Disclaimer: Anything can happen at anytime in the market; I don’t give stock picks, and as the saying goes, the market can remain irrational longer than you can remain solvent.
As I told someone at work yesterday, they have no moat and no business model, but they do have alllll the money.
As Karen Hao pointed out, it was in their interests to believe that scale is all you need. We need all the money to build AI god, and everyone who believes us can take a cut, hey Goldman 👋 Hey VCs 👋
I hope it is slow instead of a crash, maybe it will hit us all with less velocity.
It seems like, while insisting that we must do this before China, they just handed all our IP to China via their lack of moat. Sigh.
Question: Who is responsible for moat-building. Seems to me Dario could have been more circumspect if he really was worried about the dangerous potential of Fable. Was he using it as a club to destroy his Western competitors, as some argue, or did the government, rather good at digging ditches, if not moats, drop the ball.
There was never a moat, at least not without Demis Hassabis’s to-be-named 1-2 additional breakthroughs. OpenAI got the idea for their model from a public paper from Google. Anthropic could just swipe the same data as OpenAI and replicate the work. All the companies trade engineers like Pokemon cards.
A this point. Research are still public which allows fast feedback and crowd sourced innovation. If DNA was kept as trade secret, we’d be decades behind in terms of pharmaceutical innovation.
Any of the businesses involved in the ML & LLM investment bubble (eg Micron) with better earnings, where they're just selling hardware to intermediaries, but there's no end user demand to ever actually pay for it...
Are just making the investment bubble and its consequences worse and worse... More losses for someone coming in future...
This insanity has carried on for 3+ years already, so you would be a fool to guess exactly when it'll end.
As for the bubble dying slowly, once again: THE IRAN WAR.
You can NOT build a bunch of data centers valued at hundreds of billions and then RUN them with NO ENERGY.
People keep forgetting that the real impact of the loss of energy over the past three months has NOT HIT YET. It won't hit until sometime between now and September. The oil market is being manipulated by draw downs from the National Reserve, which are running out in the next month or so.
So the stock market collapse is going to be FAST and BRUTAL by end of the year.
For those who think "well, the Strait is open right now", barely ten percent of the previous volume of ships transiting the Strait is happening now. And that could be shut down again at any moment based on what happens in Lebanon or at the negotiating table. A few more bizarre threats by Trump towards Iran and the negotiations stall and the Strait is closed again. This could go on for months.
The probability of a final deal between the US and Iran is minimal, if not zero. Negotiations will take months even assuming a resumption of the kinetic hot war doesn't occur.
Oil experts predict the price of oil will never fall back to where it was in February and is likely to remain at least $100 a barrel for the future.
Bottom line: You CAN NOT consider the future of the AI bubble without taking into consideration the Iran war.
Because a recession due to the oil shock is already locked in. Look it up.
And what does the stock market do in a recession?
Does it EVER go up? Or even remain stable?
How about a Depression - which is also being predicted, if with less certainty, by economists?
And if you cite the fact that stock prices increased during 5 of 11 recession, first, that's less than half. Second, most of those recessions were not the result of oil shocks. The 1970s one was, and the oil embargo only lasted six months at a time when less oil was transiting the Strait than now and it took a decade for the economy to recover.
This recession is predicted to be worse than the 1970s one or the 2008 one.
From Investopedia:
During the recession phase of the business cycle, income and employment decline; stock prices fall as companies struggle to sustain profitability.
A sign that the economy has entered the trough phase of the business cycle is when stock prices increase after a significant decline.
And when you add in that the AI Bubble involves massively overvalued stock, well...
I see zero reason to not be confident of a major collapse of AI stocks, if not the overall stock market.
Oil Shock: What History Says About the Stock Market and Rising Energy Prices
As cynical as it sounds, I believe SpaceX was rushed to IPO so Musk could nab the “first trillionaire” badge. They had to know the market wouldn’t respond well with the greater transparency. They waste billions that could feed the hungry, clothe and house the poor, and generate good jobs for people who need and want them. They changed the rules protecting investors from IPO volatility just before rushing over the line. Bah.
As @Ed Zitron says, we’re too far into generative AI to still be debating ROI. We know that the leaders at these companies may as well be televangelists, selling their wares while lying endlessly about the performance. “It’s too dangerous to release.” “Let’s release it.” “It blackmailed me.” “It solved those exact math problems.”
If Altman and Amodei and Musk lie about anything (and we have enormous evidence that they do), why would they ever tell us the truth about what’s really happening? There is too much capital and fame at stake, even if they corrupt the entire financial system.
The tech media abets this sham by lapping up the “AI is the messiah” nonsense. Even Bezos saying, “AGI will solve all our problems,” demonstrates a fundamental misunderstanding of the technology. LLMs won’t ever lead to AGI, and we simply cannot assert that AGI won’t destroy us.
Admittedly, I returned to school (Georgia Tech computer science) because I believed in the power of machine learning and what I thought was AI. I read Kurzweil, ignoring the nonsense and lifting the resonant phrases. I always looked to the future for the brighter day rather than remaining a hostage to the past. But learning about the technology changed my mind.
I decried deep learning as just a new label to place on an existing method. It was and is a good engineering tool, but we have reason to believe there are much more efficient ones. The problem was never with the output—it was the way it was marketed and sold. AI! AI! AI! We know the human brain doesn’t operate like a DNN, with cleverly manufactured changes capable of fooling the DNN object recognizer while a human can’t see any difference.
Watson had access to all the world’s knowledge, but it broke down on Jeopardy several times (they edited together as if Watson really operated in real time.)
I thought we were wasting valuable time and energy when LLMs hit the internet. People found them interesting, occasionally useful. But did they really help anyone?
Claude is the only generative AI product I’ve ever tried, and only for maybe fifteen minutes when I understood that the presentation of it was more impressive than the plaques-and-tangles-riddled code.
Even Microsoft tries to force Copilot on us, and I can’t even hide the goddamned icon so it stops popping up. They’ve repackaged older tools in Office to appear like Copilot, ultimately trying to say that it’s all AI somehow.
The real grift is this—you may think that the product is a system capable of AI. The real product is you and your user data. If you use Claude, Anthropic benefits from what you do. ChatGPT leverages your data similarly. You pay them money to help them make more money.
Except the tokens told the truth—they are massively burning cash to make us believe these tools have ROI. They don’t. No one would pay for a service to replace two coders that costs more than the sum of the two, plus all the time in the world for you to debug and refine. And you wind up with the bill.
Grab your virtual pitchforks and torches, friends!
That's just more hype masquerading as careful, circumspective hedging. What's the advance? Why are data centers stalling in construction? Why do the spokespersons for these companies lie to us? Why are investors taking pauses? Can you explain every promise and its immediate shattering with something more concrete than that? The hurry-up-and-wait isn't helping anyone.
It's because said promises are as soulless and empty as the miscreants making them. They torch cash without any way of recovering it. Think about it--there's only so much demand for anything in this world, and for any one company to succeed, they would require almost more resources than have been invested. Their promises led to hundreds of thousands of workers to lose their jobs, then they discover that the grand output is nearly worthless.
The internet and the computer had ROI early on. There weren't high profile illusionists trying to persuade the Pentagon that these things were valuable--they could determine the military application immediately. By contrast, this technology and the gratuitous subsumption of all other tools into the bland bucket "AI" is just bad business for the rest of us.
AGI is a fairy tale--but it's been one that predates the industrial revolution and the founding. Philosophers grappled with the problem of mocking our actions with automata, statues, and so on. It's a fairy tale we all wish could be true, but the Manhattan project every month should have demonstrated verifiable, reproducible results that these systems can operate as advertised.
Economies and markets are about decisions--we fund X so we can't fund Y. There are plenty of approaches falling outside the OpenAI/Anthropic umbrella, but with all the world's money flowing into them, SpaceX, and the other tech giants just starves the rest of us.
Your emotions appear to be getting the most of you. The internet and the PC are odd examples to pick, since they’re the textbook cases of delayed ROI — the dot-com bust torched more capital faster than anything happening now, and the productivity gains from computing didn’t show up in the data for over a decade. “Payoff isn’t instant” describes every general-purpose technology, including the two you’re holding up as counterexamples. Separately, “they torch cash” lumps the labs in with the chipmakers — but Nvidia and TSMC are paid whether or not OpenAI ever clears its cost of capital. Those are different bets with different risk.
First off, I don't think you really are qualified to describe my state of mind. You cannot separate emotion from argument in any human being--you responding to me begins with emotion. I'm married to a psychiatrist.
Second, no one was flag waving, signaling, and promising the stars and moon with respect to the two examples I gave. They were implemented specifically to address problems the Pentagon wanted solved. The modern internet and PC aren't what I'm focusing on.
Further, you're not addressing the actual point--they lie to us just about weekly and daily about the ROI and when it will happen. There isn't a specific problem they're solving--it's just "everything is AI and it can do everything."
It's not clear what any objective is, beyond vague promises of the world instantly being wonderful.
They are bankrupting the global economy by subsuming other investments we badly need elsewhere.
And I've worked on this tech in my career (Microsoft, Amazon, Uber). I studied statistical machine learning at Georgia Tech and earned a graduate degree there.
I know what I’m talking about. But you don’t have to have any of that to know that the numbers make no sense. You should listen to Ed Zitron.
Some cities now are banning the building of AI data centers.
If not already started, it's time for the U. S. Government to issue licenses for the building of data centers with some strict approval criteria, including geographical boundaries related to and supported/or/not by surrounding towns, cities, and National parks, etc. and the use of resources like water and electricity, and the pre-clean up criteria for all sorts of pollution and influences on animal habitat.
None of this counts . . . because they got ahead of their skis and, like Trump, just went ahead and did what they wanted to do without giving a hoot about their responsibility for anything or anyone.
The fizzle hits companies that chased adoption metrics instead of outcomes first. When the stock narrative flips, 60 trillion tokens a month stops being a growth slide and becomes the bill nobody budgeted for.
Almost every post from Gary contains a substantial amount of self-congratulation about being right about some prediction or other. This is really off-putting and so, so tiresome. It really gets in the way of your message. Please concentrate on the news and analysis rather than on being your own biggest promoter. That guy at the party who talks that way is a bore and it's no different on Substack, no matter how smart you are.
Maybe he should prefix all these statements with something like 'this thing has been public knowledge since at least xxxx; here's an article from then'...? Would that make it better? He's scattering receipts in these posts so the latecomers can catch up, and hopefully understand that his opinions are based on protracted, well-informed study and independent thought, not a bowling score app whipped up while walking the dog. Maybe he's wrong, but he's done the work and he has a lot of receipts. You have no idea what a professional wilderness is until you become a contrarian... look at mainstream macroeconomics for another example of what can only be called professional shunning.
With the SV set value = big money and the power it generates. For most of us value = quality of life, civil rights, and constructive human interactions for all, which in turn promote useful innovation, creativity and economic prosperity.
LLMs captured the public’s and investors’ imaginations thanks to OpenAI’s decision to release ChatGPT to the public FOR FREE (very important detail) in 2022. Maybe because it’s a language model that can, at first glance, convincingly emulate intelligence and empathy, journalists, which deal in words, fell for it too (still, shame on Thomas Friedman for his NYT opinion piece on “Anthropic’s Restraint” last April, but he’s not alone).
The fact remains that, after almost 4 years of a lot of hype and money spent, no amount of ad hoc harnesses has been able to turn these probabilistic models into reliable agents, despite Dario Amodei’s fantastic assertions. To make matters worse, they also turned out to be hugely expensive to use.
As to OpenAI’s alleged USD 5.7 bn in Sales and Marketing costs in 2025, investigative journalist Ed Zitron claims (convincingly IMHO) that this figure probably refers to free token credits given away to Microsoft and other users and should have been booked as cost of sales instead.
Just curious: Are you actually claiming a trademark on "Generative AI Fizzle", as the TM superscript would imply, or is this a bit of Gary's Attempted Humor (TM)? :-)
One of the wonderful things a writer does, and an LLM does not, is to realize that Elias Thorne the lighthouse-keeper does not need a cameo in every goddamn story they write.
The bump of SPCX was probably for a large part asset managers investing in NASDAQ 100 and that one FTSE index, that Musk was able to worm his way into (luckily S&P resisted) Those asset managers are legally/contractually _required_ to rebalance their portfolios before the beginning of July, they have to buy, if they like it or not. They don’t want all of that purchasing to happen just before their deadline, because of risk management.
As I told someone at work yesterday, they have no moat and no business model, but they do have alllll the money.
As Karen Hao pointed out, it was in their interests to believe that scale is all you need. We need all the money to build AI god, and everyone who believes us can take a cut, hey Goldman 👋 Hey VCs 👋
I hope it is slow instead of a crash, maybe it will hit us all with less velocity.
It seems like, while insisting that we must do this before China, they just handed all our IP to China via their lack of moat. Sigh.
Question: Who is responsible for moat-building. Seems to me Dario could have been more circumspect if he really was worried about the dangerous potential of Fable. Was he using it as a club to destroy his Western competitors, as some argue, or did the government, rather good at digging ditches, if not moats, drop the ball.
There was never a moat, at least not without Demis Hassabis’s to-be-named 1-2 additional breakthroughs. OpenAI got the idea for their model from a public paper from Google. Anthropic could just swipe the same data as OpenAI and replicate the work. All the companies trade engineers like Pokemon cards.
No moat but there is a swamp with AI-ligators
Pokémon cards ! 😁😂
A this point. Research are still public which allows fast feedback and crowd sourced innovation. If DNA was kept as trade secret, we’d be decades behind in terms of pharmaceutical innovation.
Sings an old Elton John song; I’m still holding yeah yeah yeah I’m still holding 😂
We'll see I guess, markets are unpredictable.
Any of the businesses involved in the ML & LLM investment bubble (eg Micron) with better earnings, where they're just selling hardware to intermediaries, but there's no end user demand to ever actually pay for it...
Are just making the investment bubble and its consequences worse and worse... More losses for someone coming in future...
This insanity has carried on for 3+ years already, so you would be a fool to guess exactly when it'll end.
As for the bubble dying slowly, once again: THE IRAN WAR.
You can NOT build a bunch of data centers valued at hundreds of billions and then RUN them with NO ENERGY.
People keep forgetting that the real impact of the loss of energy over the past three months has NOT HIT YET. It won't hit until sometime between now and September. The oil market is being manipulated by draw downs from the National Reserve, which are running out in the next month or so.
So the stock market collapse is going to be FAST and BRUTAL by end of the year.
For those who think "well, the Strait is open right now", barely ten percent of the previous volume of ships transiting the Strait is happening now. And that could be shut down again at any moment based on what happens in Lebanon or at the negotiating table. A few more bizarre threats by Trump towards Iran and the negotiations stall and the Strait is closed again. This could go on for months.
The probability of a final deal between the US and Iran is minimal, if not zero. Negotiations will take months even assuming a resumption of the kinetic hot war doesn't occur.
Oil experts predict the price of oil will never fall back to where it was in February and is likely to remain at least $100 a barrel for the future.
Bottom line: You CAN NOT consider the future of the AI bubble without taking into consideration the Iran war.
“So the stock market collapse is going to be FAST and BRUTAL by end of the year.”
Another definitive prediction about something that actually is an uncertain outcome. No better than Cathie Wood’s TSLA stock price nonsense.
Why do people do this? Why not just say “it seems to me a reasonable probability that… “?
Because a recession due to the oil shock is already locked in. Look it up.
And what does the stock market do in a recession?
Does it EVER go up? Or even remain stable?
How about a Depression - which is also being predicted, if with less certainty, by economists?
And if you cite the fact that stock prices increased during 5 of 11 recession, first, that's less than half. Second, most of those recessions were not the result of oil shocks. The 1970s one was, and the oil embargo only lasted six months at a time when less oil was transiting the Strait than now and it took a decade for the economy to recover.
This recession is predicted to be worse than the 1970s one or the 2008 one.
From Investopedia:
During the recession phase of the business cycle, income and employment decline; stock prices fall as companies struggle to sustain profitability.
A sign that the economy has entered the trough phase of the business cycle is when stock prices increase after a significant decline.
And when you add in that the AI Bubble involves massively overvalued stock, well...
I see zero reason to not be confident of a major collapse of AI stocks, if not the overall stock market.
Oil Shock: What History Says About the Stock Market and Rising Energy Prices
https://finance.yahoo.com/sectors/energy/articles/oil-shock-history-says-stock-033500251.html
I’d add the catastrophic costs of the super El Niño if it arrives (high probability).
As cynical as it sounds, I believe SpaceX was rushed to IPO so Musk could nab the “first trillionaire” badge. They had to know the market wouldn’t respond well with the greater transparency. They waste billions that could feed the hungry, clothe and house the poor, and generate good jobs for people who need and want them. They changed the rules protecting investors from IPO volatility just before rushing over the line. Bah.
As @Ed Zitron says, we’re too far into generative AI to still be debating ROI. We know that the leaders at these companies may as well be televangelists, selling their wares while lying endlessly about the performance. “It’s too dangerous to release.” “Let’s release it.” “It blackmailed me.” “It solved those exact math problems.”
If Altman and Amodei and Musk lie about anything (and we have enormous evidence that they do), why would they ever tell us the truth about what’s really happening? There is too much capital and fame at stake, even if they corrupt the entire financial system.
The tech media abets this sham by lapping up the “AI is the messiah” nonsense. Even Bezos saying, “AGI will solve all our problems,” demonstrates a fundamental misunderstanding of the technology. LLMs won’t ever lead to AGI, and we simply cannot assert that AGI won’t destroy us.
Admittedly, I returned to school (Georgia Tech computer science) because I believed in the power of machine learning and what I thought was AI. I read Kurzweil, ignoring the nonsense and lifting the resonant phrases. I always looked to the future for the brighter day rather than remaining a hostage to the past. But learning about the technology changed my mind.
I decried deep learning as just a new label to place on an existing method. It was and is a good engineering tool, but we have reason to believe there are much more efficient ones. The problem was never with the output—it was the way it was marketed and sold. AI! AI! AI! We know the human brain doesn’t operate like a DNN, with cleverly manufactured changes capable of fooling the DNN object recognizer while a human can’t see any difference.
Watson had access to all the world’s knowledge, but it broke down on Jeopardy several times (they edited together as if Watson really operated in real time.)
I thought we were wasting valuable time and energy when LLMs hit the internet. People found them interesting, occasionally useful. But did they really help anyone?
Claude is the only generative AI product I’ve ever tried, and only for maybe fifteen minutes when I understood that the presentation of it was more impressive than the plaques-and-tangles-riddled code.
Even Microsoft tries to force Copilot on us, and I can’t even hide the goddamned icon so it stops popping up. They’ve repackaged older tools in Office to appear like Copilot, ultimately trying to say that it’s all AI somehow.
The real grift is this—you may think that the product is a system capable of AI. The real product is you and your user data. If you use Claude, Anthropic benefits from what you do. ChatGPT leverages your data similarly. You pay them money to help them make more money.
Except the tokens told the truth—they are massively burning cash to make us believe these tools have ROI. They don’t. No one would pay for a service to replace two coders that costs more than the sum of the two, plus all the time in the world for you to debug and refine. And you wind up with the bill.
Grab your virtual pitchforks and torches, friends!
AI is the mess-AI-ah
Praise Bot , the messed AI and Its only Son (Massayoshi)
AI is advancing very quickly. It’s a big mistake extrapolating from current output. ROI always comes later and usually much later. Caveat Emptor.
That's just more hype masquerading as careful, circumspective hedging. What's the advance? Why are data centers stalling in construction? Why do the spokespersons for these companies lie to us? Why are investors taking pauses? Can you explain every promise and its immediate shattering with something more concrete than that? The hurry-up-and-wait isn't helping anyone.
It's because said promises are as soulless and empty as the miscreants making them. They torch cash without any way of recovering it. Think about it--there's only so much demand for anything in this world, and for any one company to succeed, they would require almost more resources than have been invested. Their promises led to hundreds of thousands of workers to lose their jobs, then they discover that the grand output is nearly worthless.
The internet and the computer had ROI early on. There weren't high profile illusionists trying to persuade the Pentagon that these things were valuable--they could determine the military application immediately. By contrast, this technology and the gratuitous subsumption of all other tools into the bland bucket "AI" is just bad business for the rest of us.
AGI is a fairy tale--but it's been one that predates the industrial revolution and the founding. Philosophers grappled with the problem of mocking our actions with automata, statues, and so on. It's a fairy tale we all wish could be true, but the Manhattan project every month should have demonstrated verifiable, reproducible results that these systems can operate as advertised.
Economies and markets are about decisions--we fund X so we can't fund Y. There are plenty of approaches falling outside the OpenAI/Anthropic umbrella, but with all the world's money flowing into them, SpaceX, and the other tech giants just starves the rest of us.
Your emotions appear to be getting the most of you. The internet and the PC are odd examples to pick, since they’re the textbook cases of delayed ROI — the dot-com bust torched more capital faster than anything happening now, and the productivity gains from computing didn’t show up in the data for over a decade. “Payoff isn’t instant” describes every general-purpose technology, including the two you’re holding up as counterexamples. Separately, “they torch cash” lumps the labs in with the chipmakers — but Nvidia and TSMC are paid whether or not OpenAI ever clears its cost of capital. Those are different bets with different risk.
First off, I don't think you really are qualified to describe my state of mind. You cannot separate emotion from argument in any human being--you responding to me begins with emotion. I'm married to a psychiatrist.
Second, no one was flag waving, signaling, and promising the stars and moon with respect to the two examples I gave. They were implemented specifically to address problems the Pentagon wanted solved. The modern internet and PC aren't what I'm focusing on.
Further, you're not addressing the actual point--they lie to us just about weekly and daily about the ROI and when it will happen. There isn't a specific problem they're solving--it's just "everything is AI and it can do everything."
It's not clear what any objective is, beyond vague promises of the world instantly being wonderful.
They are bankrupting the global economy by subsuming other investments we badly need elsewhere.
And I've worked on this tech in my career (Microsoft, Amazon, Uber). I studied statistical machine learning at Georgia Tech and earned a graduate degree there.
I know what I’m talking about. But you don’t have to have any of that to know that the numbers make no sense. You should listen to Ed Zitron.
Some cities now are banning the building of AI data centers.
If not already started, it's time for the U. S. Government to issue licenses for the building of data centers with some strict approval criteria, including geographical boundaries related to and supported/or/not by surrounding towns, cities, and National parks, etc. and the use of resources like water and electricity, and the pre-clean up criteria for all sorts of pollution and influences on animal habitat.
None of this counts . . . because they got ahead of their skis and, like Trump, just went ahead and did what they wanted to do without giving a hoot about their responsibility for anything or anyone.
The fizzle hits companies that chased adoption metrics instead of outcomes first. When the stock narrative flips, 60 trillion tokens a month stops being a growth slide and becomes the bill nobody budgeted for.
> The fizzle hits companies that chased adoption metrics instead of outcomes first.
I used to work for one of those companies, and that is *exactly* what happened there. Good call!
We are literaly paying for predictions in form of tokens. Not intelligence like a prominient figure once said paying intelligence like electricity...
Chip-wreck. I like that haha. Hopefully it sticks to describe the popped bubble.
The AI-tanic
Almost every post from Gary contains a substantial amount of self-congratulation about being right about some prediction or other. This is really off-putting and so, so tiresome. It really gets in the way of your message. Please concentrate on the news and analysis rather than on being your own biggest promoter. That guy at the party who talks that way is a bore and it's no different on Substack, no matter how smart you are.
Maybe he should prefix all these statements with something like 'this thing has been public knowledge since at least xxxx; here's an article from then'...? Would that make it better? He's scattering receipts in these posts so the latecomers can catch up, and hopefully understand that his opinions are based on protracted, well-informed study and independent thought, not a bowling score app whipped up while walking the dog. Maybe he's wrong, but he's done the work and he has a lot of receipts. You have no idea what a professional wilderness is until you become a contrarian... look at mainstream macroeconomics for another example of what can only be called professional shunning.
Gary I think you are far, far more value creating when you write about how AI could develop, alternative models, recent front line research. Bw
It's alright for human beings to not be "value creating" 100% of the time. Additionally, what "creates value" for you may not apply to others.
With the SV set value = big money and the power it generates. For most of us value = quality of life, civil rights, and constructive human interactions for all, which in turn promote useful innovation, creativity and economic prosperity.
See above. I guess I'm not the only one.
LLMs captured the public’s and investors’ imaginations thanks to OpenAI’s decision to release ChatGPT to the public FOR FREE (very important detail) in 2022. Maybe because it’s a language model that can, at first glance, convincingly emulate intelligence and empathy, journalists, which deal in words, fell for it too (still, shame on Thomas Friedman for his NYT opinion piece on “Anthropic’s Restraint” last April, but he’s not alone).
The fact remains that, after almost 4 years of a lot of hype and money spent, no amount of ad hoc harnesses has been able to turn these probabilistic models into reliable agents, despite Dario Amodei’s fantastic assertions. To make matters worse, they also turned out to be hugely expensive to use.
As to OpenAI’s alleged USD 5.7 bn in Sales and Marketing costs in 2025, investigative journalist Ed Zitron claims (convincingly IMHO) that this figure probably refers to free token credits given away to Microsoft and other users and should have been booked as cost of sales instead.
Just curious: Are you actually claiming a trademark on "Generative AI Fizzle", as the TM superscript would imply, or is this a bit of Gary's Attempted Humor (TM)? :-)
GAH™
Question: Is a writer (like me) a large language model? Just asking …
no
FWIW, here's a link to a recent article in 'Psychology Today' about 'fluid intelligence', which has more to do with novelty than knowledge (a point Gary has made over and over): https://www.psychologytoday.com/us/blog/social-instincts/202606/the-most-misunderstood-form-of-intelligence
On the plus side, you can pretend to be one, which the LLM can't do in return. ;)
I bet you prolly have memory - so; no.
One of the wonderful things a writer does, and an LLM does not, is to realize that Elias Thorne the lighthouse-keeper does not need a cameo in every goddamn story they write.
I do like the term the Generative AI Fizzle™.
Another term that came to my mind is:
The Great AI Deflation™
...together with the sounds of a slowly deflating balloon with lukewarm air.
The bump of SPCX was probably for a large part asset managers investing in NASDAQ 100 and that one FTSE index, that Musk was able to worm his way into (luckily S&P resisted) Those asset managers are legally/contractually _required_ to rebalance their portfolios before the beginning of July, they have to buy, if they like it or not. They don’t want all of that purchasing to happen just before their deadline, because of risk management.
Lets hope Trump doesnt decide yo bail them out.
SpaceX:Apple::Beanie Babies:Honus Wagner cards
I like that for the creativity, but don't mistake my like for my hope. I still have a bunch of Honus Wagners I need to get rid of.