Liz Hoffman: Will AI Trigger a Jobless Food Fight?

Summary

The Gist – “Liz Hoffman: Will AI Trigger a Jobless Food Fight?”

Date: April 30, 2026
Host: Mike Pesca
Guest: Liz Hoffman (Business and Finance Editor, Semaphore; former Wall Street Journal reporter and author)

---

Episode Overview

This episode of The Gist dives into the potential economic and societal upheaval that could be triggered by the rapid acceleration of artificial intelligence (AI). Guest Liz Hoffman—currently at Semaphore and co-host of the “Compound Interest” podcast—discusses the pace of technological change, implications for jobs and markets, and the regulatory and ethical dilemmas facing industry and government. The conversation is candid, often wry, and pulls back from apocalyptic scenarios to ground the discussion in history, policy, and real-world trade-offs.

---

Key Discussion Points & Insights

The Supreme Court Decision and Its Political Impact

• [00:56–05:44] Pesca opens with analysis of a recent Supreme Court decision ending majority-minority congressional districts, forecasting impacts on Democrats’ prospects but ultimately downplaying a catastrophic effect. This prefaces a segue into the economic segment.

---

AI: Transformative Power and the Future of Work

The Scope and Pace of AI Change

• [06:21] Hoffman contends that while physical robotics (e.g., household chores) remain lagging, advances in intelligence are close to surpassing the smartest humans:

  “I think within, call it two years, [AI] will be smarter than the smartest person you know, and possibly smarter than the smartest person we’ve ever produced.” – Liz Hoffman [06:44]

• [07:41] Both agree the traditional hope that new technology will merely shift, not displace, employment is shaky. Hoffman suggests pace is key: if change is gradual, job categories may adapt or expand, but rapid change means “massive unemployment, massive dislocation.”

• Pesca presses for concrete examples of future “AI-created” jobs, highlighting the speculative nature of positive predictions.

  “If they knew it, they would be hiring for it.” – Liz Hoffman [08:52]

Historical Analogies and Limits

• Hoffman invokes the “Fourth Industrial Revolution” analogy, but with a caveat: it’s easier to see jobs destroyed than those created.
• She references Jevons’ Paradox—technology may produce more of certain kinds of work (e.g., health care, conductors with steam trains), but certain industries, like banking, are “demand constrained” and unlikely to expand their workforce.

Inequality and Sectoral Disruption

• AI’s impact will be uneven:

  “There are different buckets of sectors that are going to benefit from this in really interesting ways and ones where it is just going to turn into a jobless food fight for the share that remains.” — Liz Hoffman [11:31]

Notable Moment

• Printing Press Parallel: Hoffman notes that the printing press, while eventually positive, also “led to an unbelievable number of deaths” due to ensuing wars and upheavals. Both agree AI could be equally unpredictable:

  “I just think AI is going to be unbelievably transformative in a way that we can’t possibly comprehend.” – Mike Pesca [10:41]

---

AI Business Models and Value Creation

• [12:51] The discussion shifts to whether AI’s business models are sustainable:

  • OpenAI and Anthropic began with different approaches (consumer-facing vs B2B).
  • Consumer tech has “one winner,” while enterprise may yield several; Uber shifts toward managing fleets rather than serving individuals.

• [13:30]

  “Uber is… tucking those robo taxis in at night. It’s cleaning them, it’s parking them, it’s charging them, it’s servicing them. And their customers are in fact mostly not you and me, but are big fleet owners. …You cannot get more sort of like mundane enterprise business services than managing like a fleet owned by Blackstone.”—Liz Hoffman

• OpenAI is now pivoting toward enterprise tools, recognizing that consumer adoption alone cannot sustain their financial model.

---

AI Regulation, Ethics, and Geopolitics

Regulatory Readiness

• Pesca asks if there is anyone up to the challenge of regulating AI:

  “Congress had all the social media CEOs come down…It was so obvious they had no idea how these companies made money.” — Liz Hoffman [15:18]

• Universal Basic Income (UBI) and new tax models (e.g., “taxing compute instead of labor”) are discussed as possible, if radical, policy responses.

• Hoffman is skeptical about UBI’s adequacy, emphasizing that people need work for meaning:

  “It would only solve a tiny fraction of the problem…People don’t want to essentially live as pensioners…what happens to your drive? It’s not that pretty a picture.” – Mike Pesca [16:09–16:52]

AI Firms and Moral Signaling

• [19:08] Hoffman is asked who “wears the white hat” in the Musk vs Altman/openAI legal dispute and the Pentagon’s conflict with Anthropic.

  “We don’t understand this tech, but neither do the people making it…It is fundamentally a black box, and…we don’t know why it does a lot of things it does, and the people making it don’t even know.” – Liz Hoffman [19:32]

• She draws an analogy between raising AI and raising children: “you try to instill the values…understanding that it’s gonna go out in the world and be autonomous in some way.” [20:57]

• Regulatory lines (e.g., Anthropic’s refusal to provide “smart weapons” to the Pentagon) are as much about branding as actual safety, with different companies seeking to be seen as responsible.

Geopolitical Stakes

• Hoffman underscores bipartisan concerns about China’s potential to define global AI standards:

  “If we don’t win on this, China does, and there’s going to be global models that are…not us. They are the global South…[China] is very keen to put its models into those economies and like, who knows what kind of morals and…freedoms are written into those models.” [22:14]

• The question of what “winning” with AI actually means remains open.

---

Notable Quotes & Memorable Moments

• On AI’s Transformative Power:
  “It will decimate work as we know it. It will more than decimate work. Work will stop being work.” – Mike Pesca [07:25]

• On B2B vs B2C in AI:
  “When those venture subsidies go away and you actually have to have people paying for your product, turns out you can win in consumer. But there is generally one winner in consumer. Google won consumer search. Right. Bing did not.” – Liz Hoffman [12:51]

• On Regulation:
  “This idea of should be taxing compute instead of labor going forward. The answer to me is obviously yes, but you’re talking about rewriting a couple hundred years of the tax code to do that.” – Liz Hoffman [15:56]

• On Existential Risk:
  “The consensus, the latest sort of straw poll out of Silicon Valley was 10% that [AI] will destroy everything.” – Liz Hoffman [20:14]

• On Industry Morals:
  “Anthropic has tried to write that [moral center] into its models. It has this constitution where it tries to draw the line.” – Liz Hoffman [20:29]

---

Highlighted Timestamps

• Supreme Court Voting Rights Act Reaction: 00:56–05:44
• AI will surpass human intelligence soon: 06:21–06:56
• Transformation of work and historical analogies: 07:41–09:15
• Job creation skepticism: 08:52
• Sector-specific AI effects (health care, banking): 10:57–12:01
• AI business models and B2B shift: 12:51–14:49
• Regulation, UBI, and labor/tax innovation: 15:18–16:52
• Ethics and 'white hats' in tech: 17:49–21:14
• China and global AI competition: 21:21–22:26

---

Tone and Language

• Conversational, sharp, and self-aware.
• Candid skepticism toward tech utopianism and simple historical analogies.
• Jargon is unpacked or mocked, and both participants use humor and illustrative asides.
• Pesca challenges clichéd talking points with direct questions and empirical skepticism.

---

Summary Value

This episode delivers a sobering, witty, and nuanced look at the future of AI’s impact on work, meaning, business models, and geopolitics. It is both a reality check on “inevitable” progress and a warning that, historically, the transition is neither smooth nor necessarily positive for all. The conversation oscillates between optimism about adaptation and deep caution about the unknowable scope of change. Anyone seeking a clear-eyed, jargon-busting reality check on the hype and hope of AI’s economic impact will find this episode invaluable.

Transcript

[Progressive Insurance Announcer]

Insurance isn't one size fits all, and shopping for it shouldn't feel like squeezing into something that just doesn't fit. That's why drivers have enjoyed Progressive's Name your price tool for years. With the name your price tool, you tell them what you want to pay and they show you options that fit your budget enough. Hunting for discounts, trying to calculate rates, and tinkering with coverages. Maybe you're picking out your very first policy, or maybe you're just looking for something that works better for you and your family. Either way, they make it simple to see your options. No guesswork, no surprises. Ready to see how easy and fun shopping for car insurance can be? Visit progressive.com and give the Name your price tool a try. Take the stress out of shopping and find coverage that fits your life on your terms. Progressive Casualty Insurance Company and affiliates Price and coverage match Limited by state law

[Mike Pesca]

It's Thursday, April 30, 2026. From Peachfish Productions, it's the Gist. I'm Mike Pitt. Pesca the Supreme Court handed down its decision in Calais, which was expected, though maybe not this early. So majority minority districts are no longer permissible under the Voting Rights Act. In fact, they are seen as discriminatory against whites. The timing of the decision surprised some observers. I saw some, I don't want to say panicking, but there was urgency among Democrats and surprise. A surprising amount of surprise. There's an organization called When We All Vote and they put out a YouTube short about the electoral impacts of this decision.

[Liz Hoffman]

As states like Florida and Mississippi are redrawing their congressional maps right now, the court's decision could actually erase up to 10 majority black seats across the country, all before the midterm elections that are happening in November.

[Mike Pesca]

The organization Indivisible, also a voting rights organization, says a ruling that undermines section two could, quote, lock in up to 19 additional safe Republican seats in the U.S. house. House 19? I doubt that. I am not sure about 10, but I'm trying to figure out if that can happen when we all vote. Analysis is not wrong when they say that Mississippi and Florida are redrawing maps. Florida was going to do it anyway. Mississippi seems very excited to do it now. And that means Bennie Thompson, who's been a black Democrat in Congress for 30 years, his district will get split in half or quarters, and the 40% of Mississippi's voters who are black might have no representation in the house. Then again, 40% of Mississippi's voters are black. That might mean that there is some basis for a Democratic vote. Although when I look at the lay of the land in Mississippi, I don't know. Thompson is plus 11. That's his district. So what they do is they draw it so that the guy who is going to be Democrat gets as many Democrat votes as possible. And in Mississippi that means black votes. And that's all a proxy for Calais in this gerrymandering fight. But the other three Mississippi was Reps who are Republicans go from plus 14 to plus 21. So you can see dispersing those Thompson voters in such a way that even in this an overwhelmingly Democratic year, Mississippi goes all Republican. Now let's also keep in mind that not all states are going to change their maps in time. Louisiana, the state that Calais was fought over, that has a May 16 primary and they're trying to race the clock to eliminate their two Democratic, which is to say black representatives. I don't know if it will be able to or even possible to get to six Republican. They could go five one time is more a factor there than the ability to draw the districts. Other states still have enough Democrats in this a Democratic wave year to thwart these gerrymandering efforts. I do not think this will seriously imperil the Democrats chances of taking back the House. The prediction markets agree with me and of course you gerrymander a Senate seat. So I'm waiting for all the maps to get drawn and I'm waiting for some less surprised analysis to take over. You have Cook, who I mentioned. You have Nate Silver. There's a new guy named Lakshya Jane from the argument. He's good one. And I will be watching all of them on the show today. Over the next two days. Today and tomorrow, I will treat you to a deep dive into California law and regulations. Today it's Prop 36 which imposed the possibility of more penalties on repeat shoplifters. How's that going? Well, we're not sure, but we're sure that it's not going how some describers, how some observers say it's going. But first I've been listening to a really good Semaphore podcast. Compound interest and in fact interesting is the best way to describe one of the hosts, Liz Hoffman. I've been aware of her work and have read her work for years and now she comes on the gist to discuss all that is going on with AI economics and anything that could make or these days lose you money. Liz Hoffman up next. Liz Hoffman is now Semaphore's business and finance editor. She was with the Wall Street Journal I read her book Crash Landing the inside story of how the world's biggest company survived an economy on the brink. And I have to tell you, I don't usually read books unless I'm going to book the author. And I don't think Liz has ever been on the show. But now that she is hosting compound interest semaphore. I got her on the show and we're going to do semi rapid fire all about the economy. I maybe GLP1s. Hello Liz, welcome to the gist.

[Liz Hoffman]

Hey Mike, thanks for having me.

[Mike Pesca]

So I mentioned GLP1s because you were talking to Mark Cuban. You said there were four things important in the world. GLP1s the NBA playoffs. Let's cap in that because that was maybe a SOP to Cuban and Iran and AI and I kind of agree if with GLP1 you just counted as the category of technologically induced miracles and if maybe with the NBA you call it entertainment. So let's start with AI because that's where everyone in the economy is starting with, do you think? I don't know, give me a time horizon. But is there realistically anything I won't be able to do?

[Liz Hoffman]

It'll be a while before they can do like the robots can like install your, your FiOS. Right. Think about like. Really?

[Mike Pesca]

Yeah, because the humans are great at that in a timely manner.

[Liz Hoffman]

You know, if you've seen the videos, like the robot trying to unload the dishwasher for 45 minutes and like doing it bad, there's a long tail of like very dexterous things that will be hard. But like when you're talking about actual intelligence. No, and I think pretty, pretty much faster. I think within call it two years, I will be smarter than the smartest person you know and possibly smarter than the smartest person we've ever produced.

[Mike Pesca]

Yeah, it's probably the same person for both of us, but I don't want to get into names or at least he thinks so. So I agree with everything you're saying and I think they also will get into the robotics that is blue collar workers or fans thereof are trying to cling on to that saying you'll always need the working man. Probably not. So with this massive world changing technology, I don't even want to get into the so called 20% chance it destroys everything. It's just very, it's hard to contemplate, but it seems inescapable that not only I don't want to use the wiggle word of change, it will decimate work as we know it, it will more than decimate work. Work will stop being work. I mean, is there any way around that?

[Liz Hoffman]

Okay, I think the question is around the pace of change. So, like, not to go all like historical on you, but like, you know, people are talking about, this is the fourth industrial revolution. We had steam, we had combustion, we had semiconductors. Now we have this. And at every stage people said exactly what you're saying. And we have more people employed now than ever in different jobs. The question is how quickly it comes, because I think it's a lot easier right now. And this is probably what people were saying when the steam engine came along. It's a lot easier to, to see the jobs that it will replace than the jobs that it will create. And if that happens slowly enough, right, like in a couple of years, we'll be sitting here saying, well, there's this whole new skill that we need that we just didn't think about because we couldn't imagine the way that AI would rewire the world and the economy. That's great. But if it happens too fast and you can't retrain the people or get people into those skills in time, then you're going to have massive unemployment, massive dislocation. And so to me, it's a question of how fast it happens.

[Mike Pesca]

I used to agree with that. I used to hold on to that hope. But it would really help me concentrate, concentrate on that eventuality or that prediction. If you could name one of these things, one of these skill type jobs that might exist that we can't even think of now.

[Liz Hoffman]

You know, I've asked literally every smart person I can that question. And like, if, if they knew it, they would be hiring for it.

[Mike Pesca]

Ok, so sorry, let me do a timeout. But with steam, couldn't they have done that instead of the steam engine or instead of the coal engine? It's like, all right, well, this other technology will now power our ships. So but that's not that. So that analogy doesn't seem to apply with AI.

[Liz Hoffman]

Well, okay, so there's another piece which is that we may not need jobs we can't imagine. We actually need more of jobs that we're already kind of familiar with. And this is what you hear people talk about Jevons paradox, right, which is that when something gets cheaper and easier to produce, we get more of that thing. And so, you know, maybe to take your steam analogy, like steam, steam trains were a lot. You know, people started taking trains because they were cheaper and easier. And so we got a lot more like conductors and People tending bar in the train cars and like those jobs do exist. The print printing press is a great example of that.

[Mike Pesca]

Guys pack those little nut packages on the Amtrak. I mean those are really good ones. You know, the prepackaged nuts and the little wine that you peel off the

[Liz Hoffman]

entire modern media industry that has popped up since the advent of the printing press. I mean those are jobs. When it became easier to share information, we got a lot more people like you and me who were in the information gathering and sharing business. So I'm trying to put like a slightly more optimistic spin on it. I would not say that I'm terribly optimistic. I think this is going to be really painful.

[Mike Pesca]

Really painful. And by the way, the printing press, you know, you listed steam and some other great technologies. There are other ways to look at it and you could say like antibiotics and the printing press and a few others. Agriculture were the big changes. And I would say the printing press, glad we have it. Wish we had some more of it for about 100 years. It led to an unbelievable number of deaths.

[Liz Hoffman]

You know, we did get a revolution out of it probably.

[Mike Pesca]

Well, yeah, we got, we got the 30 years war and we got arguments over everyone who had their own view of Christianity could now print it. I'm an optimistic person. I'm a. I'm a fundamentally optimistic person. And I also think humans will change and adapt. I just think AI is going to be unbelievably transformative in a way that we can't possibly comprehend.

[Liz Hoffman]

One thing I would, I have been thinking about is, is it will. It will hit the economy unevenly. So I think health care is a great example of this. If it were cheaper and easier to go to the doctor, we'd all go to the doctor more and we'd probably be better off for it. And by the way, we don't have. There's a massive shortage of clinical care workers and doctors and a bunch areas like that feels great to me. For example, I have covered finance, I covered M and A for a long time. If it were cheaper and easier to have like the AI models decide which company should buy which company, we would see a lot more M and A. I doubt it. Like there are industries that are just what we call demand constrained. And like for those industries, I don't think that is a pie that is going to sort of structurally grow and shrink but is not going to explode with AI. I could be wrong about this. We could get to the point where it is frictionless to sell your H Vac company. And so Goldman Sachs is out using M and A to do those deals and is hiring a lot more bankers maybe. But I do think that there are different buckets of sectors that are going to benefit from this in really interesting ways and ones where it is just going to turn into a jobless food fight for the share that remains some

[Mike Pesca]

gigantic portion of M and A activity. And economics is just the lawyers looking at contracts. And that seems totally, very, very well handled by AI.

[Liz Hoffman]

Totally.

[Mike Pesca]

What about the latest. So here's what my takeaway is from the latest charts and the latest guidance from anthropic and OpenAI. OpenAI not meeting its goals. I think maybe the consumer focuses on the consumer products. How many people sign up for this? But I'm pretty sure that the companies themselves, that's not their business model. In fact, I've even heard that the amount of debt that the companies had if they got everyone in every consumer in America to sign up for their products wouldn't even and cover the amount of debt they have. So it's not a consumer facing play, is it?

[Liz Hoffman]

I think it's interesting. You have seen, you know, OpenAI and Anthropic in particular were kind of birthed with different business models. You know, and you are right, this sort of we call B2C business selling to consumers was a very fun, sexy business model in the 2010s. It's what Silicon Valley was funding. We got Uber and then we got Uber for everything else. And. And you know, I think we really saw the limits of that because when those, when those venture subsidies go away and you actually have to have people paying for your product, turns out you can win in consumer. But there is generally one winner in consumer. Google won consumer search. Right. Bing did not. There are a lot more winners than

[Mike Pesca]

is Lyft viable by the way. I'm thinking of that one.

[Liz Hoffman]

You know, it has hung on. I mean it is worth a fraction of Uber and is much less ambitious. By the way, we had Dara, the CEO of Uber on our show recently and his future is not direct to consumer. It is B2B. He has this vision of a world where robo taxis are doing almost everything and that Uber is the wrapper that is literally tucking those robo taxis in at night. It's cleaning them, it's parking them, it's charging them, it's servicing them. And that their customers are in fact mostly not you and me, but are the big fleet owners. I mean you cannot get more sort of like mundane enterprise business services than managing like a fleet owned by Blackstone So I think you're right.

[Mike Pesca]

Sell that to Wall Street. They're like, oh my God, we could manage taxis. We could be like the Louis de palma of the 21st century.

[Liz Hoffman]

But no, but you're right. I mean, OpenAI started with this real Dr. Direct to Consumer model. Anthropic started with a real business to business model, selling to their services to mostly software developers. OpenAI is starting to realize that that was perhaps a mistake and they have too many eggs in one basket and are trying to build these coding tools and these enterprise tools. We'll see. But I think you are right. Like if everyone in the world uses your product, you will win. But if like you really have to get them all. And that goes back to a trust problem that's hanging over AI in general and OpenAI in particular.

[Mike Pesca]

Okay, let's say what is wrought by AI isn't totally disastrous. There'll be some winners and some losers, but also a massive amount of losers. And what the government's job is, is to address this. So I have two questions on this one. Have you come across anyone who is in a position to be a regulator or pass laws so it can be elect and elected or someone who's actually in the firmament of regulation who you find to be up to the task?

[Liz Hoffman]

I mean, you know, we'll remember the. A couple of years ago the Congress had all the social media CEOs come down for a hearing and at some point it was so obvious that they had no idea how these companies made money. I mean, just literally didn't understand the tech at all. So there is a massive gap. I do think there are some people who are thoughtful about this. There's actually an interesting congressional race happening right now in New York City where Alex Bors has become kind of the boogeyman for a lot of of the AI community because he would like to perhaps thoughtfully regulate them. We're talking more about, you're talking about how to sort of policy position ourselves ahead of this wave. You're starting to see universal basic income kind of become a sort of a mainstream talking point. This idea of should be taxing compute instead of labor going forward. The answer to me is obviously yes, but you're talking about rewriting a couple hundred years of the tax code to do that.

[Mike Pesca]

That. Well, I also think universal basic income is one of those things, one of those pie in the sky things that we were to, if we were to actually achieve it, it would only solve a tiny fraction of the problem. In fact, wouldn't even solve it would displace some of the pain. But people don't want to essentially live as pensioners or on fixed incomes or like the, like the passengers on the spaceship in Wally. Right. There's a whole, there's a whole model for what happens when say entire towns are living on disability, you know, government largess. And what happens to you intellectually? What happens to your worth? What happens to your drive? It's not that pretty a picture.

[Liz Hoffman]

No. You know, it's hard to know how much of that is that there's stigma attached to those benefits, that if they were truly universal they might look and feel different. But, but I totally agree. Like we have jobs because we want meaning and we find fulfillment in it. And you know that that's not gonna. I don't think we're gon into the top of. What is it? Maslow's pyramid. Sort of all living in the, in the leisure class, having, having deep thoughts all day, you know, Self actualization. Self actualization?

[Mike Pesca]

Yes.

[Liz Hoffman]

Yes, yes. No, I think much.

[Mike Pesca]

Well, we're not there yet. I mean we're like somewhere mid pyramid.

[Liz Hoffman]

You're still working pretty hard. You're working pretty hard.

[Mike Pesca]

That's right. I feel like even the pharaohs didn't get to the top of the pyramid. And of course my people came from the builders. Hey, hey, AI, tell me what's going on with this Musk and Sam Altman trial. I know that there are, you know, big personalities who hate each other. Clash of the titans. Lot of money for the regular folk. What should we watch out of this one, this lawsuit?

[Liz Hoffman]

I should say I'm not out there at the courthouse. There are plenty of reporters who are. But it is, I think, a little bit of a proxy for the both deeply personal grudges that exist in the AI community and also kind of the broader. What is this technology for the things we've been talking about? How dangerous is it? How should it be deployed? Musk has been testifying this week there is absolutely no love lost between him and Sam Altman and that crew. And in fact, Elon was the original founder of this thing. And it fell apart in kind of spectacular fashion. He took his toys and went elsewhere and really has a very for profit, very futuristic view of what he's doing with AI, which is mushing all of his companies together, SpaceX and Tesla and Xai and Grok and into one sort of of AI super company. But yeah, I mean it's crazy at the heart of these commercial disputes that you've seen both between Sam and Elon and between Anthropic and the Pentagon, they're fighting about contracts and profit streams. But what they're actually fighting about is, is this existential technology in the hands of people who can be responsible. And it is like one of the strangest uses of the legal system I've ever seen.

[Mike Pesca]

Yeah, but who wears the white hat? Who. Who represents safety in this particular trial?

[Liz Hoffman]

Well, in this one, probably nobody. Pretty clearly, right now, it's Dario Amade, Anthropic, who has sort of drawn a bright line about how he does and does not want their tools used and is in a world of trouble with the Pentagon because of that. But that is part of it, too, I think, which is that we don't understand this tech, but neither do the people making it. I mean, they're pretty open that they don't really know where this goes, that they are scared of it in some way. And that is just, like, really not reassuring. It is fundamentally a black box, and we don't know why. It does a lot of things it does, and the people making it don't even know.

[Mike Pesca]

But I would say I heard a guest on your show, maybe it was, I can't remember which guest, talking about Dario's stance, which was to say, we're going to draw some lines on smart weaponry as being, as much as anything else, signaling. Signaling that we're the kind of people who are the good guys in this. They want to wear the white hat.

[Liz Hoffman]

That.

[Mike Pesca]

But still, they are anthropic, and they are going full steam ahead in this technology that has, you know, I don't know if it's 20%, but some percent chance to really upend things.

[Liz Hoffman]

Yeah, the consensus, the latest sort of straw poll out of Silicon Valley was 10% that it will destroy everything.

[Mike Pesca]

So, you know, how could you even put a percent?

[Liz Hoffman]

Yeah, you take your odds. No, that was. That was a partner, a guy named David Ulavich. He's a partner at Andreessen Horowitz, and he was fairly critical of Dario. He said, yeah, you can say if you want to use my technology, you got to call me on the phone. But that reveals a pretty deep God complex. And there's a question of, should there be some moral center to this? Anthropic has tried to write that into its models. It has this constitution where it tries to draw the line. I mean, look, I don't make AI models for a living, and I don't have kids, but I suspect that it is a similar thing, which is that you try to sort of instill the values in the thing that you are developing and raising, understanding that it's gonna go out in the world and be autonomous in some way. And, you know, I think parents go through their lives a lot of anxiety about whether they've prepared their kids the right way. And, you know, I like to think that Dario and his peers are grappling with that too.

[Mike Pesca]

Do you think a Democratic Pentagon would have gotten in the same fight as this Pentagon did with Dario?

[Liz Hoffman]

I don't think that they would have taken the actual technical step that this Pentagon did, which was to designate anthropic a supply chain risk, and essentially threaten to put it out of business. No one, no, for some.

[Mike Pesca]

Some of that is because it's a loser. That's never going to work, is it? It's not a supply chain risk.

[Liz Hoffman]

It is, of course it is not a supply chain. It's fine to say that, you know, we don't. We're canceling the contract because we've. We've come to a disagreement that we can't get past. So I think in some ways this. This administration as, you know, kind of like wakes up and chooses the fight at all times. But I do think that a lot of this discussion is actually going to continue, whether there's a Democrat or Republic White House in 2028. A real consensus opinion, which I think I hold too, which is that if we don't win on this, China does, and there's going to be global models that are, you know, the fastest growing countries in the world are not. Not us. They are the global South. And. And China is very keen to put its models into those economies and like, who knows what kind of morals and. And, you know, freedoms are written into those models. That is a very real and pretty bipartisan opinion.

[Progressive Insurance Announcer]

Opinion.

[Mike Pesca]

Yup. So if we don't win, China will. But what is winning, by the way, I followed to some extent, I think Emil Michael, who's the undersecretary of defense for research, and he was making, I would say, a decent enough case for how frustrating it was to bargain or negotiate with Dario. This should have all already been taken care of. And he accused Dario of essentially grandstanding. And maybe that's true. Right? He, Dario, has to stick up for his people or brand himself as the white hat. So I'll say that. But my big question of, would a Democratic administration have done it? I mean, is it part and parcel of a more warrior, hawkish mentality to insist on them having the keys to their own weapons, including automatic kill devices? Right. Would even the okay, so we're not positing a word world where the guy who's in charge of the Pentagon is, you know, a hippie or someone who, you know, Bernie Sanders, the guy in charge of the Pentagon is still a military general. Would a democratically appointed military general say no, no outside contractor gets that say over our weapons?

[Liz Hoffman]

I think that's Pentagon standard. I mean, when Lockheed sells a fighter jet to the Pentagon, they don't get to decide what it gets used in. That doesn't seem crazy to me. This is obviously a more dynamic and so sort of constantly it's more of a two way street on what the tech is and where the tools plug in. But like, I think Democrats would probably have a bigger fight, I mean, hanging over all of this is that a lot of Silicon Valley got massively red pilled over the last 10 years and have become themselves more conservative, more hawkish, more America first. They are friendlier to this administration. They bankrolled a lot of Donald Trump's second campaign. And so I actually think there's an argument that aside from the, the sort of process piece that, you know, liberals tend to be sort of process fetishists. And I don't think they would have done what, what the Pentagon did here on the supply chain risk side, but I think that probably the rhetoric would have been, would have been stronger. I think the Democrats have, have less faith in these guys than Republicans do.

[Mike Pesca]

Oh, so do you think a different kind of messaging might have served the Pentagon's interests well if what they wanted was the keys to their own weapon system?

[Liz Hoffman]

I think the Pentagon should just find a defense contractor that agrees to their terms.

[Mike Pesca]

But I thought that only, only Anthropic had that. Now I guess they're letting OpenAI AI train on some of their, and OpenAI

[Liz Hoffman]

took, you know, parts of that contract. And actually I think Sam Altman sort of said he had a little, some regrets about how that all happened. But you know, I think a lot of, particularly a lot of these models are going to converge and become kind of off the shelf commodities in some way. And so yes, you can have, you can, you know, Lockheed and Raytheon both make great airplanes and you can give one this contract and won that contract. And that's just a procurement process. It's not all that interesting.

[Mike Pesca]

And we will be back with more of Liz Hoffman, including what can capitalism do to save itself? What can socialism do to ever discredit itself? That will be tomorrow on the gist. And now the spiel In 2024, the state of California overwhelmingly passed Prop. 36. Residents had seen a barrage of images of shoplifters marauding stores and had been inundated with numerous stories of retail theft. Residents also experienced themselves going to stores, seeing items locked behind glass, seeing stores closed. So what Prop 36 sought to do was to allow for prosecutors at their discretion to charge third time thieves of items worth over $950 to charge them either with a misdemeanor or a felony. There was also a provision in Prop 36 which was a major selling point for people struggling with addiction, which makes sense given that the prop was called the homelessness, drug addiction and theft Reduction Act. It's been almost a year and a half since it's been in effect and the results have been inconsistent. There is almost no debate among prosecutors, civil libertarians and advocates that because the treatment programs passed without any funding for the programs, they have been a big zero, wholly inadequate. But I want to focus on the theft reduction part, part of the homelessness, drug addiction and theft reduction Act. So I recently came across an article in the LAS the Los Angeles based news organization affiliated with local public media. The L. A S says you come to L AIST because you want independent reporting and trustworthy local information. Normally I get it. Not in this case. Story checking on Prop 36. Unnecessary crackdown on repeat offenders or a return to mass incarceration. So that's it. Those are our two choices. If it didn't really work to crack down on the repeat offenders, or if it wasn't totally necessary, we've now returned to mass incarceration. It couldn't have worked just ok. It couldn't have generally changed things that were kind of changing anyway. No, it has to be either total success or mass incarceration. Much of this extreme dichotomy rests in how the article by Frank stoles defines mass incarceration. Actually, Stolz doesn't really define it. He outsources that to advocates. Here's some quotes Supporters say one thing. Critics say Prop 36 has, quote, been a return to mass incarceration. And then they back this up with a quote from the executive director of the California Public Defenders association, quote, it really is a return to mass incarceration. All right, what are the facts? Los reports that in 2025, prosecutors filed more than 19,000 Prop 36 felony drug cases and 15,000 felon theft cases. Most defendants were released on bail pending the outcome of their case. 35,000 cases and 900 Californians have been sent to prison under Prop 36. According to the article I found different and better, more accurate stats anyway. The article then also talks about county jail populations that have grown since December, but it doesn't directly connect that to an increase in Prop 36. It includes a statistic from LA county where it says there are 1150 individuals in jail because of Prop 36 and that's a 9% increase in the jail population. So a 9% increase doesn't seem like mass incarceration. It seems like a small increase. We could debate if it's a necessary increase or a useful increase in terms of combating crime, but that's all they have for mass incarceration. You also have to take into account, well, what was the starting point? So let's zoom out. According to the Public Policy Institute of California, which is very well respected, sort of center, maybe center left. California's prison population April 2026. So now it stands at the lowest point in more than 30 years. California's prison population peaked at 173,000. Now it's at little over 90,000, a drop of 27% since December 2019. So incarceration is massively down. There recently has been a slight uptick according to or stemming from Prop 36, but that is credibly reported as a return to mass incarceration. That's kind of crazy. What it has led to is not mass incarceration, it's led to more incarceration. That's exactly what the law intended it. But it would seem kind of stupid to oppose a law whose very purpose is we're going to put some more people into jail in prison to impose that with the objection, no, no, no, you're going to put some more people into jail and prison. So the opponents, including Gavin Newsom, said you are not just going to put some more people, you are going to put many more people. It will lead to this concept that has been derided and decried, mass incarceration. And when it clearly didn't get there, here is the ever so clever rhetorical gambit. Just say it did. By the way, Prop 36 passed 70 to 30. So the people were clearly saying yes, we in fact want some more incarceration. And opponents are saying said then, still said now. Well, this is exactly what we predicted, mass incarceration. But it is clearly not the implication that comes through in the story as quoting the advocates on the anti Prop 36 side, but quoting them with credulity. The implication is that any increase in incarceration is inherently bad. And I would say it's not. It's not Happy. But it is a useful tool in achieving that other public policy goal, which is combating crime. So incarceration went from 170,000 in California to 90,000. I don't know. Maybe 90,000 is not the optimal floor for incarceration. Maybe the optimal floor is slightly higher, like 91,000 or 91,500, with that difference being made up of repeat shoplifters. And the reason I chose that number is that's about the number. According to the California Policy Lab, which tracked every prison admission from Prop 36 offenses, there have been 1957 admissions to California prisons. As of last month. Month drug cases accounted for 1.9% of prison admissions and theft cases for 4.1%. Over the course of the law's implementation, about 1% of people admitted to prison were there on drug cases and 2.5 on Prop 36 theft cases. 795 individuals, not mass incarceration. We could debate if they deserve to be there. If them being there is lowering the amount of shoplifting going on. And that debate is addressed in the article. It quotes the DA of Los Angeles who says the statistics are not in yet and that is fair. The ridiculous assertion of mass incarceration I think I've made clear now that I think it's ridiculous does not answer the question does the law work work? The article itself quotes the LADA but also throws some cold water on the question of if the law works. It conveys this stat, but crime was on the way down before Prop 36 passed. Violent crime, that's not a Prop 36 thing, fell 6% and property crime dropped 8.4% in California the year Prop 36 passed. It then dropped further. But the implication is is the already on the way down. I always get bothered by this because what it really says is the only way to judge that a law works or that a punishment works is if the only effect were to be seen immediately after the punishment. So if shoplifting goes down the year before the law passed, the year after the law passed, and the year after that that how would you be able to disprove that the law had some effect on shoplifting? I'll throw another complication into the mix. I read that article's stat. Here's another headline from the Public Policy Institute of California from last year. Overall crime fell in California last year, but shoplifting continued to rise. So we can't even be sure if the subcategory of property crime called shoplifting, which this law Prop 36 was meant to address. We can't even be sure what happened. But I'll tell you this. Crime ebbs and flows for a number of reasons and crime is falling across the country by a lot. But that doesn't mean the shoplifting law doesn't have a positive effect in discouraging would be shoplifters. I mean, think of it this way. Even if the base rate of shoplifting was a 10% decrease year on year, year without the punishment, it might have only gone down 5%. And it might be the case that the 10% decrease, which is followed by, let's say, an 8% decrease, that 8% decrease was in large part attributable to the law, the necessary law that puts more repeat shoplifters in jail at the discretion of prosecutors. There are few other government interventions that I can think think of where the promised outcome was achieved, where the law was in fact enacted, where the reaction to the law was exactly as predicted, where 70% of the people were in favor of it, and where reasonable people are saying no, it totally failed, it's a disaster. If all of our failed disastrous government initiatives work this way, the world, or at least America would be a much better place. Not a perfect place. There'd still be shoplifting. Quite sadly, a lot of those shoplifters would still be going to jail. I don't know, maybe there's something going on with the language. Maybe amongst activist circles there is a line that all incarceration is mass incarceration or that we already have mass incarceration. So any more incarceration just adds on to the mass incarceration. But if that were the case, why was it always framed as will lead to mass incarceration? So I don't know if goalposts are shifting or if the ball is being hidden. I just know in terms of basic facts and basic concepts that I try to understand, there might not have been a massive change in shoplifting, but there might have been enough of a decrease. And there definitely wasn't mass incarceration. The only thing mass was, I'm going to say the delusion in thinking that's what happened. That's it for today's show. Cory War produces the gist. Kathleen Sykes runs the gist list. There's a good post op about Bruce Springsteen and building things@mike pesca.substack.com that's free. That's free for all of you. Jeff Craig runs How to and Ben Astaire is our booking producer. The stylish Michelle Pesca is COO of Photo Shoots and Peach Fish Productions. Thanks for listening.