language...[1] https://transformer-circuits.pub/2025/attribution-graphs/met...
Anecdotally, $dayJob consumes Anthropic models via Azure subscriptions which lend themselves pretty neatly to the spending dashboards Ed mentions are missing from Anthropic themselves, and finance seems ok with the current usage, but there's no real hard incentives internally for AI usage either.
I guess Q3-4 are going to be interesting to see where this all goes.
I interpret the exact same evidence in the opposite direction. A year ago the idea that a company would spend $1,500/month/employee on AI tooling felt absurd, what could people possible want to do with AI that would cost that much?
Then coding agents (and, increasingly, general purpose agents) happened and suddenly companies are having to set limits because otherwise the demand from their employees is too high.
The TAM of these AI companies just leapt up to $1,500/knowledge-worker/month, how is that "slowing down"?
My company provides employees with API keys and soft limits, but as soon as you approach ~$400/month they ask that you get a Claude/Codex Max subscription instead. Curious if it's not the same case at Uber.
While this seems to be allowed because the current ToS don't seem to explicitly forbid it, I'd be surprised if this loophole stayed open for long... Why would they even distinguish between business and (much cheaper) individual plans if companies can work around it by telling employees to just pay for the latter themselves?
Another way you could take it is, avg Uber salary is what $300k/yr? Does Uber think LLMs make their engineers at most .5% more productive?