I don't understand the people who have the patience to listen to Zitron. It's all one-tone takes with no place for nuance, you kind of know what he's going to say and the fact the mixes genuine criticism with bad faith arguments doesn't help.
> I don't understand the people who have the patience to listen to Zitron. It's all one-tone takes with no place for nuance...
But that's why some people give him the time of day. A lot of people prefer to see things in black and white because it's easier on the brain.
Basically, Ed is doing the easy part (pointing out the malinvestment) and not addressing the harder part, which is to predict what comes next in realistic terms. Because the idea that LLMs are just a $30-40 billion/year TAM and there's going to be an epic implosion that leaves only rubble is not the likely outcome.
It's a bit like the first .com boom. There was a huge amount of malinvestment and the bubble popping was painful, but there really was a business for people to buy stuff online, to consume paid content online, etc. The malinvestment got sorted and a couple decades on, immense wealth has been created.
> A lot of people prefer to see things in black and white because it's easier on the brain.
people also tend to sheild their sacred cows with gestures toward nuance. maybe i'm misreading you, but it sounds like your criticism of him is he has correctly identified a problem and is reporting on it, but he hasn't taken the next step of predicting the next twenty years of consequences. sounds to me like a reporter doing his job well.
> and a couple decades on, immense wealth has been created
For whom? "Everyone". Maybe we can quote some "Everyone is better off now/the global standard of living has increased bullshit". Have you tried to get a job that pays enough to buy the average house in your area?
> It's a bit like the first .com boom. There was a huge amount of malinvestment and the bubble popping was painful
That's the point though, isn't it? How detached from reality must you be to think this is ok?
"Painful" is carrying a lot of weight in that sentence. Painful means people lost their jobs, families broke up, depression, probably suicide, other forms of violence...
There must be a better way to go about these things.
I don't disagree, but there's a big difference between 'this is massively overinvested and valued and that's distorting the market around a useful product' and 'this is all basically a scam with no value to it whatsoever'. For some reason a lot of AI critics seem to be really hard pushing on the latter part despite it being by far the least credible take at this point. It might be overused, it might have some big negative externalities (though these are often overstated), it might have sucked up way more capital than it deserved, but it's also still very useful and valuable for quite a lot of people.
(to me it's a bit like criticising oil companies by claiming that oil doesn't actually produce any useful power after its refined. There's a lot to criticise about them but the fact that their product is very useful is in large part why it's so hard to do something about the rest of the problems they cause)
I won't dismiss the human cost but boom and bust cycles have existed for a very long time. This isn't a new thing today, and it wasn't new then either.
You can view these things 100% cynically, or you can also consider the possibility that markets over (and also under) shooting are also a form of discovery in which the value of new business models, technologies, etc. are determined. While I'm not personally a fan of today's financial engineering and have concerns about direct government investments in particular, net-net it's probably better to live in economies where capital is abundant and malinvestment is possible, than to live economies where the opposite is true.
From the .com days, I personally observed four categories of outcome:
1. Total financial ruin. If we're being honest, this was often the result of individuals who got rich very quickly and spent even more quickly.
2. Survival. Lots of people got laid off and eventually found new jobs when the economy recovered. Their experience was valuable. A decent number of these people went on to ride the recovery boom and are comfortable if not "rich" today.
3. Survival with a story to tell. People who survived and came away with a story to tell ("at one point I was worth millions on paper and 6 months later I was laid off").
4. Huge windfall. Some people made a lot of money and through luck and/or smarts, managed to walk away from the implosion wealth intact.
The truth is that nobody can predict with certainty the future. Markets are the place where humans make bets about the future. Expecting entirely orderly markets in which everything moves at a predictable pace in a predictable way, where no participants win or lose too much, just isn't realistic.
> net-net it's probably better to live in economies where capital is abundant and malinvestment is possible, than to live economies where the opposite is true.
This is a false dichotomy. Not everything needs to exist at the extremes.
> boom and bust cycles have existed for a very long time. This isn't a new thing today, and it wasn't new then either.
Murder and domestic violence have existed for a very long time. Racism existed for a long time. Slavery existed for a long time.
Things can be fixed, or at least improved to reduce the human cost associated with what's essentially a governance choice.
Not sure how you would do things better without a centralized planning committee. Most people believe that bubbles are bad, but what's the alternative? People want to get rich quick.
I find more nuance in his arguments than on the typical AI hyper in yhe HN comment session. I can sort of predict the talking points I find here.
Anyway, I think Zitron is wrong on the usefulness of AI, but I don't need to agree with him on everything. That is the weaker part of his arguments anyway (and, unsurprisingly, where critics here choose to engage with).
He has a very solid point on the viability and economics of AI, and I am still to see facts that contradict his analysis there.
In fact, what spooks me about AI is not that it can replace me. It can't, that much is clear. I am spooked about the probable economic downturn that it will spawn. All the reckless spending is fun until the bill has to be paid.
The TAM figure they arrive at is ~$36 billion by 2030. And for 2026 they claim a TAM of $10.6B
OpenAI alone is rumored to be on track for $30B of revenue in 2026. Add in Anthropic, Google, Microsoft, Meta, and Chinese providers, and the revenue being generated from LLM's in 2026 is plausibly in the range of $50-100B already.
Whatever your thoughts are on the cash burn to get there are irrelevant. There's at least $50B of LLM usage being paid for in 2026. 5x higher than the figure these research report companies are providing.
I think it’s nice to have a voice criticizing the fundraising aspect of these companies. I do think we’ll see at least one of them blow up. The technology is obviously useful though. Hundreds of thousands of developers have already changed the way they were working for decades. Some of the criticisms that the technology doesn’t work at all go a bit too far.
> Hundreds of thousands of developers have already changed the way they were working for decades
I agree on this, but it doesn't mean that there is an automatic benefit on business side ... and business is what is paying our wages & tokens!
We are still in the discovery phase, but we don't know yet if there will be enough return to repay those hundreds of billions already invested and other few trillions that will be invested in the near future.
The header looks not too bad until you realize that 2/3 of the industry revenue is Nvidia ...
If you remove the shovel sellers (Micron/Nvidia/AMD) revenue is left at $183B.
It actually does work for a whole range of things.
Is it a panacea? Of course not, but it is potentially a sustainable business.
Except - open models are barely months away from the same performance and there is no moat, so we will see commodity pricing, and the billions being heaped on the fire currently will probably not see a return, unless someone comes up with a genius trading bot perhaps.
How many days in a row can you do this, and then still be able to get things done when Codex is unavailable or the price has been raised to $UNAFFORDABLE?
I logged into my account today to look at my balance and saw a notice that they're changing pricing to charge more at peak times. I haven't looked into to it more yet but I'm guessing it will still be significantly cheaper than the American models.
When you say "just", which innovation are you referring to? Did they literally just announce something or are you referring to something like v4 flash?
I don't understand the people who have the patience to listen to Zitron. It's all one-tone takes with no place for nuance, you kind of know what he's going to say and the fact the mixes genuine criticism with bad faith arguments doesn't help.
> I don't understand the people who have the patience to listen to Zitron. It's all one-tone takes with no place for nuance...
But that's why some people give him the time of day. A lot of people prefer to see things in black and white because it's easier on the brain.
Basically, Ed is doing the easy part (pointing out the malinvestment) and not addressing the harder part, which is to predict what comes next in realistic terms. Because the idea that LLMs are just a $30-40 billion/year TAM and there's going to be an epic implosion that leaves only rubble is not the likely outcome.
It's a bit like the first .com boom. There was a huge amount of malinvestment and the bubble popping was painful, but there really was a business for people to buy stuff online, to consume paid content online, etc. The malinvestment got sorted and a couple decades on, immense wealth has been created.
> A lot of people prefer to see things in black and white because it's easier on the brain.
people also tend to sheild their sacred cows with gestures toward nuance. maybe i'm misreading you, but it sounds like your criticism of him is he has correctly identified a problem and is reporting on it, but he hasn't taken the next step of predicting the next twenty years of consequences. sounds to me like a reporter doing his job well.
> and a couple decades on, immense wealth has been created
For whom? "Everyone". Maybe we can quote some "Everyone is better off now/the global standard of living has increased bullshit". Have you tried to get a job that pays enough to buy the average house in your area?
You should read his What If We're In a Bubble 3 part series.
> It's a bit like the first .com boom. There was a huge amount of malinvestment and the bubble popping was painful
That's the point though, isn't it? How detached from reality must you be to think this is ok?
"Painful" is carrying a lot of weight in that sentence. Painful means people lost their jobs, families broke up, depression, probably suicide, other forms of violence...
There must be a better way to go about these things.
I don't disagree, but there's a big difference between 'this is massively overinvested and valued and that's distorting the market around a useful product' and 'this is all basically a scam with no value to it whatsoever'. For some reason a lot of AI critics seem to be really hard pushing on the latter part despite it being by far the least credible take at this point. It might be overused, it might have some big negative externalities (though these are often overstated), it might have sucked up way more capital than it deserved, but it's also still very useful and valuable for quite a lot of people.
(to me it's a bit like criticising oil companies by claiming that oil doesn't actually produce any useful power after its refined. There's a lot to criticise about them but the fact that their product is very useful is in large part why it's so hard to do something about the rest of the problems they cause)
I agree with you, for what it's worth.
I won't dismiss the human cost but boom and bust cycles have existed for a very long time. This isn't a new thing today, and it wasn't new then either.
You can view these things 100% cynically, or you can also consider the possibility that markets over (and also under) shooting are also a form of discovery in which the value of new business models, technologies, etc. are determined. While I'm not personally a fan of today's financial engineering and have concerns about direct government investments in particular, net-net it's probably better to live in economies where capital is abundant and malinvestment is possible, than to live economies where the opposite is true.
From the .com days, I personally observed four categories of outcome:
1. Total financial ruin. If we're being honest, this was often the result of individuals who got rich very quickly and spent even more quickly.
2. Survival. Lots of people got laid off and eventually found new jobs when the economy recovered. Their experience was valuable. A decent number of these people went on to ride the recovery boom and are comfortable if not "rich" today.
3. Survival with a story to tell. People who survived and came away with a story to tell ("at one point I was worth millions on paper and 6 months later I was laid off").
4. Huge windfall. Some people made a lot of money and through luck and/or smarts, managed to walk away from the implosion wealth intact.
The truth is that nobody can predict with certainty the future. Markets are the place where humans make bets about the future. Expecting entirely orderly markets in which everything moves at a predictable pace in a predictable way, where no participants win or lose too much, just isn't realistic.
> net-net it's probably better to live in economies where capital is abundant and malinvestment is possible, than to live economies where the opposite is true.
This is a false dichotomy. Not everything needs to exist at the extremes.
> boom and bust cycles have existed for a very long time. This isn't a new thing today, and it wasn't new then either.
Murder and domestic violence have existed for a very long time. Racism existed for a long time. Slavery existed for a long time.
Things can be fixed, or at least improved to reduce the human cost associated with what's essentially a governance choice.
Not sure how you would do things better without a centralized planning committee. Most people believe that bubbles are bad, but what's the alternative? People want to get rich quick.
This is also a false dichotomy. There's a lot in between neolib free markets and centralised planning.
Yeah, I used to be a fan of his critiques, but it's basically the opposite of reading Tyler Cowen's blog now.
I find more nuance in his arguments than on the typical AI hyper in yhe HN comment session. I can sort of predict the talking points I find here.
Anyway, I think Zitron is wrong on the usefulness of AI, but I don't need to agree with him on everything. That is the weaker part of his arguments anyway (and, unsurprisingly, where critics here choose to engage with).
He has a very solid point on the viability and economics of AI, and I am still to see facts that contradict his analysis there.
In fact, what spooks me about AI is not that it can replace me. It can't, that much is clear. I am spooked about the probable economic downturn that it will spawn. All the reckless spending is fun until the bill has to be paid.
[flagged]
Just keep in mind that Zitron is in the business of content creation. It doesn't mean he's wrong[1] but his job isn't making successful predictions.
[1] Actually I think it's right on the overinvestments and ROI claims
We've gone from asking "Can an LLM do this?" to "Is an LLM actually the best tool for this?" That's a healthier conversation.
A big portion of the global economy is currently banking on LLMs being able to do everything.
A big portion of the global economy is pure wishful thinking and hype.
No one wants to admit it for the obvious reason that the ruse is very profitable.
So, the "big portion" is money, not people.
"Is an LLM actually the best tool for this yet?"
I believed that $10-30 billion/year TAM figure is sourced from this research report:
https://www.precedenceresearch.com/large-language-model-mark...
The TAM figure they arrive at is ~$36 billion by 2030. And for 2026 they claim a TAM of $10.6B
OpenAI alone is rumored to be on track for $30B of revenue in 2026. Add in Anthropic, Google, Microsoft, Meta, and Chinese providers, and the revenue being generated from LLM's in 2026 is plausibly in the range of $50-100B already.
Whatever your thoughts are on the cash burn to get there are irrelevant. There's at least $50B of LLM usage being paid for in 2026. 5x higher than the figure these research report companies are providing.
> There's at least $50B of LLM usage being paid for in 2026
Just out of curiosity, where did you get the 50B figure?
Ed Zitron,
He gets things others are missing.
He also misses things others are getting.
I think it’s nice to have a voice criticizing the fundraising aspect of these companies. I do think we’ll see at least one of them blow up. The technology is obviously useful though. Hundreds of thousands of developers have already changed the way they were working for decades. Some of the criticisms that the technology doesn’t work at all go a bit too far.
> Hundreds of thousands of developers have already changed the way they were working for decades
I agree on this, but it doesn't mean that there is an automatic benefit on business side ... and business is what is paying our wages & tokens!
We are still in the discovery phase, but we don't know yet if there will be enough return to repay those hundreds of billions already invested and other few trillions that will be invested in the near future.
Would you be able to quantify how useful it is? Trillions are at stake here, so you’ll need to get specific.
I can't quantify how useful GitHub is as a software developer either. That doesn't prove it's useless, just that it's hard to quantify.
I never said it was useless. But you better be able to quantify the value of something you’re spending trillions of dollars building.
This seems like common sense to me.
Some numbers for the people in the back.
https://isaiprofitable.com/
The header looks not too bad until you realize that 2/3 of the industry revenue is Nvidia ... If you remove the shovel sellers (Micron/Nvidia/AMD) revenue is left at $183B.
Yep you see how deep the scam goes. It’s like in the gold rush if the ONLY people that made money were the one shovel seller.
Some logic for those who don't do numbers:
Token rates need to double in order for the industry to "break even".
In reality, just "breaking even" is not enough. Venture capital expects a sizeable return on their investment. So look for token rates to triple.
In reality, most companies are not at all prepared to feed AI vendors what they need in order to become profitable.
Uber is an early example of what is in store.
https://aimagazine.com/news/why-uber-has-already-burned-thro...
https://www.forbes.com/sites/janakirammsv/2026/05/17/uber-bu...
It actually does work for a whole range of things.
Is it a panacea? Of course not, but it is potentially a sustainable business.
Except - open models are barely months away from the same performance and there is no moat, so we will see commodity pricing, and the billions being heaped on the fire currently will probably not see a return, unless someone comes up with a genius trading bot perhaps.
Folks if you believe this Ed guy, you're soon going be really disappointed.
LLMs are big, Codex app is huge. You can control the browser from within codex app with just natural language.
Tell codex to perform a few actions, you go get coffee and come back to work done. This is real boost in productivity.
How many days in a row can you do this, and then still be able to get things done when Codex is unavailable or the price has been raised to $UNAFFORDABLE?
Meanwhile Deepseek just 6x-ed their inference efficiency at zero quality loss.
I logged into my account today to look at my balance and saw a notice that they're changing pricing to charge more at peak times. I haven't looked into to it more yet but I'm guessing it will still be significantly cheaper than the American models.
When you say "just", which innovation are you referring to? Did they literally just announce something or are you referring to something like v4 flash?
I imagine GP is talking about this - https://news.ycombinator.com/item?id=48696585