This is Dalton + Michael, and this feels a little different. We're going to do a little bit of a 2025 in review. Hot takes.

A hottest of takes, Michael.

A hottest of takes. Dalton was inspired by the Colin Jost, Michael Che, end of the year SNL skit where they give each other extremely challenging jokes to say out loud that will get themselves canceled.

Let's not do that.

We're not going to do that. No. But it's a cool idea. Maybe next year. Anyway, where do you want to start? What's the first hot take topic?

Let's start with San Francisco, Michael — let's start at the hottest of hot takes. What's your hot take on the status of the city and politics and the whole shebang?

My hot take is, if you're reading it on Twitter, there's a greater than 50% chance it's wrong.

And what way?

Just wrong. Just assume anything that is trending about San Francisco, either positive or negative on Twitter, is wrong.

So is it better? You're just saying it's different.

It's just different. There's no signal. The signal that you're getting on the things that are memeing — you should ignore that signal. It's the fuzz. It's not real signal.

I'm sure they're all going to listen to you. You solved it.

I think my second hot take: people give politicians way too much agency. I don't care who the people leading San Francisco were. If you have COVID, if you have the emptying of downtown's office spaces, and if you have a defund-the-police movement, you're going to be in a challenging situation. I don't care who's leading a city — if you're emerging from those things, you're going to be in a better situation.

All too often, I feel like humans need viral narratives to explain simple things. Yes, there's more work to be done and blah, blah, blah, all these things, but those two trends were generally going to hold. Probably the people running the city are just trying to do their best with the general situation that they're in.

What's that phrase where the king goes down to the thing and Charles is, I'm going to show all of my citizens how powerful I am. And I'm going to command the tide to go out. And the tide doesn't move. And he's like, that's my power.

I don't know that one.

Those are my moves. Anyway, I think that's my hot take on San Francisco. The Twitter narrative is basically completely incorrect. All right.

What do you got for me?

Hot take depopulation. Is humanity going to make it, Dalton? And will it be too few people or too many?

Yeah. The graphs look bad.

They look really bad.

I remember a long time ago, when I was an undergrad at Stanford, I remember seeing a building on the campus. There was a dumpster next to it. It was full of books that had clearly been thrown out of the building.

Interesting.

Of course, this is my natural curiosity — why would you throw a bunch of books and journals into a dumpster? It was all from something called, I think, the Stanford Food Research Institute. I looked into it a little bit at the time, and basically it was a whole department slash think tank that, as far as I could tell — I could maybe get some of the subtlety wrong — was basically encouraging people to stop having children because there wasn't enough food to feed people in the world. There may be more to it than that, but I guess Stanford shut that down, which is why all the journals they were studying, the theory of this, ended up getting shut down. Obviously this theory became very popular in certain countries.

I might argue it became part of public policy globally.

It did. I think if you read a lot of books in the 60s and 70s, this was the population bomb. If you were really into environmentalism and trying to save the world, trying to slow the rate of population was the coolest thing you could be. Wow, is that different now?

And by the way, I think the evidence that this wasn't the problem was a bit hiding in plain sight.

Well, it's almost like it was so effective at raising living standards in places that it all kind of solved itself. Basically the abundance of raising the standard of living fixed all this stuff.

It was almost as if we have an equilibrium and we thought we had to manage that. That's a very human perspective. Obviously we have to manage our population equilibrium. We've never done that ever until this point. We've been around for a while, but obviously now we're smart.

It looks bad. It looks bad for Japan. I don't know. Look, it's not my line of work.

It looks bad for America. I see these graphs where it's like China's population is going to kill everything, right? And then America fine. And then it's like America fine includes immigration. And you're like, well, great. That's not repopulating the world. That's just us opening our doors, which I love immigration, but it's not solving a larger problem.

I want to give PG a little bit of credit here. I think that going to Yale and seeing people talking about ideas who didn't always seem that smart, I think I just dismissed the power of ideas a little bit. I think I dismissed academia a little too much. I remember PG talking about some things around, hey, this stuff happened. I just feel like, oh, we just ignore college campuses. Just ignore that. Yeah, whatever. Who cares?

There are some ideas that when they take hold, they can have a lot of power. I think this is one of them. You would assume at some point someone would do the math and be like, no, no folks. I think there are still people who aren't having kids today because they believe that we're going to overpopulate. It's about environmentalism. Wow, that's a powerful idea.

What's crazy is the solutions are all kind of random from government. Wow, we'll give people a tax credit.

The solutions are not good.

The solutions are basically tax credits. And I think I read that in all the attempts thus far to do those, it's like 0% effective.

I read that basically when you do it, you basically push up the people who are going to have kids into getting the incentive. And then you have this falloff because, well, I had my kids early. I'm shocked that government isn't great at figuring out incentives for humans to do that. Especially things that involve watching something for 18 years. That's a big one.

All right, so that's one. Next one, AGI Dalton. January 1, 2025, did you think we were going to be closer or further from AGI than you believe today?

I think it depends on how you define it. I think the models improved faster than I would have expected January 1, 2025. Just in the past two or three weeks, there have been a bunch of new models that dropped. Literally in the past two weeks.

Google's coming on strong.

Everyone put out new models and they all seem pretty good. And the other funny thing about the new models is it takes a while to know how good they are. You can't just run the benchmark and automatically know if they're good. And so I think the rate of improvement is higher than I expected. But I'm not sure, depending on how you define AGI, this is way better than I would have expected at the beginning of the year. I still think the timelines are similar to where we started the year.

I'll answer the same way, but I think because I was skeptical of AGI in January and I'm skeptical now. I think that AGI is such a cool and almost viral idea and such an impactful idea that it can suck up all the oxygen in the room. I remember talking to a founder in August and she was like, oh, I talked to all the people working on this stuff. And they're like, yeah, we're six to no more than 18 months away from just superintelligence. And they were sitting there just imagining what the world would be like, as if they're making plans. Oh, your near-term plans, right?

It's weird. I feel like that did not help this founder at all. One, I think that those people while making their plans were not operating their businesses as if AGI was six months away. So it was this weird kind of duality that's hard to capture. But two, I think I'm always skeptical of ideas that take all the oxygen out of the room.

I think this is related to our next hot topic, which is self-driving. Let's start with what I think the parallels are. In 2015, it felt like we were at the precipice of self-driving working, and I spent a lot of time being like, man, my kid's never going to need to learn how to drive. Remember, we had a colleague who didn't know how to drive, and he pre-ordered a Tesla full self-driving in 2016, because he was like, oh, well, this is going to self-drive in 2017. I think we were living in an environment where it really did seem like ubiquitous self-driving would happen in 2017, 2018. Your former co-founder Kyle was the founder of Cruise. I think we were really drinking the Kool-Aid. I think with hindsight, we were totally right. I think in hindsight it took longer than we thought. The thing that we thought was going to happen in 2017 is now happening in 2025.

The tricky thing is like, I would say it's not, it's happening. It's started to happen.

It's happening in the Bay Area. Yeah, across the US. I took a, I'm in the Tesla RoboTaxi Beta. I took one here. And I take Waymo constantly.

I would still give that a maybe. What is the ratio of tele operators to cars on the road at Waymo?

I still think this would have been within the range of what we would have satisfied us in 2017. I'm not saying I know the number. You don't think it is? I think it is.

I think that in 2015-17, no, no, I think I thought self-driving was going to be a solved problem. There are no humans in the loop when I'm doing a Google search. And then to your point, and then of course, have we rolled it out to yet?

And actually there was a technical bottleneck. Again, this is the other parallel. My understanding from people that work on this when I've spoken with them is that they ended up deleting a lot of hard-coded C++ rules and it's just all ML stuff that wasn't even invented. Computer science had to catch up — that's my understanding. Again, maybe someone in the comments will disagree, but that is my understanding from people that work on this stuff.

I think that's what kind of frustrates me about these topics is that if we're directionally correct but the time frame's off, it really impacts how you should operate your life. There's a meaningful difference between an AI model that can do a lot of work for you, that makes you a superpower. I mean, let's say an AI model or a foundation model company that can make every person feel like they get the value that cursor gives an engineer. There's so much distance between that and super intelligence.

I wouldn't be surprised if we had a whole new job of tele operators. And there wouldn't be hundreds of thousands of tele operators that are helping manage a self-driving fleets as they expand across the country. And we would call that self-driving. And it would work as a user, even if you were like, right, I'm not driving. To me, there's a parallel there, right? Like versus like AGI super intelligence, I think we have to define it kind of across the line. And I think like we could be way further away.

I think that's fair. But again, I think the parallel is the crazy utopians. I think we're right about self-driving. I would basically make the bet that the wacky AGI people are right just on a longer timeline. Do you know what I'm trying to say? I think they'll probably be right about every wacky thing we're hearing, I think is going to end up being true. That's my bet.

I think long enough timeline, I would completely agree with you. I just think that when I start seeing people making strategies about their startups or making strategies about their life. Assuming we're on an 18-month timeline.

Delaying having kids, say it.

Yeah, I'm like... The timeline caveat's a large caveat, I would say. Yeah, I would say.

All right, what's the next one? Space. This is an interesting year for space. We have seen Starlink basically go mainstream. We've seen a lot more space startups being funded. We've seen space defense kind of take a little bit. What's yours?

Yes, I think I put out a YC video request for startups about this, but basically as a side effect of SpaceX working, the cost per pound to get something into space is dropping rapidly. It's like Moore's law, where remember the number of transistors in a chip was doing this 📈. The cost to send something into space is doing this 📉. So it's getting exponentially cheaper. Now you can do all sorts of wacky stuff because you can get things in space. Sounds like the SpaceX IPO is real. So I think this is a really exciting time.

My space hot take is very similar, which is we haven't experienced cheap launch yet. People who've been in the space industry for a while are like, oh my God, we finally have cheap launch. I think we're going to look back at this time and be like, God, that was expensive launch.

If you told me we've reached the iPhone moment where it makes sense to start building iPhone apps, I would believe you. If you told me that we haven't yet, I think I still might believe you. I still might believe Starship is the iPhone moment.

I'll give you another one that was interesting though. What do you think about space companies having to compete with SpaceX for launch space? It's tricky. I've been in multiple companies now that have gotten bumped off their launches, and I'm like, oh, can you negotiate? Who was it? And they're like, it was SpaceX. It was Starlink. I was like, yeah, I guess you can't really negotiate, can you?

So what you're saying is we should buy SpaceX stock at IPO.

No, no.

This is not an investment.

Another big topic from the COVID days. What's your hot take on in-person versus remote?

Obviously in-person is better, and it's depressing how much of big company stuff is putting the toothpaste back in the tube when they went remote. I do think there's stuff that stuck from that era that was probably good — people are way more willing to do Zoom meetings now. I think the norm has changed in some ways in a positive way. Imagine how weird it would have been to try to do nothing over Zoom.

Yeah, yeah, yeah, that would have been...

I'm glad I can just hop on Zooms and do certain types of meetings.

I agree.

That's way better.

Real-time communication chat got way more popularized.

Yeah,

I am torn about this in-person versus remote. I think this is what I've come to believe. I think large companies are always remote. You get over some number of employees and the chances that the three people you work with most closely have their desks next to you basically goes to zero. I never really thought this was a fair debate because I think the debate was really for startups. It always seemed obvious that you should be in person. I felt like this was a fight that we had to fight for a while, which was just kind of silly, and now it's like, okay, can we go back to normal? Have we not run this experiment involuntarily? To me I think the shocking thing was that a bunch of startups that did not want to run like big companies were copying the best practices of big companies. Where have we left.

Are we in a bubble Michael?

I think that as long as there's another bubble that keeps whatever bubble's going inflated.

Yeah, there's always going to be a new bubble.

Yeah. I'm going to say no. And I'm usually the skeptic. I don't think we know the amount of impact that even today's AI tools can have in business. I think we were so early in the adoption cycle that even if all AI development stopped today, we would still get tons of improvement because adoption is very, very small.

I think my nuanced take is the following. As a founder, I learned that debt is bad because debt causes startups to blow up. PG has talked about this, right? This is one of his evergreen topics — never put debt on your startup. It feels like free money. It feels like cheap money. There are all these covenants and the contract that comes when you take debt so that when debt goes bad, it goes really bad and you lose your company.

Super fast.

Okay. So debt and startups is like a bad mix. And then if you look at prior bad economic cycles, debt is in there somewhere. I think if you go all the way back to the 1929 stock market crash, people were taking out loans to buy equity. So taking out debt to buy equity, I think we can say is probably a bad idea if you don't know what you're doing.

The great financial crisis was very similar.

It was debt driven. And so my nuance here is of all the things that I see that kind of concern me, it would be debt issuance to finance data center spend, to finance buying graphics cards. And that is downwind of all the stuff we do. Again, we're not in that business, man. But if you look at the whole machine, there's a lot of debt that's being issued right now.

If we were to reframe it, like was the internet a bubble? No. Was there a bubble in the deployment of fiber optic cable? Yes.

Did the people in the business of laying the fiber optic cable lose their ass? Yes, they definitely did. And we as consumers benefited from the fiber optic cables getting laid. Our startups benefited from these things. We're all downwind of it. But the actual people that were in the business of financing that fiber optic cable, that was bad.

I am quite positive that somebody is over building or inefficiently building at some level of the AI stack. Like it seems like almost impossible. Yeah, wouldn't be the case.

That does worry me, but that doesn't mean that this stuff doesn't work and isn't solving problems for customers and all that on the user-facing stuff. But I do wonder how much debt is actually in this thing. That worries me, man.

I mean, from a perspective of a participant in the stock market, those kinds of things can have short-term negative impacts on the stock market. But in terms of the...

Well, and I do wonder, Let's say all the AI stuff works, which again, I think it will. The value of a lot of the legacy software companies that are going to get killed goes down. There's going to be some churn, man.

It's so funny. You're defining bubble in a how-someone-might-experience-the-world way. I think I'm defining it in a very narrow way. But yeah, could there be a lot of AI-caused turmoil in public markets? Probably there will be. Sure. If you define bubbles like that.

Yeah, who might use the actual definition, Michael?

If you define this as like, is this a hoax?

No.

Like, you know, when we were giving out like mortgages to people who couldn't afford mortgages because we could resell them before anyone would figure that out. That's like a hoax. Like that's, there is no value creation happening. I think in this situation, we definitely have value creation, but like, Probably market turmoil could be caused for all kinds of reasons. Right?

It's fair. Anyway, of all the things I've seen in the past, in 2025, what has worried me — some of this data center stuff seems a little scary, man.

Yeah, public market turmoil.

All right. And the last one, this is probably secretly my favorite, abundance versus affordability. Let me set this up. The Democratic Party is searching for a message. The Trump era theoretically is coming to a close or a change. And it seems like there's two competing messages. There is more of a moderate abundance. Can we make government work better? And there is more of a progressive affordability. What are the things that happen in the economy or amongst business or even in the government that makes it harder for you to pay for everyday things. How do you think about this dichotomy here?

Technology has consistently delivered us abundance. If you just think about all the cool stuff we get — smartphones and good healthcare and cheap clothes — all this good stuff, we live in an age of abundance and it's only increasing.

It's funny you say that because I think somebody that was examining the fossil record would obviously agree.

Yeah, if you're like, what was life in the 1930s like? I don't think they had all this stuff and I don't...

Objectively. But what's weird is — isn't it interesting — what percentage of the population of America would say I feel like I'm living in a time of abundance?

I think stuff used to be cheaper. Again, this isn't my area of expertise, but the housing stuff is clearly an affordability type. There is not an abundance of cheap housing. I think it's easy to look at the past with rose-colored glasses and gloss over — remember when gas used to cost whatever per gallon and when a house cost $100,000 — and not admit that there's a trade-off, that you had to live without the internet. There are all these other trade-offs. I think it's easy to forget that there was a trade that we are all implicitly taking. Maybe some people would say, I don't want to make that trade. I'd love to go live in 1986 again, right?

Well, you go back to the physics. A big trade is the number of employed women in the market and the workforce. There are big trades.

I think what scares me about this question is — and this is what both scares me but comforts me about politics and government — affordability is clearly the short-term campaign message that is a better campaign. Like if you ask me, hey Michael, you can run and they'll be like, your banner can either say abundance or affordability or like done, affordability. Abundance is clearly the good governance. Like if we actually want our government to deliver goods and services, we should be doing the abundance work.

I think the problem that I have, but then also the hopeful side is like, I just don't think we do long-term planning well. I think abundance is like, hey, you know, if you eat your vegetables, when you're 70, you'll live another 10 years. And I think affordability is like, hey, do you want to go to the movies and like do something like this weekend? And I think that as humans and therefore a society's built of humans, we don't do long-term well. However, we made it. Something about this like what seems like awkwardly short-term optimizations...

seems to somehow keep working.

Yeah. And so it's like I can sit back and judge academicians like, well, you're not, you're not eating your peas and carrots. But like, maybe we should take a second and not shit on a system that is kind of working, right? Like, I think that's when my experience just even locally in San Francisco, like at first you're like, you do it this way? Like, how could that work? And then you're like, well, it is kind of working. Like I was looking at like the Department of Public Works, like they pick up tons of trash every day. If you were to say, Michael, would you design it like that? I'm like, maybe I would, maybe I would. But like, it's kind of working. Let's let you shit on it. Cause like, no, I mean, that's not the way I would have done it.

So I don't know for this abundance versus affordability — I'm kind of like, oh man, my heart is in abundance. But I think my brain is in affordability and that's what scares me a little. My brain is in the campaign message over the governance. That's a little scary for me. But maybe that's just a deep-seated lack of faith that we will do good government quickly.

We're talking about 2025. What does that mean for 2026? What does that look like in practice?

That means I believe the Democratic Party will have a populist message that bad people make things too expensive. Just like I think the Republican Party has had a bit of a populist message that immigrants make the country worse. It's a little scary that we're going to have two populist messages fighting each other that I think are kind of hiding the real facts. But it's not the first time you've had populist messages. Yeah, we've gotten through this stuff before. When people are like, oh my god, social media is making things horrible, America — there's never been a worse time in America...

I don't want to be that guy.

I know these are problems, but we've dealt with worse problems. When you're down, it's hard. Especially for a process-oriented person like me.

Yeah, you're on the front lines. I'm glad this is not my... Yeah, I don't have to go fix this one.

Well, I guess that was 2025. We'll see what happens when AGI takes us over. This will be produced by AGI in 2026.

They have enough footage of us. They can definitely replicate it.

They do. They do. And I've been saying thank you every time I use a foundation model. So we'll be safe.