Network effects, AI medicine, and the fight for free speech

In this episode, Reid and Aria are live from New York as they unpack why predictions about the “death” of San Francisco and New York keep missing the mark and how network effects continue to anchor these cities as the world’s leading tech and finance hubs. Reid also shares advice for young founders choosing where to build and explains how to align your startup with the right economic network by breaking down lessons from companies like Shopify and Spotify that scaled outside Silicon Valley. The conversation then shifts to the future of AI in biotech as Reid offers an update on Manas AI and why curing disease hinges on regulation as much as technological breakthroughs. The episode closes with a candid discussion on media, political pressure, and the dangers of “pre-obeying” authority. Reid reflects on free speech, institutional courage, and what a volatile post-midterm landscape could mean for American democracy.

• 0:00Speaker 00

  Reid, it is delightful to have you here in New York in the Notion office. First question is actually about cities. So I feel like for the past few years, everyone's been talking about the death of San Francisco, the death of New York, sort of anointing Miami and Austin as the next tech hubs. The data that we've seen from Silicon Valley Bank says that actually, over the last three years, San Francisco is the only city that has seen company growth in terms of VC-backed companies being started. And this past January, David Sacks said, and I quote, "As a response to socialism, Miami will replace New York as the finance capital and Austin will replace San Francisco as the tech capital." Why do these people keep saying these things? And why do they keep being proven wrong?

• 0:45Speaker 00

  Or do you think something's going to change in the next five, 10 years?

• 0:49Speaker 01

  It's certainly not going to change. And I think that the question is, you know, Silicon Valley and New York are not the world organized according to how David Sacks thinks they should be organized. So if he uses as a proof point Miami's taking over of the tech capital or the financial capital as a proof point for his beliefs, which of course he won't, then he would have to acknowledge he was wrong. Because the actual thing about these things are they create, the cities and economies create network effects. And people who understand network effects, understand how they have persistence, growth, compounding. That's one of the fundamental things about what made Silicon Valley Silicon Valley.

• 1:32Speaker 01

  And I'm less knowledgeable about the New York financial scene, but I suspect that it's the same way there as well. Now, as Americans, we'd like to have as many different tech-growing hubs as possible. The fact that the others aren't growing is not a fact that I take any joy or delight in. I would prefer it to be 10 or 20 different cities. As a matter of fact, I try to help that in various cities in the US and also in other countries that I think we have values alignment with, like for example, London and Paris and, you know, Rome and Milan and other places as ways of doing this. So I think it's a, it's a good thing if most of these, these things are the case.

• 2:15Speaker 01

  You know, I think that the, we want as many of these tech things as possible. But it's also one of the reasons why, like, you know, the, the death of Silicon Valley is something that's been said for the last At least 20 years that I've been tracking this, Peak Silicon Valley, et cetera, et cetera. And all of that comment doesn't understand the fundamental strength of network effects. Networks of talent, networks of capital, networks of knowledge. And it draws it in, it is part of what leads to blitzscaling. And so the Silicon Valley bank data does not surprise me at all with the Silicon Valley growth. Now, it does say we need to focus some more on these other cities because they should have growth too.

• 2:59Speaker 00

  Yep. And so if you're a 25-year-old founder today, well, I feel like you're gonna say San Francisco no matter what. So I'm gonna say, if you're a 25-year-old founder today, you can't start a company in San Francisco. How should you choose where to start? Proximity to the industry that you're trying to disrupt? Your own network, universities, vibes, a city on the rise? Like, what should you be thinking about when you're choosing the city to house your startup?

• 3:24Speaker 01

  Well, fundamentally, it's the network, the economic network, or in LinkedIn terms, economic graph, of the city that fits the industry company that you're going for. And so, for example, if someone says, hey, I'm moving from New York to San Francisco to create a new finance thing, like that's pure finance. Now, tech is important on everything, so Silicon Valley and San Francisco helps on that. But it's like, well, actually, the finance industry is here more. Be a little bit more assertive. I'm going to Silicon Valley to start a fashion brand. That seems like, like, an impossible mission. So where's New York or LA would make sense? So it's where the network for... And how do you know where the network is?

• 4:06Speaker 01

  Well, it's kind of where the scalable companies emerge from, where they, financing emerge from. Where, where when you're hiring talent, because you use your scale, the talent, including how to go to global scale, because of real relevance, blitzscaling is always going to global scale, is where are the networks best for doing that? Now... Sometimes there's like nowhere, then it's like, okay, well, think about talent networks, think about capital networks. So it could be, for example, that something that's near university with a bunch of raw talent, a bunch of innovation. It could be that if you're creating something that's radically new or different, being somewhat askew works. And like, for example, part of like the way to say, hey, like, I always take things like Shopify

• 4:48Speaker 01

  and Spotify as interesting lessons because they said, okay, we know that we don't have the benefit of being in the Silicon Valley network. But how do we play this technology game? Like which technology games work as a kind of a, we learn from, but we build separately in order to do. And frequently that's build in a different direction. Do something like, for example, you know, with Spotify, it's like, well, in order to get it to critical mass, we had to first just guarantee the record labels Scandinavian revenue and could do it within that focused area and then get it to scale. Similarly, because Silicon Valley kind of thought, oh, all shopping is owned by Amazon, you could play the long game compounding in a network effects with Shopify because

• 5:31Speaker 01

  no, no, actually, in fact, I'm betting on the network effect of the internet relative to e-commerce and I'm building up a connect culture around it. But like, for example, Toby Lutka, who's one of the world-class anywhere, not just Silicon Valley, but anywhere founders and CEOs, and look, I'm learning intensely from it. He does board meetings there, you know, some of his board meetings, he reads constantly. He makes sure that the team is, is drawing from the knowledge base of Silicon Valley as much as he can to build his alternative path and build something that's, that's, that's extremely successful and competitive elsewhere.

• 6:09Speaker 00

  Awesome. So talking about another industry, let's talk about biotech. So you have AI, which is compounding so quickly. And we're sort of at the intersection of these two different curves. And one curve is AI has been getting so much more efficient, so much better and so much cheaper over the last few years. But then you have the biomedical curve and it is heavily regulated. It is full of failures. You have few. And it takes a long time, even with the introduction of AI, it still moves much more slowly than the world of technology. And you often talk about sort of the difference between the world of atoms and the world of bits. And how do you get technology to help the other grow more quickly? So as many of our listeners know, you started with Siddhartha Mukherjee, the company Manus AI.

• 6:54Speaker 00

  Can you give us an update on how that's going? What is the latest that we can expect from this company that hopes one day to have cures for cancer?

• 7:02Speaker 01

  One of the things that Sid and I talked a lot about was where should Manus be? And because of the fact is, is the classic Silicon Valley approach to, you know, kind of the pharma, biomedical, etc. Tends to be too oriented only in the world of software and bits. So like, you know, hyperbolic examples are, oh, we're just going to create new drug research or AIs and they're going to do everything. Or we're just going to create everything in simulation and the simulation will output everything. And by the way, if those things work, that'll be amazing transformations. It was like, okay, well, we need to have this new approach that is not just bits, but kind of biological atoms, right?

• 7:44Speaker 01

  So it's not just, you know, physical atoms, but biological atoms are kind of that midpoint. They're, they're dynamic, they're somewhat programmable. They have, you know, the language of biology, the language of cells, the language of chemistry in terms of how it's operating. And part of the reason why Sid and I kind of thought that the time is now for this specific approach was that we were like, okay, what are the challenges? That in development of what is essential for human well being and prosperity is like, you know, curing cancer kills everybody, kills young, old, men, women, you know, races, nations, like everybody. And yet, you know, our solutions are very, very barbaric and old school.

• 8:29Speaker 01

  Now, if we catch it early enough, we can do a simple surgery in a number of them. Not all, many not, but like you frequently can't, it doesn't happen that way. Or is that okay, now we're going to chemo or CAR T? Which is basically, we're going to.

• 8:46Speaker 01

  Try to kill you and the cancer at the same time and hope that we kill the cancer, you know, first. It's like, okay, there's gotta be better ways of managing all this. And the reason, the insight that's adjacent to the classic Silicon Valley way of doing this is say, well, AI is like matching these search problems, right? It's like a Google for what should the right, you know, kind of small molecule be that could be a, you know, a solve, a cure, a therapy, et cetera. You can also look at it as kind of like the intersection with generative AI. It's like, okay, what would be the, like, you know, in your ideal universe, you would go and say, I would like to, you know, solve triple negative breast cancer, and it'll go, and here's your molecule, and every input-output

• 9:30Speaker 01

  is kind of a, yes, is kind of a way of operating. Obviously, the models are nowhere close to being able to do that, other than pure hallucination. And so it's like, okay, what would be the things that are reconfiguring this that allow us to use all this amazing new scale compute, scale learning, et cetera? And so, you know, what we've done since we've kind of announced is we've hired a whole team, we've identified, you know, problems that go all the way from kind of like, how do you create new kinds of generative models that are the best predictive for therapeutic molecules? They're not like, you know, which is great, we use it, alpha fold, you know, to say all protein folded, because we don't,

• 10:15Speaker 01

  what we care about is therapeutic molecules. We care about what makes a difference for the human else. So what's the set of things that we, you know, said, okay, part of, one of the things that everyone knows is, you know, kind of limitations on data. So we've got multiple solves that one of them, we've been public about, which is a partnership with Schrödinger, where we go, okay, we use the fact that they've encoded a lot of chemistry and physics knowledge to configure the data set for, you know, how we find those molecules. And we've got a bunch of others. And then we've also got thoughts about, you know, how it goes through clinical testing, how it goes through to business models, because you have to look at the whole field. And it makes it much more complicated, makes it slower, but...

• 10:58Speaker 01

  Now, here in New York, we have a whole company that's doing it. And, you know, I don't know exactly when we're planning on being more noisy, but but but progress is in full motion.

• 11:07Speaker 00

  But it's exciting. So some of the heads of the Frontier Labs have said that they expect within 20 years, practically all diseases are going to be cured because of AI advancements. So do you agree with that? And/or, and I know you hate predicting, what's the end number of years you think is going to come before the first sort of AI drug that cures a disease?

• 11:29Speaker 01

  I think the number of years before AI drugs cure disease is probably more driven by FDA and other processes than it is like creation. Like, I think the technology was probably, like, already in our earliest prototyping phases, we've actually, you know, filed some INDs, have some things that are prospective patents, other things. And that's just us prototyping our technology to get it, to get the the factory running. So I think that the the notion of that there will be AI discovered, invented, found, you know, drugs, I think is very small n years.

• 12:11Speaker 00

  Great.

• 12:12Speaker 01

  But the process by which we get that into clinical trials, the process by which we get that into affecting human health outcomes, that is also a key part of that. And that's part of the reason why it gets harder to do now. If you said, do I think that it's completely feasible that because of AI techniques, all current human diseases will have cures, or the vast majority? I think the answer is, that seems very plausible. It's possible that there's a drug or there's a disease or two that just because something structural about how biology works, it's just like there's no fix for it that isn't also like mortal, right? Or something as kind of equivalent.

• 12:54Speaker 01

  Like, it just doesn't, there may be some that just don't have that infinite solve capability. And it's also that, you know, diseases evolve. So just as we

• 13:05Speaker 00

  We don't know what the disease landscape is going to look like in 20 years.

• 13:08Speaker 01

  So the all disease cured. Unclear, but I think, look, the advancement of like part of what we the part a major part of human progress over the last, you know, 50 plus years has been like, like solving diseases through, for example, vaccines and other things, which all of the biomedical folks, you know, kind of know and do. It doesn't mean that, you know, they say, well, you said there were no vac, there were no side effects for vaccines. No one smart ever says that. Of course, for some individuals are side effects, just as like for some individuals are side effects for aspirin, for some individuals are side effects for, you know, kind of eating

• 13:50Speaker 01

  a particular, like, you know, shellfish and so forth. There's of course side, but the question is, is, no, no, for a vast majority, this is life saving. And for some, there's some side effects and then you try to navigate that, that, that's essentially the statistical trade.

• 14:04Speaker 00

  It's really hard for people to wrap their head around life expectancies in 1900.

• 14:10Speaker 01

  Yes.

• 14:10Speaker 00

  And how radically different they were. And that has changed because of vaccines, public health, technology, like it is pretty wonderful. And we just forget honestly that what it was like to live 100, 120 years ago.

• 14:23Speaker 01

  And I think AI is the next phase of acceleration.

• 14:26Speaker 00

  Absolutely. So to end the episode, I want to switch gears a little bit. Recently, we got a very 2026 sentence. Stephen Colbert says CBS wouldn't let him air an interview with a Democratic Senate candidate because they were scared of the FCC. For those of you who are listening or watching, and you haven't seen the Colbert episode on this, please go watch. It is hilarious and also informative, but it's a little scary. What happened was that Colbert taped an episode with James Taralrico, who is running for Senate in Texas. He is one of the Democratic candidates. And the CBS lawyers called Colbert and said, you cannot run this episode citing the FCC's equal time provision because Brendan

• 15:07Speaker 00

  Carr wanted to change it. And so Brendan Carr did not himself call CBS, but CBS was pre-obeying. They were guessing about what was going to happen in the future. They saw the Trump administration get mad at various media outlets, threatened to pull licenses. So again, this is a pre-obeying that in some ways is a little scarier because it means that people aren't even fighting against it. Luckily, Stephen Colbert said, I don't care. And he put the episode on YouTube and now. I think as of when I saw it, it had 10 million views and 60,000 comments. People were really pissed that censorship was happening.

• 15:46Speaker 00

  So, do you think companies are afraid of retribution from the Trump administration? Is this a normal course of events? And what do you think about sort of the pre- obeying versus actually the administration cracking down on this?

• 16:01Speaker 01

  Well, it's precisely part of the autocratic playbook to create fear in citizens, but also in organizations like media organizations. Because if everyone stands up and disobeys, then it's very hard for them to do something. So it's kind of the make an example, do something. This is the whole Jimmy Kimmel, you know, kind of early way of doing it and saying, "We can hurt you very badly." And so everyone goes, "Okay, well, don't hurt me." Like, maybe you hurt somebody else, but don't hurt me. And so, and this is precisely where I think as citizens, as leaders, it's like, look,

• 16:42Speaker 01

  when you're feeling fear for this is a time to kind of show courage. And I think that, you know, we all have a responsibility to humanity first, and society, and then to business in terms of doing it. And so, how do we have a responsibility to society in doing this? Now, I found the Kimmel stuff bemusing because, you know, maybe for the first time ever, I found myself on the same side as Ted Cruz. I found myself on the same side as Tucker Carlson, who are both much smarter than a lot of these other crazy people who are supporting this administration's crushing of free speech, crushing of free assembly, I mean, being anti-constitutional on this. Because they're like, "Well, wait a minute. If that happens now, what happens in the next Democratic administration?"

• 17:27Speaker 01

  Like, if you're all these, you know, kind of crazy... You know, kind of.

• 17:33Speaker 00

  Just be free speech, the speech that I like.

• 17:35Speaker 01

  Yes.

• 17:35Speaker 00

  Because what happens when there's a different administration who doesn't like my free speech?

• 17:38Speaker 01

  Yes, exactly. So you have to be, and of course, I find it, I find who is really legitimately free speech or not is if they were speaking up against, you know, Carr, who frankly, I think on this, you know, I think there was, there would be a good reason for him to lose his job because of that. But as opposed to that, it's like, no, this is the kind of intimidation and fear we want. It's the kind of, of bullying that I think is anti-America, anti our 250 years of creating, you know, kind of freedom of press, freedom of assembly, to, to be able to have a free democracy. It's precisely the anti-American that everyone should be speaking up on. And as I mentioned, my amusement was like, well, actually, in fact, now people

• 18:22Speaker 01

  I rarely agree with, I agree with them. And the other people who are between quiet and supportive, I think are demonstrably un-American.

• 18:29Speaker 00

  Absolutely. And credit where credit is due. You can disagree with 90% of what someone says and actually say thank you for speaking out.

• 18:35Speaker 01

  Yeah.

• 18:36Speaker 00

  Because you're representing the Constitution. So some people are saying that after the midterms, and nothing is a guarantee, especially with the shenanigans of the Trump administration, but a lot of people think the Democrats are going to win the House. And so after the midterms, you could argue that perhaps this media, this retribution is just a blip. And after the midterms, since the Democrats will have the House, we will see less of this. Other people will say, oh, no, no, this will be a lame duck administration. They will see that they're losing their grip and they will tighten even more to ensure that all this negative press doesn't get out there. How do you see the landscape shifting after November?

• 19:13Speaker 01

  I think that it'll get worse and more volatile.

• 19:16Speaker 00

  Yeah.

• 19:16Speaker 01

  I think that part of that's because, I mean, it's like, for example, the, the public loyalty oath for the Trump administration is that the 2020 election was stolen. Now, you have to like say that in public. Now, it's like, okay, let me just repeat what I understand that claim to be, which is when Trump was president, with a Republican Senate, and, and mostly in states that had Republican Secretary of Election.

• 19:43Speaker 01

  Sleepy Joe Biden stole the election.

• 19:45Speaker 00

  Yep.

• 19:46Speaker 01

  But when Biden was president, he didn't.

• 19:49Speaker 00

  Right.

• 19:50Speaker 01

  That's literally the claim. I was like, that's the claim. And it's amazing to watch people do gyrations of, well, but there were activists who were bust. And you're like, well, why weren't there activists bust then too? I mean, like, it's just like, it's like, if it, if, like, what do human beings do? If it worked once, you're gonna do it again.

• 20:07Speaker 00

  Totally.

• 20:07Speaker 01

  Right? And it's like, it's not, you know, so it just, it's, it's literally the, I must say, you know, up is down.

• 20:15Speaker 00

  But also, they only stole it for the presidency. They decided to not steal it for Senate seats or House of Representatives.

• 20:20Speaker 01

  Which of course will deliberate a very, a very detailed conspiracy plan.

• 20:24Speaker 00

  Exactly.

• 20:25Speaker 01

  And so, now, what I think will happen is Trump will either go one of two directions if the Dems win the House. He'll either go, that was a fake election.

• 20:35Speaker 00

  Yep.

• 20:35Speaker 01

  Like, the kind of prediction that I see, it was like, oh, the Democrats are trying to steal the election with illegal immigrants, you know, blah, blah, blah. And so that, of course, now that we have this massively funded ICE, which is mostly terrorizing blue cities, right? As opposed to, like, for example, doing stuff on the border, which is, I think, what Americans want, right? Is, you know, kind of be kind of crazy. So it's either gonna be the, oh, the Dems stole the election, or the Republican Party is so weak because they're not using me, right? You know, etc. Like, but it won't be the, no, this is actually, in fact, a good referendum on how bad things have been in only a year.

• 21:13Speaker 00

  Yep. No, that all makes sense. Well, Reid, it is always a pleasure. Thank you so much.

• 21:18Speaker 01

  Thank you. Possible is produced by Palette Media. It's hosted by Ari Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Tenasi Delos, Katie Sanders, Spencer Strasmore, Imo Zoo, Trent Barboza, and Tefazwa Nirmurundwe.

• 21:34Speaker 00

  Special thanks to Surya, Yala Manchili, Sayida Sabieva, Ian Alice, Greg Beato, Parth Patel, and Ben Ralas.