60: Patrick Hsu - Speedrunning the Infinite Game

So let's say in the 1940s, the vast majority of R&D spent was from industry.

The NIH, a lot of our kind of post-war government funding apparatus became the dominant way by which we kind of funded, incentivized kind of biomedical innovation in this country.

And then I think now, today, we seem to be...

in a place where that kind of 50 year sort of default is now historically weak and the sort of market share, right, of kind of corporate agenda and ideas and leadership of innovation is now ascendant, right? And that seems to be a trend that's likely to only continue, right? Eli Lilly, for example, is the largest bout tech company in the world.

It's valued about a trillion dollars.

The GLP-1 revenue is going to be something like $90 billion this year.

And they're going to scale their internal R&D budget linearly or along with their GLP-1 revenue, which they project over time, it will be half of the entire pharma industry.

So half will be GLP-1, the other half will be everything else.

Now, this is interesting because Eli Lilly spends more money on R&D than the entire country of Germany, not just the government of Germany, but the aggregate national output of all German companies in health and biopharma.

And so by the early to mid 2030s, they'll surpass the United States.

if GLP-1 revenue continues to ramp the way that it has.

Sort of the power law we saw in software and technology happening in pharma on some basic level, or is that maybe that's an oversimplification? The question is how repeatable this will be,

right? GLP-1 is really special because it's so horizontal, right? Other medicines, while really important, generally work for a very specific disease.

But targeting obesity seems to target so many other comorbidities and copathologies.

Everyone's trying to figure out what's the GLP-1 for sleep, for vitality, for health span.

I think ultimately why everyone's so intrigued and interested and skeptical about peptides, like why are peptides such a thing? Why does everyone ask me about them at dinner? It's sort of this idea that, you know, Modern medicine has failed us.

The drug industry has failed us.

It's given all these really expensive, really targeted drugs that work only for a very specific thing with crazy side effects.

And for the first time, we found that there's something that feels magical, right? Just the product.

feels magical yeah and we can see it we can feel it it's very tangible and visceral it's not you know these weird pharma ads with happy people now able to stand up again and walk around but you you see all of your own friends your entire friend group or mine is kind of often is being shaped by this, right? And the idea that this could extend to, you know, other parts of, you know, things that plague humans or could enhance humans, I think is incredibly tantalizing.

So this sort of proof of existence of biomedical product that can be for everybody, combined with this revolution in, you know, all of the AI models for biology, I think kind of is creating today a massive wave of excitement for what geniuses in a data center can do when they break outside of the data center,

right? Is the implication there that you think there could be more GOP1s, not in the sense of more peptides for X, but more, that category goes from one to N in terms of these sort of like interventions that are broadly useful across a whole suite of categories?

One of the funny things about innovation is it's very obvious post-talk, right? And for example, you know, I was recently thinking about this, you know, like, you know, with a four minute mile, right? As soon as, you know, Roger Bannister ran faster than four minutes, so many people like 10 people broke it within some very short period of time right as soon as you know we knew that you know you um are able to make a nuclear bomb right you know other people are able to replicate it and so both in scientific discovery and in quant trading right oftentimes just knowing that there's alpha and that someone is making progress in this general area you don't need to know a lot of specific details That's sufficient, right? And this has come up recently in the discussion between the frontier AI labs about who's actually solving the millennium price problems and, you know, kind of how do they learn about specific mathematical approaches, you know, kind of.

I'm not well-versed in the merits of that kind of academic credit discussion, but I think the fact that multiple people and multiple models are able to kind of converge on the same problem at the same time is something very kind of deep in how discovery happens.

And I think that's something that's really interesting that just knowing that this is possible, creates the existence proof that gives people and markets somehow the courage and changes something about the plumbing of the universe to make it possible, right? And so I just think it is going to happen, right? Because now people think it's possible.