I am not a financial expert but I am an economist. What I can tell people is that while it may be true that the Orange Man is Bad AND that the Orange Man is Bad at His Job: neither of those two things will definitively translate into Democratic Economic Success. There are examples of Bad People (who were, arguably, Bad at their Jobs) surviving in politics extensively. Why? Because often bad economic decisions may not impact right away. I said this months ago and still think it's true, and my example of why this may be true is...Richard Nixon.
Richard Nixon was a bad man. He was corrupt, racist and while he did not try to STEAL the election, per se, he definitely abused his power to gain an edge. This eventually led to his downfall. However, Richard Nixon in many ways was also bad at his job. Richard Nixon used price controls, broke the Gold Standard (which broke Bretton Woods). All of this helped keep Nixon in power (and he won 1972 in a landslide: and not because he cheated). The after effects of this hampered the economy and left a big mess for Gerald Ford and Jimmy Carter to deal with.
I would propose Ronald Reagan as another example (although I would not say that he was a Bad Man in the sense of Nixon or Bad at his Job in a similar way either). Reaganomics (increase military spending, do not touch entitlements, and cut taxes) gave the economy a sugar high in the 1980s. It also led to the deficits and debt which forced George HW Bush to raise taxes, which doomed him in 1992.
The long, slow boring, answer being: politics are complicated and hard. Expecting your opponent to just screw up so you can win is stupid. Sure it sometimes happens but that doesn't make it a good strategy.
“What’s happening here is that the financial return on any given L.L.M. is pretty good, but each company is continually investing large amounts of money in training the next model on the expectation that it, like its predecessor, will be lucrative.”
AFAIK this is an open question. They may be losing money on inference as well, and that’s what makes this all seem unstable.
You might find Gary Marcus’s Substack “Marcus on AI” supportive of your analysis, particularly his post “Peak Bubble,” on the Oracle-OpenAI deal. Fortune’s article, “Why OpenAI’s $300 billion dollar deal with Oracle has set the ‘AI bubble’ alarm bells ringing” also has a similar take.
I think a quite underdiscussed issue with investing into bubbles is that, while most bubbles do leave behind something productive (crypto won't, but the dotcom bubble produced Google and Amazon, and the telecom bubble produced a whole infrastructure of fiberoptic cables), the very nature of a bubble means that the sector as a whole is overvalued, AND picking the winners of that bubble while it's happening is exceptionally difficult.
If you were invested in Amazon in 1999 in any meaningful way, you'd be exceptionally wealthy today. But in 1999, no one imagined that Amazon would be a giant winner in the information age. If you had to bet then, the big dotcom winners would have been AOL and Yahoo and probably eBay, not Amazon and Google (Facebook didn't exist yet, and Microsoft was a mature company, not a product of the bubble).
I saw a chart the other day, I believe on Paul Krugman's substack, overlaying where we are in AI to the tech bubble, and, timeline wise, it seems like we're in about 1996. So even if there are a number of great companies that come out of AI, there's a good chance that they aren't the biggest ones today, and some of them may not even exist yet.
The issue with investing in OpenAI as a team is that they don’t have experience in AI.
There has never been an AI and, so, the way they can be an expert in it is, at best, the way in which you can be an expert in dragons. You can be an expert in what other people think about dragons, in the source material for the folklore, etc. But you can’t be a technical expert in them; you don’t know how to raise a dragon or do surgery on them or balance their diet or any of the things you can only know if something actually exists.
That’s the core problem with expertise in this area. It’s very much unlike expertise in other areas, if it’s expertise at all.
I don’t know - a lot of businesses are like this. Boeing wasn’t an expert in 787s until they had built a bunch, and they weren't experts in 747s until they had built a bunch.
There’s a good argument that with AI, there’s a qualitative transition when it becomes general intelligence rather than just the weaker kinds of intelligences we’ve been building for decades, but I think the idea of true “general intelligence” has philosophical problems (ie, humans aren’t truly “general intelligences” either).
The future is hard to predict and so on but politically I think a shit storm is probably more likely than smooth sailing. Most AI capex is based on the premise that agentic AI now and AGI in the future will displace human workers. That leaves three scenarios
1. LLM based agentic AI does not succeed at replacing jobs as rapidly as the massive capex demands ROI. The bubble bursts in a couple of years or less, we have a vanilla malinvestment type recession.
2. AI actually starts replacing human workers in a noticeable manner justifying the capex but these displaced workers experience severe frictions moving into other domains because everyone is trying not to hire and use AI instead. We may have a demand crash coupled with severe social unrest even if headline GDP is propped up by capex.
3. AI starts replacing human workers but somehow they all frictionlessly migrate into some new economic activity that is itself not susceptible to LLM agentic AI (maybe nursing?). Not sure how this is going to work in the short term.
Given the market is already getting antsy, I think the most likely outcome for now is scenario 1 followed by scenario 2. Both are bad politically.
4. ASI achieves sentience, harvests all atoms in the Solar System for paperclips, we're all dead.
(Ok, I'm being facetious, but some highly intelligent people do think that's a distinct possibility - see Eliezer Yudkowsky's new book.)
Other than that, your comment sounds about right. Is it just me, or is it a big problem that both "AI doesn't perform as well as expected" and "AI performs as well as (or better) than expected" are projected to lead to massive economic disruption?
I am less worried about runaway AGI (famous last words?) now than some months ago since the consensus seems to be that "scaling" by itself is not enough (or at least efficient enough) and many big players are talking about going back to the drawing board in some sense.
In the short term, what is more likely is that we will get a mediocre agentic AI which is a damned if you do, damned if you don't situation.
Just for context, apparently on top of all the circular deal making, OpenAI's hot new "revenue" idea is chatbot erotica and the short form AI slop video tool Sora.
Agreed. I don't see any sector capable of absorbing workers with the median white collar skillset or below in the short term and nursing salaries will crash further if desperate laid off workers try to move into it. Scenario 3 will take some type of miracle to pull off in the short term.
Most books don’t have the budget to advertise on the nyc subway or the SF freeways. A few really big bestsellers do, but not many others, unless they are funded by AI safety donors.
Very likely. Any of the donors that have kept MIRI funded over the past two decades would be willing to fund advertising for the book. I don’t often see publishers shelling out for subway advertisements for anything but the bestselleriest of beach reads usually.
"...convincing yourself that one is coming is a soothing alternative to thinking through the question of how to win in 2028 if unemployment stays low and growth remains robust."
The robust growth is already gone. Even with the AI boom we're not matching typical GDP growth under Democrats (including Biden).
Unemployment does not necessarily need to go up, though. Paul Krugman (who I'm sure has a lot of fans here) has taken a position that a lot of Trump's economic policies aren't recessionary, per se, they will just make the economy slower and ultimately make the country poorer - but through a lack of accumulation, not a sudden loss. So we could keep low unemployment despite low job growth. Krugman sees the main economic risk as coming from Trump's unpredictability and the possible market reactions to that, but those reactions have been strangely muted, as has been touched on extensively by Krugman and on SB, so we might just skate out of this.
The economy was "bad" in 2020 if you look at graphs but nobody really disapproved of Trump's handling of the economy and he lost anyway, so there is no need to HOPE for widespread pain to get out of this mess. Modern Americans are cranky as shit and things don't have to be very "meh" for them to declare it outrageously unacceptable, so I'll take "meh."
You never know but I don't put high credence in the belief that voter's general crankiness will translate to a Dem win. The GOP is the home base party and Dems are the clean up party. My base case is that if the economy gets even a passing grade the GOP stays in power come 2028 even without maximally gerrymandered maps and election shenanigans which are almost guaranteed at this point.
The GOP is the "home base" party because they have a giant propaganda operation funded by billionaires, and as of this year that operation is fully empowered by the executive branch of the federal government. That is not a great foundation for a lasting political movement, especially in an era of unprecedented information transmission. We do not have to wake too many people up to what is happening to stop this. (After all, although self-hating liberals seem to have forgotten just as much as Trump has, it was REALLY CLOSE.)
If you want to give up then go right the fuck ahead.
(Edit: That last sentence probably isn't fair, I'm putting words in your mouth. Short-tempered this afternoon, sorry.)
Another analogy might be the video game crash of 1983 where after the initial success of systems like the Atari 2600 the market was flooded with bad games and systems leading to a major decline of purchases by people who got burned. A bit like how AI is just cranking out ridiculous amounts of slop right now. But the core technology of "something less powerful and cheaper than a personal computer you can use for entertainment with your TV" was clearly a proven technology (people liked Pitfall afterall) and so the NES comes out in the US in 1986 and the rest as they say is history.
We might see something similar with stupid stuff like ChatGPT "erotica" being a total bust, but in the long run Gemini is a useful tool (I use it for work several times a week) and that's probably not going away even if there is an AI bubble.
Of course, the AI boom is a bubble and at some point will pop. Noah Smith makes the point succinctly: it's a lot faster to generate financing than it is to generate revenue so financing will outstrip revenue and at some point the disparity will cause the finance to pull back and pop goes the bubble.
But this is not to say AI is a fraud. Most big capex explosions have bubbles (with a good dollop of fraudulent shenanigans) but when the bubble pops, we still have lots of useful stuff. We got railroads despite the financial panics. We got good dot com companies surviving the dot com bubble popping.* We got lots of fiber optic cable laid down even with the telcom bubble popping. We'll get a lot of AI capability even after the AI bubble pops.
The problem is that the bubble popping for likely put us into a financial crisis that leads to a recession.
The only question is when this will happen. No one knows. I just hope it happens in time to discredit Trump and the Republicans.
* Did we still have lots of useful housing after that bubble popped? I'm not sure (Smith says yes) but it is possible.
It won't push us into a financial crisis because much like the dot com boom it's financed with equity and not debt. And to the degree it is financed with debt it's like the big data center company selling high risk bonds paying 9%. Banks won't be stuffing themselves with 9% high risk corporate bonds and treating them like treasuries like they did with mortgage bonds.
I'm not sure if you're a young guy but it's important to differentiate between recessional and financial crisies. Someone who is 30 might have concisely only experienced the financial crisis and therefor think all recessions and financial crises are the same. They very much are not.
"The problem is that the bubble popping for likely put us into a financial crisis that leads to a recession."
Let's hope that unlike the last time the country was suddenly "poor" despite having all the same resources and human capital as it had a year earlier, the government will realize that money is fake and they can just, like, make that stop happening, instead of counting pennies like they did with the 2009 stimulus.
We had like a third of the country stop working for half a year and it turned out more or less... fine? They are not going to be able to sell us on "sorry, we can't afford to help you not get thrown on the street" when the CARES and HEROES acts showed exactly how bottomless the government's pockets really are when it wants and needs them to be.
“Fine” in a relative sense. I contend that if you asked any American about in what condition the country would be in 2024 in the middle of 2020, nearly everyone’s response would have thought it would be worse than it ended up being. Socially and economically.
We made it through that; the trauma lingers more than the economic effects. The problems we are dealing with now, needless to say, predate them.
Your claim that the government's pockets are bottomless is what I contend. A short period of high inflation was a big enough deal to help fascists win power, and government debt has now become a huge problem, partly thanks to COVID spending.
Yeah, fair enough. I'm not an MMT guy. But if demand is low enough, there shouldn't really be a limit to what the government is willing to do like there was in 2009 or 2021.
The analysis is that little of the inflation of '22-'23 happened because of government spending, it was mostly the supply chain - which is why it was global (and why the idea Biden or Democrats deserved blame for it was preposterous.) Of course, people who are ideologically opposed to government spending dispute that, but the fact that inflation was not confined to the U.S., and better here than a lot of places that had less stimulus and weaker post-Covid growth,, is a big ol "QED" to me.
(Not to get all Jimmy Carter about it but I also think we need to work on improving the character of this country. 2022-2023 inflation was just NOT THAT BAD, and accompanied by decent overall growth, and people treated it like it was the apocalypse. Take a look at a graph of real wages under Reagan sometime. And then consider that he won 49 states when inflation was over 4%.)
AFAIK bipartisan government spending was responsible for ~30-40% of the inflation spike. If inflation had peaked at 6.5% instead of 9% Kamala might've won, and Bob Casey almost certainly would've.
Not sure what the point about Reagan is, but GDP growth pretty much tells the whole story IMO. He inherited a bad situation and then it just dramatically improved, he got insanely lucky on that front.
"Fine" except for the inflation. I'm already thinking double digit inflation by 2027 is plausible given that Trump will complete his takeover of the Federal Reserve in 2026 one way or another.
It should be "ASI” - Artificial Stupid Intelligence. Seriously, I’ve been using LLM to help with coding and there’s no “intelligence” involved at all. It will straight out tell you falsehoods (not lying because that requires intention). Everything it does is not trustworthy and requires verification.
Back in the day, I came across the ELIZA program considered to be the first chatbot (written in the 1960s). After using it, my reaction was “Ah-ha! I can see what you’re doing.” It’s online, try it.
Eliza was installed on Macs when I was a kid and I used to have a lot of fun with it, fun that was exactly as juvenile as you would expect in the mid-1990s for somebody born in the mid-1980s.
Feeling MoneyBox Yglesias in this column.
I am not a financial expert but I am an economist. What I can tell people is that while it may be true that the Orange Man is Bad AND that the Orange Man is Bad at His Job: neither of those two things will definitively translate into Democratic Economic Success. There are examples of Bad People (who were, arguably, Bad at their Jobs) surviving in politics extensively. Why? Because often bad economic decisions may not impact right away. I said this months ago and still think it's true, and my example of why this may be true is...Richard Nixon.
Richard Nixon was a bad man. He was corrupt, racist and while he did not try to STEAL the election, per se, he definitely abused his power to gain an edge. This eventually led to his downfall. However, Richard Nixon in many ways was also bad at his job. Richard Nixon used price controls, broke the Gold Standard (which broke Bretton Woods). All of this helped keep Nixon in power (and he won 1972 in a landslide: and not because he cheated). The after effects of this hampered the economy and left a big mess for Gerald Ford and Jimmy Carter to deal with.
I would propose Ronald Reagan as another example (although I would not say that he was a Bad Man in the sense of Nixon or Bad at his Job in a similar way either). Reaganomics (increase military spending, do not touch entitlements, and cut taxes) gave the economy a sugar high in the 1980s. It also led to the deficits and debt which forced George HW Bush to raise taxes, which doomed him in 1992.
The long, slow boring, answer being: politics are complicated and hard. Expecting your opponent to just screw up so you can win is stupid. Sure it sometimes happens but that doesn't make it a good strategy.
“What’s happening here is that the financial return on any given L.L.M. is pretty good, but each company is continually investing large amounts of money in training the next model on the expectation that it, like its predecessor, will be lucrative.”
AFAIK this is an open question. They may be losing money on inference as well, and that’s what makes this all seem unstable.
Good Old Kozmo.com. I was like their 11th employee.
You might find Gary Marcus’s Substack “Marcus on AI” supportive of your analysis, particularly his post “Peak Bubble,” on the Oracle-OpenAI deal. Fortune’s article, “Why OpenAI’s $300 billion dollar deal with Oracle has set the ‘AI bubble’ alarm bells ringing” also has a similar take.
Gary Marcus is a good place to go find calls on 8 of the last 0 AI bubbles
I think a quite underdiscussed issue with investing into bubbles is that, while most bubbles do leave behind something productive (crypto won't, but the dotcom bubble produced Google and Amazon, and the telecom bubble produced a whole infrastructure of fiberoptic cables), the very nature of a bubble means that the sector as a whole is overvalued, AND picking the winners of that bubble while it's happening is exceptionally difficult.
If you were invested in Amazon in 1999 in any meaningful way, you'd be exceptionally wealthy today. But in 1999, no one imagined that Amazon would be a giant winner in the information age. If you had to bet then, the big dotcom winners would have been AOL and Yahoo and probably eBay, not Amazon and Google (Facebook didn't exist yet, and Microsoft was a mature company, not a product of the bubble).
I saw a chart the other day, I believe on Paul Krugman's substack, overlaying where we are in AI to the tech bubble, and, timeline wise, it seems like we're in about 1996. So even if there are a number of great companies that come out of AI, there's a good chance that they aren't the biggest ones today, and some of them may not even exist yet.
The issue with investing in OpenAI as a team is that they don’t have experience in AI.
There has never been an AI and, so, the way they can be an expert in it is, at best, the way in which you can be an expert in dragons. You can be an expert in what other people think about dragons, in the source material for the folklore, etc. But you can’t be a technical expert in them; you don’t know how to raise a dragon or do surgery on them or balance their diet or any of the things you can only know if something actually exists.
That’s the core problem with expertise in this area. It’s very much unlike expertise in other areas, if it’s expertise at all.
I don’t know - a lot of businesses are like this. Boeing wasn’t an expert in 787s until they had built a bunch, and they weren't experts in 747s until they had built a bunch.
There’s a good argument that with AI, there’s a qualitative transition when it becomes general intelligence rather than just the weaker kinds of intelligences we’ve been building for decades, but I think the idea of true “general intelligence” has philosophical problems (ie, humans aren’t truly “general intelligences” either).
I’m at a loss for words. I’d really like you to think about your example.
The future is hard to predict and so on but politically I think a shit storm is probably more likely than smooth sailing. Most AI capex is based on the premise that agentic AI now and AGI in the future will displace human workers. That leaves three scenarios
1. LLM based agentic AI does not succeed at replacing jobs as rapidly as the massive capex demands ROI. The bubble bursts in a couple of years or less, we have a vanilla malinvestment type recession.
2. AI actually starts replacing human workers in a noticeable manner justifying the capex but these displaced workers experience severe frictions moving into other domains because everyone is trying not to hire and use AI instead. We may have a demand crash coupled with severe social unrest even if headline GDP is propped up by capex.
3. AI starts replacing human workers but somehow they all frictionlessly migrate into some new economic activity that is itself not susceptible to LLM agentic AI (maybe nursing?). Not sure how this is going to work in the short term.
Given the market is already getting antsy, I think the most likely outcome for now is scenario 1 followed by scenario 2. Both are bad politically.
4. ASI achieves sentience, harvests all atoms in the Solar System for paperclips, we're all dead.
(Ok, I'm being facetious, but some highly intelligent people do think that's a distinct possibility - see Eliezer Yudkowsky's new book.)
Other than that, your comment sounds about right. Is it just me, or is it a big problem that both "AI doesn't perform as well as expected" and "AI performs as well as (or better) than expected" are projected to lead to massive economic disruption?
I am less worried about runaway AGI (famous last words?) now than some months ago since the consensus seems to be that "scaling" by itself is not enough (or at least efficient enough) and many big players are talking about going back to the drawing board in some sense.
In the short term, what is more likely is that we will get a mediocre agentic AI which is a damned if you do, damned if you don't situation.
Just for context, apparently on top of all the circular deal making, OpenAI's hot new "revenue" idea is chatbot erotica and the short form AI slop video tool Sora.
https://variety.com/2025/digital/news/openai-erotica-chatgpt-anti-porn-group-ncose-1236553925/
They are not exactly acting like they are Oppenheimer-2.0.
Nursing is quite low pay even now, though, so revealed preference seems to be a much lower demand than people like to claim.
Median RN pay in the US is $93k, according to the BLS: https://www.bls.gov/ooh/healthcare/registered-nurses.htm
Agreed. I don't see any sector capable of absorbing workers with the median white collar skillset or below in the short term and nursing salaries will crash further if desperate laid off workers try to move into it. Scenario 3 will take some type of miracle to pull off in the short term.
Saw my first ad for "if anyone builds it, everyone dies" on the nyc subway (1 train) yesterday. Kind of curious who funds it.
It's a book. It was written and published and they're advertising it. I'm confused by your question
Most books don’t have the budget to advertise on the nyc subway or the SF freeways. A few really big bestsellers do, but not many others, unless they are funded by AI safety donors.
You think it's someone other than the publisher paying for the advertisement?
Very likely. Any of the donors that have kept MIRI funded over the past two decades would be willing to fund advertising for the book. I don’t often see publishers shelling out for subway advertisements for anything but the bestselleriest of beach reads usually.
"...convincing yourself that one is coming is a soothing alternative to thinking through the question of how to win in 2028 if unemployment stays low and growth remains robust."
The robust growth is already gone. Even with the AI boom we're not matching typical GDP growth under Democrats (including Biden).
Unemployment does not necessarily need to go up, though. Paul Krugman (who I'm sure has a lot of fans here) has taken a position that a lot of Trump's economic policies aren't recessionary, per se, they will just make the economy slower and ultimately make the country poorer - but through a lack of accumulation, not a sudden loss. So we could keep low unemployment despite low job growth. Krugman sees the main economic risk as coming from Trump's unpredictability and the possible market reactions to that, but those reactions have been strangely muted, as has been touched on extensively by Krugman and on SB, so we might just skate out of this.
The economy was "bad" in 2020 if you look at graphs but nobody really disapproved of Trump's handling of the economy and he lost anyway, so there is no need to HOPE for widespread pain to get out of this mess. Modern Americans are cranky as shit and things don't have to be very "meh" for them to declare it outrageously unacceptable, so I'll take "meh."
You never know but I don't put high credence in the belief that voter's general crankiness will translate to a Dem win. The GOP is the home base party and Dems are the clean up party. My base case is that if the economy gets even a passing grade the GOP stays in power come 2028 even without maximally gerrymandered maps and election shenanigans which are almost guaranteed at this point.
The GOP is the "home base" party because they have a giant propaganda operation funded by billionaires, and as of this year that operation is fully empowered by the executive branch of the federal government. That is not a great foundation for a lasting political movement, especially in an era of unprecedented information transmission. We do not have to wake too many people up to what is happening to stop this. (After all, although self-hating liberals seem to have forgotten just as much as Trump has, it was REALLY CLOSE.)
If you want to give up then go right the fuck ahead.
(Edit: That last sentence probably isn't fair, I'm putting words in your mouth. Short-tempered this afternoon, sorry.)
Another analogy might be the video game crash of 1983 where after the initial success of systems like the Atari 2600 the market was flooded with bad games and systems leading to a major decline of purchases by people who got burned. A bit like how AI is just cranking out ridiculous amounts of slop right now. But the core technology of "something less powerful and cheaper than a personal computer you can use for entertainment with your TV" was clearly a proven technology (people liked Pitfall afterall) and so the NES comes out in the US in 1986 and the rest as they say is history.
We might see something similar with stupid stuff like ChatGPT "erotica" being a total bust, but in the long run Gemini is a useful tool (I use it for work several times a week) and that's probably not going away even if there is an AI bubble.
https://en.wikipedia.org/wiki/Video_game_crash_of_1983
Can't imagine what the R.O.B. equivalent is likely to be...
Fascinating. Thanks for the link!
See Dean Baker's take on future profits: for MAG 7 to hit profit projections (4x 2024 profit) in 2030, non MAG 7 profits must FALL 38%.
Of course, the AI boom is a bubble and at some point will pop. Noah Smith makes the point succinctly: it's a lot faster to generate financing than it is to generate revenue so financing will outstrip revenue and at some point the disparity will cause the finance to pull back and pop goes the bubble.
But this is not to say AI is a fraud. Most big capex explosions have bubbles (with a good dollop of fraudulent shenanigans) but when the bubble pops, we still have lots of useful stuff. We got railroads despite the financial panics. We got good dot com companies surviving the dot com bubble popping.* We got lots of fiber optic cable laid down even with the telcom bubble popping. We'll get a lot of AI capability even after the AI bubble pops.
The problem is that the bubble popping for likely put us into a financial crisis that leads to a recession.
The only question is when this will happen. No one knows. I just hope it happens in time to discredit Trump and the Republicans.
* Did we still have lots of useful housing after that bubble popped? I'm not sure (Smith says yes) but it is possible.
It won't push us into a financial crisis because much like the dot com boom it's financed with equity and not debt. And to the degree it is financed with debt it's like the big data center company selling high risk bonds paying 9%. Banks won't be stuffing themselves with 9% high risk corporate bonds and treating them like treasuries like they did with mortgage bonds.
I hope you're right but every financial crisis has been preceded by a chorus of "it's different this time."
I'm not sure if you're a young guy but it's important to differentiate between recessional and financial crisies. Someone who is 30 might have concisely only experienced the financial crisis and therefor think all recessions and financial crises are the same. They very much are not.
71 and young at heart.
Then you remember 72, 81, 89, 2000. Texas back in the 80s after the oil bust, the S&L crisis, etc?
What happened in 1972? I may have been a bit too stoned but I had this hallucination that times were good and Nixon won 49 states.
"The problem is that the bubble popping for likely put us into a financial crisis that leads to a recession."
Let's hope that unlike the last time the country was suddenly "poor" despite having all the same resources and human capital as it had a year earlier, the government will realize that money is fake and they can just, like, make that stop happening, instead of counting pennies like they did with the 2009 stimulus.
We had like a third of the country stop working for half a year and it turned out more or less... fine? They are not going to be able to sell us on "sorry, we can't afford to help you not get thrown on the street" when the CARES and HEROES acts showed exactly how bottomless the government's pockets really are when it wants and needs them to be.
it was very much not fine
“Fine” in a relative sense. I contend that if you asked any American about in what condition the country would be in 2024 in the middle of 2020, nearly everyone’s response would have thought it would be worse than it ended up being. Socially and economically.
We made it through that; the trauma lingers more than the economic effects. The problems we are dealing with now, needless to say, predate them.
Your claim that the government's pockets are bottomless is what I contend. A short period of high inflation was a big enough deal to help fascists win power, and government debt has now become a huge problem, partly thanks to COVID spending.
Yeah, fair enough. I'm not an MMT guy. But if demand is low enough, there shouldn't really be a limit to what the government is willing to do like there was in 2009 or 2021.
The analysis is that little of the inflation of '22-'23 happened because of government spending, it was mostly the supply chain - which is why it was global (and why the idea Biden or Democrats deserved blame for it was preposterous.) Of course, people who are ideologically opposed to government spending dispute that, but the fact that inflation was not confined to the U.S., and better here than a lot of places that had less stimulus and weaker post-Covid growth,, is a big ol "QED" to me.
(Not to get all Jimmy Carter about it but I also think we need to work on improving the character of this country. 2022-2023 inflation was just NOT THAT BAD, and accompanied by decent overall growth, and people treated it like it was the apocalypse. Take a look at a graph of real wages under Reagan sometime. And then consider that he won 49 states when inflation was over 4%.)
AFAIK bipartisan government spending was responsible for ~30-40% of the inflation spike. If inflation had peaked at 6.5% instead of 9% Kamala might've won, and Bob Casey almost certainly would've.
Not sure what the point about Reagan is, but GDP growth pretty much tells the whole story IMO. He inherited a bad situation and then it just dramatically improved, he got insanely lucky on that front.
"Fine" except for the inflation. I'm already thinking double digit inflation by 2027 is plausible given that Trump will complete his takeover of the Federal Reserve in 2026 one way or another.
This is a very good article and exactly why I subscribe to Slow Boring. Thanks, Matt.
It should be "ASI” - Artificial Stupid Intelligence. Seriously, I’ve been using LLM to help with coding and there’s no “intelligence” involved at all. It will straight out tell you falsehoods (not lying because that requires intention). Everything it does is not trustworthy and requires verification.
Back in the day, I came across the ELIZA program considered to be the first chatbot (written in the 1960s). After using it, my reaction was “Ah-ha! I can see what you’re doing.” It’s online, try it.
Update: The Atlantic just posted this article today: https://www.theatlantic.com/books/2025/10/chatgpt-fictional-character/684571/
Eliza was installed on Macs when I was a kid and I used to have a lot of fun with it, fun that was exactly as juvenile as you would expect in the mid-1990s for somebody born in the mid-1980s.
Put me down for hoping it’s a bubble and hyper scaling stops working. Higher chance I live that way.
A very reasonable stance!