E51: Seth Bannon on Deep Tech, Biotech, and VC Strategies

23:01Or if you look at edit, again, you know, the previous technique for editing was called talons, super messy, super long, super expensive. And you wouldn't even be sure that you made the edits that you wanted. Thanks to CRISPR-Cas systems, which people have probably heard of, for which the Nobel Prize came out, we can now super easily and accurately edit genomes. It's so easy that, you know, high school students are editing yeast genomes with CRISPR kits that their teachers bought online. And then when it comes to design, you know, I think largely because of the advances in computational biology and things like LLMs, a lot of what used to happen in the head of a medicinal chemist can now be done extremely fast, extremely cheaply, and very accurately using computational techniques, which is all to say that the cost and complexity of building in biology has dropped by a couple orders of magnitude over the last five years.

23:47And we think it's going to continue over the next 10. If we're in 1995, what's going to get us to 2008, 2009? What's going to be the iPhone moment here that's going to 10x the big impact that it's already having? So we talked about when it comes to read, right now, we're really good at reading DNA, but there are other omics, right? So there's things like the proteome or the transcriptome or the morpholome. And it really does feel like on the horizon are going to be similar drops in the cost complexity of reading each of those, but we're not there yet. We definitely need to get there. I think once you see the sort of like Illumina level breakthrough for proteomics, we will see a massive explosion of animation.

24:31We're really excited about that. Two, right now, though DNA writing has been abstracted away, you can only do it for short oligos, so short DNA strands. And then you have to basically piece them all together in the lab if you want to do something long. There's a new way of writing DNA called enzymatic DNA synthesis. So the current way we write DNA is a technique called phosphoramidite chemistry. Long story short, it's a man-made way of doing it. Every living thing in the world makes DNA using enzymes. And so for short, enzymes are the right way of doing it. Enzymes should enable us to write DNA strands thousands and thousands of MERS long.

25:09We should be able to write a gene, a chromosome. When we're able to do that, our ability to engineer biology will radically jump forward. There are new editing techniques that are coming that are even better than CRISPR. So CRISPR has a few downsides. It suffers from what are called off-target effects, where every now and then you will make an edit somewhere you didn't want to make the edit. And so once we can even more accurately edit, it's going to be really, really powerful. And then the thing that is happening literally right now, I mean, as we speak, is the application of large language models to biology.

25:40And it has been pretty wild, just as surprising as the advancements we've gotten in, you know, you know, image creation or video creation or, you know, things like chat GPT, it is wildly surprising how effective they are at helping us engineer biology. Like you can literally just sort of translate DNA sequences, treat them as language or small molecule structures and treat them as language and throw them at these large language models. And they will help you get to a level of understanding faster and cheaper than ever possible before. So, so that last one, I think is like, it's actually happening right now.

26:12Maybe, I don't know if I want to say it's already happened, but like, it's like this year and the other ones I think are coming over the next five years. Say more about what that last change will mean for investors in the space. Is it that a set of startups will now be, you know, investable that previously weren't? Is it that, you know, it will kind of diffuse, the benefits will diffuse throughout all the startups in the space or how do you, or more incumbents maybe, how do you think about like unpack more what this change will mean when it happens? Yeah, it's pretty interesting. So I think AI in general is very interesting in that it seems like a large amount of the value, say outside of biology, will accrue to the people that already have large distribution, right?

26:53Like the speed of incorporating large language model technology into existing products has been really fast. And so while some of the model companies, is I think my benefit, it's unclear to me if the vast majority of the rest of the value is going to go to new startups that are based off of this or the existing companies that already have distribution that just quickly, you know, insert that technology into what they're doing. For biology, I'm much more confident that the majority of the value will come to startups because the pharma companies historically have never moved as fast at incorporating new technology as like a Facebook or even a Salesforce, right?

27:37And over the last 10 years, one of the trends we've seen in that space is that they've sort of functionally almost outsourced their R &D to acquisitions. So they basically wait for a startup to build something cool and they go and acquire it. And that's sort of how they innovate. Which means I think that the benefits we're able to get via these large language model advances will go to startups. And a lot of them will be new startups, but a lot of them will be existing startups. Because one of the things that we now know when it comes to ML in general is that what you can do, the value you can get from these techniques is in large part dependent on the quality and the proprietariness of the data sets that you have.

28:14And a lot of these bioengineering companies, that is their core innovation. Their core innovation is some way of seeing biology in a way no one else can do. So, you know, we have a company called Digital Biology, for instance, and they've developed a new technique. It's out of the Wies Institute at Harvard. essentially they can, using a photolithography, paint DNA barcodes onto cells in a way that allows them to, in very high throughput, sequence what's going on with full spatial context. So what does that mean? Spatial context, you might have three cells, cell A, cell B, cell C. Cell C might be in a disease state, and you want to know why.

28:51And it might only be in a disease state because it's next to cell B, which is sending some weird signaling molecule to it. and cell B might only be in that state because cell A, which is next to it, is sending another signaling molecule. And if you don't have that spatial context, cell A is next to cell B is next to cell C, you might have no idea why cell C is in that disease state. And the solution, the therapeutic, might require you to treat cell A, not cell C. And currently, the paradigm is you can either have that spatial context, but in very low throughput, one at a time, which is very slow and expensive, or you can do super high throughput, but you lose the spatial context.

29:26You don't realize that they're next to each other. Digital biology for the first time can give you super high throughput and the full spatial context, right? And so they're going to be able to generate a data set that literally no one else has access to, which means they're going to be able to leverage these advances in ML in a way that other people aren't. And I think that is a dynamic that we're going to see across a lot of these bioengineering companies. That's a fascinating overview. And it makes me ask, as kind of a side note, you don't have a PhD yourself, right? No, I didn't even graduate college.

29:54Okay, exactly. How have you educated yourself and how would you advise others who are fascinated by this revolution but don't have the formal training to get up to speed on it? Where's the best bang for your buck? So biology is very interesting in that unlike physics, you can struggle over a paper and understand what's going on as long as you have Google by your side. So physics is hard in that there's typically some sort of esoteric mathematics. and if you don't know that math, like you're not gonna understand that paper. There will literally be alien symbols on the page and you're like, I don't know what that means.

30:31And it's not that easy to like Google it because like, what do I even Google? I don't know what that symbol is called. I mean, maybe now with GPT-4 and like the multimodal stuff, it'll be easier. But, you know, physics is very, very hard if you don't have deep technical training to even grok what's going on. It's like super counterintuitive and really, really difficult. Biology, the most, in a biology paper, the most advanced mathematics you'll typically see is some statistics, right? And it's not that hard to understand statistics. And there will be words you don't understand, but you can just Google those words.

30:58And by doing that, you can grok generally what's going on. You can grok what the advancement is, why it's important, what the challenges are ahead. And so for me, I've just read now probably thousands of academic papers. And then you make notes in the margins of the things you don't understand. And then when you meet the entrepreneurs, they educate you on what's going on. I mean, it's like one of the craziest things about being in DC is that every day we're getting educated by the absolutely most brilliant, passionate pioneers of a space who are literally building the future of their fields. Right.

31:30And because they're entrepreneurs, they typically are slightly better communicators than your average academic. And so my knowledge has literally just come from, you know, all these incredible conversations with all these founders that we've had over the years. Yeah. And you talk to founders who teach you, and then you talk to LPs and you teach them. Yeah, exactly. Which is a segue to, let's say, we're at Perspective LP and we're asking you, hey, could you create kind of like almost, we're commercially motivated or minded. Can you create like a market map of how you think about the 50 years kind of investable scope?

32:04Like what are the things that you're trying to invest in, the opportunities, the subsectors that you think are ready for commercialization now versus ones that may not be ready or may not be. We would say, no, we cannot do that, potential LP. So there's a lot of VC funds that are what are called thesis driven. So they basically say, here's how the world's going to pan out over the next 20 years. And therefore, I want 15 % of my portfolio to be in this area of healthcare and 10 % to be in space technology and 30 % to be in this other thing. We are intentionally antithesis. One of our core values as a firm is beginner's mind, right?

32:40And why is that? I think it's particularly important if you're a deep tech firm, because the nature of every single fundamental advancement in deep technology is that at some point, there were five to 10 people in the world that even knew it was possible, and therefore saw how the world was going to change because of it. And those were the five to 10 people in the lab that just figured it out. And if you're not one of those five to 10 people, you're not going to be aware of it. And so the only thing I can guarantee about VCs, you know, sitting somewhere in Silicon Valley, like whiteboarding where the world is going, is that they're going to be terribly, terribly wrong.

33:14And so we, instead of focusing on being thesis driven, focus on being network driven. That sound familiar, Eric? Network driven. And so what does that mean for us? That means making sure that we are building, you know, resources like the spin-off playbook or programs of which we have several that are helpful to the researchers and the engineers and the scientists that we think are gonna be among that group of five to 10 so that when they have that breakthrough, they come and tell us about it and then to make sure we have the capabilities of doing our own analysis of whether or not we agree with them.

33:49One of my favorite books, it's called Profiles of the Future. I actually have it here on the table here. And it's by this guy, Arthur C. Clarke. And it makes, I think, a really compelling case that you cannot predict the future. No matter who you are, VC, scientists, engineer, nobody. And there's just a few examples. So there was this scientist who basically published a piece on rocketry, and he was saying that spaceflight, human spaceflight would be possible. And in 1924, Nature, which is one of the premier scientific publications, it's probably the premier, Nature published a review of it. And, you know, the nature of you say, quote, in these days of unprecedented achievement, one cannot venture to suggest that even Herr Obath's ambitious scheme may not be realized before the human race is extinct.

34:35And then literally 40 years later, everything he was saying happened. Right. So like they were they were claiming that literally human race might go extinct before it happened. And then, boom, 40 years later, it happened. there was a guy named Lord Rutherford who helped maybe more than anyone elucidate the internal structure of an atom. And people in Lord Rutherford's day, and this is like the, you know, 1910, 1915, were starting to talk about harnessing atomic energy, nuclear energy. And he literally like mocked them, like, like mercilessly, like mocked, like made fun of basically saying that that would just simply never happen.

35:14Literally five years after he died, we harnessed the first controlled chain reaction. And this was someone who was literally at the forefront of his field. He literally was responsible for the most recent advancement, and he got it completely wrong. And so I know if these people get it wrong, as VCs, we're going to get it wrong. And therefore, we really focus on just talking to the best and brightest and letting them tell us where the world is going. And then we need to be able to run our own analysis of that. that's a strong strong support of the of the beginner's mind investment philosophy i i definitely agree with you there hey we'll continue our interview in a moment after a word from our sponsors i'm gonna shift gears a little bit um one of the things that we were discussing previously is this idea of that you have of the sort of tech bio verse biotech talk about that that framing and and why it kind of required that shift in the first place yeah so so biotech has been an area of VC investment and innovation for a long time.

36:10It's largely, I would say, dominated by the Boston area. And recently, we've seen innovation in tech bios. So first, what is the difference? And why are we coming up with two different names for something that is essentially aiming towards the same goal, which is curing disease via therapeutics? So a biotech company typically finds a, say, small molecule or a collection of small molecules that it thinks treats Alzheimer's. And the entire company is about validating that hypothesis that these small molecules treat this disease. And that's it, right? And it's interesting in that there's like a playbook for that.

36:49There's not a lot of creative problem solving required. You literally, you do your in vitro, then you do your mouse studies, then you do your phase one, phase two, phase three. Even before the in vitro, there's like a playbook on how you go about it. And it either works or it doesn't work. And if it works, it's going to be worth a lot. If it doesn't work, it's, you know, very sad. Shut it down. Very recently, there have been entrepreneurs that have the exact same goal, which is we want to treat disease with small molecules or gene therapies or cell therapies or whatever. But they don't start with, we have these things that we think will work.

37:19They start with an engine. They start with some computational engine, some high throughput assay that they think will lead to many, many, many such assets over time. And it's interesting in that though the goal is the same, the way you run these companies is radically different, right? The way you should fund these companies is radically different because with a tech bio company, there is no formula on how you do it. And so creative problem solving is required at every turn, right? And so with biotech, the idea of hire CEO, have them turn the crank sort of works. With tech bio, it doesn't work at all because when you have to creatively problem solve, no one is better at that than the founder, right?

37:58The inventor of the technology. It's also interesting in that it opens up completely new models for monetizing and running the business. For instance, in the old biotech model, you do not want to give up your assets until they've manifested their full value because that's the company, that's it. That is the company, right? So like the idea of taking your asset and selling it and only getting 2 % of the value would be absurd because like, well, now your entire company is only gonna manifest 2 % of the potential value. But for a tech bio company, right? If you really believe you can generate many, many assets over time, while giving up 2 % of the first one, who cares, right?

38:37That might be totally fine, because you might have 100 that are going to follow. And so it actually unlocks new ways of running these companies that are very, very interesting. So oftentimes, these companies have a technology platform that a Pfizer or Novartis or Genentech would really like access to, because they go, Oh, man, wow, that's cool. That could let us see biology in a way that might help us develop drugs. And so that company might go, Yeah, sure, here, you can have access to it, Give us a couple hundred thousand dollars and you can have access to it. Right. But only for a couple of months.

39:06And then, and then, and then, yeah, it's true that Genentech might discover a drug that makes them a billion dollars because that two month, you know,$200 ,000 period, but you know, it's fine because that's, that's only a little piece of what we can do. And then after that two months ago, Hey, did you like that? Cool. Well, guess what? You know, you now, if you want access again, you got to pay us$500 ,000. And if you find a drug, you got to pay us even more. And Genentech says, Oh, fine. Sure. We'll do that. And then if that works, you go, well, guess what? Now, if you want access to it, you got to pay us a million dollars.

39:35If you find a drug, you have to pay us money. And then if that drug makes it to market, you got to give us a percent of the revenue. Right. And this is really cool because this company can be fairly capital efficient because they're just building technology. Right. Whereas if you're building a drug, you have to spend tens and sometimes hundreds of millions of dollars pushing it all the way through clinical trials. And then over time, this company can vertically integrate. Right. So in the early days, it's just letting other people use the platform. But then maybe eight years in, it's saying, you know what?

40:04Other people, you can't use our platform anymore because we're going to use our platform. We're going to take the full upside of all these therapeutics that we make. And it's really nice because Jeff Bezos has this idea called type one and type two doors. Type one door is a door you can walk through. If you don't like what's on the other side, you just walk back out. Type two door is a door you walk through you don't like. Well, it might lock behind you. You might not be able to get back out. And for type one doors, he says, forget about it. just don't think, just go. Walk through the door. If you don't like it, come back out.

40:29Bias for action. And for type two doors, he says, whoa, whoa, whoa, you got to slow down. You really got to strategically analyze because you might not like it over there. It might be really bad. And developing your own drugs is a type two door because in order to know if it's going to work at all, you have to spend at least millions of dollars, sometimes tens of millions of dollars, sometimes hundreds of millions of dollars. And therefore you want to kick that can down the road, that type two decision down the road as long as possible. And this is why we really like this new model where in the early days, companies partner.

40:57And yes, they give up the vast majority of upside of those early things that the platform finds. And then over time, they vertically integrate. Once they have more capital, they have a bigger team, and they've learned all this amazing stuff from all these pharma partners they've worked with. Yeah, that's a very helpful way of framing it. Talk more about how these companies should think about what platform strategy is most optimal for them? So, I mean, at the end of the day, one of the most important ingredients to this, which people don't really talk about that much, it's just like founder enthusiasm.

41:29I know it sounds silly, but the most important scarce resource in any startup is the energy of a founder, right? Literally every other scarce resource from capital to talent to customers can be solved so long as the founder energy is high. And so right off the bat, the founder should really think about like, in what directions would I be most excited to point my platform and like rank those things very, very, very high. And then you should do customer development. And this looks a lot like customer development in the enterprise SaaS world, right? You literally, you pull up a list of like the 55 initial partners that you might have.

42:04You create a list of your assumptions. This is what we think those people are going to want. And here's why. And then you call them up and you validate. And you just have calls and calls and calls and calls and calls. We have a company called Manifold Bio, which really turned this into a science. They created this incredible spreadsheet where for every single hypothesis on every single conversation, they marked it from a negative two, meaning very negative reaction to positive two, meaning they were very, very interested. Zero is a neutral, which might be even the worst. And then they would leave it blank if they didn't come up in that conversation.

42:34And the first time you go about that, you literally approach it from a point of curiosity. You don't sell your platform at all. You're literally just learning about those customers and their problems and what they care about and how they see the world. And then you take that, you synthesize a little bit, and then maybe you start to formulate some hypothesis around how your platform could add value to them. And then you go back for round two, and that is 70 % learning, but then 30 % selling, 30 % going, hey, I really think our platform might meet this need of yours. Does that sound right to you?

42:59And then you go back and you synthesize again. And then the third time, maybe only then are you 80 % selling and then only 20 % listening. And what's really interesting is that in those initial conversations, if you go to that sort of negative two to positive two sort of framework, the things that are most interesting are the ones where you have a lot of twos and a lot of negative twos, right? So the things where some companies are like, oh my God, that's incredible. I would love to do that. And some are like, that makes absolutely no sense. It's a terrible idea. Those are probably the things you should be focusing on most because it means that it adds a lot of value and it's probably not a consensus way of going about it.

43:33And so we always encourage founders when they're thinking about this to prioritize one, what do you want to do? Like what would excite you? And then two, just do the traditional customer discovery and validate those hypotheses. Yeah, that's well said. I'm curious how you think about alternative funding models. You know, Zach Weinberg just launched to Curie and then also sort of D-Sci movement more broadly. I'm just curious how you think about, like how do we better capitalize these companies? Yeah, yeah. And I love, Zach just became an LP. I love what he's doing with Curie. We backed a team called Molecule, which is sort of at the forefront of the design movement.

44:09So I think we need way more experiments. I think we need way more experiments because these companies are, they have to exist, right? These companies are solving incredibly important problems and they've traditionally been neglected by the VC world. The non-VC world, you've typically either had to go to that sort of government sources or some philanthropists. And there's, I think, a huge opportunity need to run a lot of experiments. And so I'm super excited about FROs, you know, focused research organizations, which are sort of quasi philanthropic, but where the goal is actually to potentially commercialize to create an endowment.

44:42That's a really interesting idea. Very exciting that the UK announced that they're going to put$50 million of UK money behind UK based FROs. I think that's like incredibly exciting to see something that was a fringe idea. And no one was really talking about three years ago, get support of a major European nation. I think we need to explore models where patient advocacy groups can help translate research that maybe they care about, but other people don't, and then get a percent of the upside if it works. This is something that I think Molecule and what they're doing with the IPNFT is enabling. And then I love the idea of Curie, which is saying, hey, there's a lot of stuff that you have to do as a biotech company that everyone else has to do.

45:24It's really, really expensive to do all that stuff. And so we're going to create a massive team that does that for you. And so that you can just focus sort of on the core science and the business strategy and hiring and all that jazz. So, yeah, I think I just we need more and we need even more experiments. We're experimenting with a model for funding translational research ourselves. So we call it manifest grants. We tested it out in the female reproductive health space, which is a very underfunded area where there's not a lot of companies commercializing because there's not a lot of funding for, you know, for science that can be commercialized.

45:56And so we raised philanthropic capital as a VC fund, create its own separate entity. We got a committee of reviewers together. So we ended up having 40 reviewers. These are sort of experts in the space, a one and a half page application. And then within three weeks, we would get your response of whether or not you get funding. We backed over 14 projects and it worked incredibly well. And so we're now actually going to be launching this for other areas of science. We were inspired prominently by FastGrants, which was launched by Patrick Collison and Tyler Cohen during COVID to do this exact thing.

46:28And we talked to Patrick, talked to Tyler, and we're literally just took their exact formula and are just now going to roll it out for a bunch of different areas of science so that hopefully we can, and these are no strings attached grants. So we don't take any IP. There's no deal that we have to be able to back the company. It's just, here's some money, go do great science. And then our goal, hopefully, is that we create enough goodwill such that if these scientists do want to start a company someday, they come and talk to us. it's fascinating you know i've read patrick you know a few years ago he wrote with michael nielsen this this idea that science is slowing down on a kind of per capita basis and and one of the the reasons he believes is because our the you know the funding institution is just far too centralized and we need more diversification of of sort of funding sources and and you know jose has forgetting his last name he wrote was written about how we should fund people instead of a projects sometimes.

47:20I say this all as a segue to ask you, if you could change anything about how we fund science in the US on a structural level, you could wave a wand. Where do you think would be the best bang for the buck? So I think I would love to incentivize failure. So right now, the currency of academia, sadly, is publication, right? It's what high priority journals do you get published in and then secondarily, like how many people cite your work. And therefore, there's actually an impetus to work on something that you're fairly sure is going to work and that will lead to at least a top 10 publication journal and that hopefully will be relevant to other people.

47:59And it disincentivizes the sort of like really big swings. And so I actually think if I could change something, it would be, I mean, I actually think there should be a publication that is of the stature of nature, cell or science that, but you can only get in if you failed, if your research didn't work out. But you had to be just as rigorous in explaining what happened as if it did work out, right? Because actually, it's super, super, super relevant. Like a lot of scientists, when another project fails, and they go and talk to that scientist, they learn a ton of stuff that then informs their own research.

48:30And so I think if you actually had a very high impact journal that only included failures, a lot of them would get cited, right? And it might create this incentive to aim for really, really ambitious things. And if universities and PIs were themselves trying out more ambitious things, I think the funding would follow. So that's actually, it's a little bit tendential to your question, but that's the one thing that I would change. That's very, very helpful. Mark Andreessen tweeted the other day, he asked, what percentage of published academic research papers are worth anything? Roughly half of the fields are useless.

49:0490 % of the research from real fields is marginal at best. And half or more of the remainder doesn't replicate. Do you think there's any truth to that? Or is he totally off his rocker? Obviously, he's a smart guy. How do you think about sort of the state of these fields? Oh, I completely agree. I mean, you have to realize that a lot of what is considered science is science in air quotes, right? It's like soft science, you know, the journal of metropolitan sociology, right? Like, you know, I don't know any physicist or biologist that would look at that and go, yeah, that's science. It's like, not really.

49:41But those are considered scientific publications. And those far outnumber nature cell and science and the journals of biotechnology and things like that. So for sure, yes, the vast majority is quite silly and completely impossible to reproduce because it's not real science. It's not hard science. If you took out those and just looked at the sort of the hard sciences, I do still think that his point holds, there's some merit in the point. Again, the vast majority of people are just looking to publish and therefore going for things that they're fairly sure are going to publish. In biology, we have a massive reproducibility problem, which I actually don't believe comes from nefarious reasons.

50:20It's just that the field has not been professionalized. If you go into an academic lab, you will still see people who literally have studied for nine years of their lives to be there pipetting, like literally with their hands moving liquid from one place to another, right? And of course, that's a problem because no two people pipet the same way. We have a company called Opentrons, which makes a$4 ,000 lab pipetting robot. They've grown like Wildfire,$2 billion company now. Why is it cool? It's cool because if you run your experiment on Opentrons at Harvard, and someone at the University of Tokyo wants to reproduce it, they can download your protocol.

50:52And if they have an Opentrons, it will run the exact same way, right? So it helps with the sort of reproducibility problem. So yeah, I agree. If you take the most interesting journals, maybe like The Nature and Cell and Sciences, I think a pretty good percentage of them are very, very interesting and are potentially incredibly impactful. Many of them are not because of the problem we talked about earlier, which is that a lot of this incredible research ends up in that journal, and then nothing happens, right? Because there's still a lot of work to manifest it in the world, and the people that are best positioned to do that, the PhDs that invented the technology are not incentivized to do that and don't think of themselves as potential entrepreneurs.

51:31And so I think if we could solve that problem, if we could inspire more PhDs to start more companies, then what Mark said would no longer be the case. Yeah, I love, this is a perfect place to wrap the interview. You've, in this interview, you've brought some realism about the challenges that we face today, some optimism about, you know, what's on the other side if we're able to solve some of these challenges, but then while also recognizing that we are, solving some of them and then some practical advice on how we can get there. So this has been a very complete interview. Seth, thank you so much for coming on.

52:01For people who want to learn more about 50 years, what other plugs do you want to leave us with? Yeah, 50years.com, 50years on Twitter, and I'm Seth Bannon on Twitter. We're pretty noisy. We share our opinions quite a lot. So give us a follow. Awesome. I've learned a lot from your content and excited for more to come. Thanks so much for coming to the podcast. Thanks, Eric. This was fun. Turpentine VC is a podcast from Turpentine, the network behind Moment of Zen and Econ 102. If you liked the episode, please leave a review in the Apple Store or rate us on Spotify.