Zach Weinberg, Co-founder of Flatiron Health & Curie.Bio, on Fixing the Biotech Funding Model

But there's many dimensions of a business, and sometimes coming from the outside can be advantageous. So you're bringing a different skill set, and you can learn within the domain what you need to as long as you're motivated enough or you're pulled toward it.

Yeah, you have to be willing to do the 100 phone calls, right? You have to want to. Okay, maybe your outside perspective is great, maybe it's not, but you need to understand why the industry works that way. Whether you agree with it or not is irrelevant. First you have to understand it, which usually means having a lot of conversations, a lot of reading. I remember starting Flatiron, I had notebooks — this is before you took notes on a computer because I'm old — like you had notebooks of stuff, layers and layers of notes. In biotech, if you look at my calendar in 2020, 2021, and you went day by day, it would be like out of 40 phone calls a week, something like that, 30 phone calls a week, just asking questions: how does it work this way, why does it work that way. I think you have to be willing to do that.

People always say experts in particular. I think we underestimate how much knowledge is not written down. It's just not in a blog post, sorry, it's not in a McKinsey industry report. I mean, that stuff is good and it gives you a lay of the land, the terminology and who the players are, it gives you the surface layer. But the really important nuance stuff is just in people's heads, and you have to find the people and then you have to figure out how to get it out of them and ask the question the right way. So my process has always been talking to people, doing expert calls, getting intros to the next set of experts, digging, digging, digging. To the point where you've done it well where you talk to one expert and then they recommend two others, and you're like, I've already talked to them. When you get to that point where you're starting to see patterns in the people that are being recommended and you've talked to 80% of them, now you're in the belly of the beast. Now you know all the right people, and then it's just a matter of are you good enough at getting the information out and synthesizing it. Some people aren't. But it's been expert calls. One of my co-founders at Curie — there's four of us — he said, if I took a scientific problem that I know nothing about, literally I don't even know what the words mean, and if you gave me 20 hours and 25 calls, I will be a better expert at that thing than you reading a paper any day. And he's right, because there's only so much in the reading, there's only so much in the paper. If you know the right people to call, you'll figure it out better than the next person because most of that information is here. So that's what I've gotten really good at. I don't think I knew it back in the day that this was the right approach, I kind of discovered it, but that's how we did Flatiron. 18 months, give or take, to really find the idea. And then Curie was like 12 to 18 or so, and I had a head start because the first time is the hardest. Nobody wants to meet you, they're like, who the hell are you? So you don't get the meeting, your hit rate is low, and the bar for people sharing information with you when you're a first-time founder is really high because you're viewed skeptically. And now I send someone an email and they answer. I have an advantage. So this one we could shortcut it. That's why second-time founders and third-time founders have better chances of success, just because your hit rate with people is higher. But that's been the process, just out in the field. Now it really needs Zoom to be clear, but effective.

No AI. We had a little bit, we had a small team, about 10 people or so out of 500, to give you a sense of scale. Everything in healthcare is in context. There are no rules in healthcare. The rule is understand the problem that you're working on, and each one is a little different. The incentives are different, the diseases are different. People would come to me and say, oh, I'm doing Flatiron for XYZ disease, and that has absolutely nothing to do with what we did over here because the disease characteristics are different. You can tell those people just don't understand the details. For us, the thesis for Flatiron was that when you see your oncologist, post-diagnosis, there's a lot of information about you, your disease, and the treatments you are receiving, the diagnostics, imaging scans, as well as your response or lack thereof to the treatment, that sit in the electronic health record. Specifically in the record because oncology is a very in-person treatment. Most of the treatments are infusions, you're actually in the office, the physician, your oncologist, kind of becomes your PCP. So we felt there was this really rich data set inside the EHR, particularly in oncology, which is not true in other disease areas. In diabetes, for example, it's not like your chart has all the stuff. In diabetes, what matters is outside the office, what's happening to you throughout the day. So the data set that matters there is very different. But in oncology, it was in the chart. There were some really interesting nuances about how the treatments, the drugs in oncology, the trial is run on a certain population of a certain type of a certain size, and then when that drug makes its way out into the real world, the population looks different. It looks different for a variety of reasons: people are older, they're sicker, they're more minorities because minorities don't really participate in clinical trials, which is a whole other topic. So the pharmaceutical companies would be interested in understanding what was happening in the real world because the trial data was only one piece of the picture. We built an entire business, a money-losing business, to aggregate data from cancer centers through the EHR. We built this data curation infrastructure, which is a fancy way of saying nurses looking at charts and abstracting data from the charts. Then we built a pharma business, which is where the money was made on the other side. So we actually had to build almost three companies in one. Healthcare data is really messy, a lot of natural language, a lot of imprecise language. 'Suspicion of' — what does that mean? We felt there was a quality bar for medical data that was higher than what ML, specifically language models, could do. So we had a human curation team, and we used machine learning in smart ways primarily to do search space reduction. If you're looking for patients with a certain disease characteristic and you have a million charts to look through, how do you find the 20,000 that you probably should look through? We would use machine learning in really clever ways to reduce the workload, but it was not a core tool in the actual end product. That was just unique to our business.

We would try it and see how accurate it was. I don't know. We'd have to really understand what the language models can and can't do, and the only way to do that would be to just try it. So yeah, we probably would have a slightly bigger focus because theoretically the quality would be higher than it was when we were starting. And then we would see what was actually possible.

I get to do the Curie pitch now. Okay, all right, cool. Try not to be the salesman. Okay, so here's the background and the thesis for Curie, and I will answer this question of what's exciting. If you look at software over the last 20-some-odd years, starting around 2003, 2004 with cloud computing, the cost to start a software company came down. The amount of money you have to spend right in the beginning, Day Zero, went from $7, $10, $12 million — that's in 2000 dollars, so the equivalent of $25, $30 million today — to like $2 million, because you could rent the infrastructure, you didn't have to build the data center, buy all the equipment, the servers, the disk space. The tools got better and better, and all of a sudden, starting a software company in your garage became a reality for almost any idea. Most of these companies were started in effectively a garage, two people in a room. So the cost to start software companies comes down, the cost to try — how much money do I need to figure out if this thing could work — comes down. What do we see happen? Seed funds pop up because all of a sudden you can write a $2 million check instead of a $10 million check. So for my $10 million, I get five shots on goal instead of one, so I'm going to do more of it. Founder ownership goes up because you can raise less money in the beginning, so you give up less of your own company. All of a sudden it's financially really interesting to be a founder. You remember all the acqui-hires that were happening back in the mid-2000s? I knew people that started a company and nine months later it was $10 million because the economics were really good. So you had all these factors coming together at the same time, and software just explodes over 25 years because the risk-reward of how hard this is going to be and how much am I going to own kind of lines up. It's this beautiful perpetual cycle, and I think it was enabled first and foremost just by cloud. The cost structure came down. Biotech is really interesting because if you look historically — and when I say biotech, I'm using this term very loosely, I'm talking about therapeutics, companies that make drugs — the cost to start a therapeutic company has been materially high. We're talking $20, $30, $40 million. You can't start a therapeutics company in your garage. It's the first slide in our deck: it's not possible yet. The history of biotech is really expensive Day Zero costs because you have to build a lab. It's an expensive proposition: physical space, hire all these people, put the equipment in there. By the way, you can't just laugh, it takes a year. You're doing these really expensive places in Boston and San Francisco, which makes no sense. So lots of money at the earliest phase, founders get diluted to hell, they don't really own their own company, and few ideas get funded because the amount of money you have to put in is really high, really risky, so the risk appetite from investors comes down. It's just math, it's not good or bad, it just kind of is. What is changing in biotech is that the cost to do early pre-clinical, pre-human work has been coming down for a variety of reasons. It's not one thing. The cost of sequencing is much cheaper, costs to synthesize chemicals have gotten cheaper. The other thing is the vendors that will do this for you have gotten better. There are vendors out there, most of them are ex-US, not in the US, but think like WuXi, or Pharmaron, or ChemPartners, or Evotec in Europe, groups in India, heavily Asia and Europe, where you can basically outsource the initial set of experiment work because they have the infrastructure at scale and they're better at it. They have teams of people, they've hired them already, they've done this 20 times, they have specialized expertise. This didn't really exist eight or nine years ago at any sort of quality. So what's happened recently is they just got better. It wasn't overnight, it was just little stepwise improvements, and all of a sudden you realize this group actually is better at doing this work than I could. When I was getting into this, I learned about this idea that costs were coming down. People were pretty consistent: yeah, they're coming down, not for everything, but for common modalities — small molecules like pills in a bottle, biologics like engineered proteins, and a few other areas. I was like, where the hell are all the seed companies? Because you would assume if the costs come down, you would see more seed-stage startups, more founders. And they didn't really see it. Just to give you some stats, something like 200 to 300 new therapeutic companies, biotech broadly, are funded every year. Software was like 14,000. Order of magnitude different. So what's going on? Our thesis came down to that it's not just the money, it's the money and the access. The access to the right people, to those vendors, to experts, to all the things you need as a founder to get started. A lot of this stuff is really hard to do when your email address is [email protected] and not at harvard.edu or at Flagship or at Third Rock or whatever fancy place you are. The minute you're on your own, nobody pays attention to you. Unfortunately, this is the real pitch, by the way. That's how I raise, I raised this money on this story, so you're hearing the full thing. This is the first group I've actually pitched to externally. When you've gone out to start a biotech company, you end up with these series of places you can make mistakes. You pick the wrong target, you select the wrong vendor, you design the wrong program. So many little places you can go wrong. You don't find the information you need because it wasn't in the paper, it's actually in that dude's head over there, and you just couldn't get on the phone with him. There's a lot of that. Unfortunately, mistakes in biotech are very expensive. It's not like software where if you make a mistake in the code or the product, you can just go and fix it that night. In biotech, we're talking about six to 12 months and a million dollars, and you get all the way back to the beginning. So I think of it as this giant catch-22: at the moment in time when you are most likely to make the mistake because no one's helping you, it's also the point in time where the mistakes hurt you the most. That's the synthesis of the whole thing. Mistakes are really expensive and it's hard to not make them. So Curie is a new type of entity we designed. We took a venture fund — we have a $270 million seed-stage venture fund — and we basically stapled it to a services entity that is deep drug discovery expertise. These are the best people in the world, and access to the experts around them who don't just pick the company but materially help you as a founder, in a model where you still own the majority of your own thing. You can take advantage of the cost structure because you can raise less money, and we can guide you through this process, but you get some of the best drug hunters and drug makers in the world co-piloting alongside you. You can think of it as Y Combinator on steroids but focused in biotech. If you're familiar with the industry, I like to think of it as Flagship for everyone. How do you take the depth of expertise inside one of these incubators and bring it out in a model where the founder doesn't get diluted from an ownership perspective? My theory is that if you make being a biotech founder a better career decision, more people will do it. If you have more people do it, you have more ideas, more shots on goal, more medicines. The industry will grow, like what happened to cloud in 2003. The reason it's been so challenging is that you have to raise a lot of money. We raised over half a billion dollars privately to do this, which we'll announce in about two weeks. So think of it as the best place to start a therapeutics company, even though it's not actually a place. It's probably the better way of saying it: the best group to start a therapeutics company with. My co-founders — two of them are basically ex-Third Rock Ventures, which is one of the best biotech incubators in the world. They're expert partners. They have been founders of $30 billion of biotech companies. I'm just the idiot that raises the money, basically, which is great. At least I know my role, and they do all the science. Hence that little pithy tagline we have: 'Free the founders.' That's what we mean. How do you get more — in particular scientists, it's really focused on physical scientists, not necessarily computer science, but we do both — to start their own companies? Take a risk, and now it's an easier decision for you to make. I'm sure if there are any biotech founders in the room, many of the things I mentioned, you're like, oh yeah, that was happening to me. Nobody paid attention, it was hard to find the vendor, I couldn't even get the vendor to return my call because my budget was too small. That's what we're solving. In two weeks, three weeks, we'll announce it. We've done four companies so far quietly, we funded last year, and we're hoping to do another 15 or so maybe this year. It's about 25 people, will be 60 at the peak, and it's almost all scientists, effectively all drug discovery PhDs, chemists, biologists.

Almost everyone. The way you make money at the end of the day is you make a drug. If you sell software to other biotechs, thank God you exist, but that's not what we're funding. Our expertise is not in building your technology, it's in applying your technology to a drug program. I'll give you a real example. This is one of our portfolio companies. It's actually two CS guys, two guys with CS backgrounds, so no drug discovery experience, who have a really unique mathematical model to find what they call cryptic pockets in a protein. Basically, places you could drug a protein that aren't easy to think through, aren't easy to find because when you crystallize the protein, it crystallizes into a particular conformation, a particular shape, but that's not actually how it looks in the real world because it's doing this. So there are other spots. Cool. The attack is really awesome and it actually works, driven by a lot of the advancements in recent ML stuff that I don't understand. But okay, cool. Which proteins? Of the billion things you could possibly do, where do you apply it? Why? In what order? How do you actually test if the pocket you found is real? There's a series of questions that immediately come out, and that's what we do for them. We work on the target list, the vendor list, the assay development, and guide them end to end. We do it with our team and with a series of experts. In a funny way, a lot of what I learned about getting experts at the table is part of what we actually do at Curie. We get you the expert because we don't have all the expertise in-house, but we find you the person who is the expert in whatever it is you need. That's a real company, one of the ones we'll announce in a few weeks.

Yeah, we're pre-clinical. We're pre-IND, meaning you're not yet putting this or close to putting it in a human. There's a series of regulatory steps that happen kind of after us. For the most part, there's no regulatory expertise needed because it's just not what we're working on. I'm sure there will be some exceptions, and we'll go find an expert where we need to. But it's really more about biological cell assays and mice and monkeys, not yet at humans. The hope is eventually for each of these companies, they get to a human, get to the clinical development program, and then you see if it works. Because everything else is just the best guess. So it's all science. There's no software, which is weird for me. There's no regulatory people. It's real, it's really all science.

Well, I'm for it. I don't know. I mean, it seems like a good thing, of course. No, I know. It's hard, right? Because what do you measure? That's the first question you have to ask: how do you know it's working? We don't have great or any really longevity biomarkers. We have things that maybe correlate with lifespan, but honestly, if I could ask you three questions — your BMI, your family history, and whether you smoke — I bet I'd be more accurate than almost any model with just those three questions. So the question for longevity has always been: what are you measuring that shows if it's working or not? If the thing you have to measure is death, and it's going to be a long time before you know whether anything works, and then you have all these confounders that come in. The longer the study runs, the less you know whether it was the thing you were doing that's driving it versus just random chance. So I think it makes it very difficult for longevity-type companies to exist because there's no clear path. We'll see. I've seen a few really interesting companies that are trying to do earlier detection of neurodegenerative diseases, like Alzheimer's and dementia. Theoretically, if you get a good accurate diagnostic of early dementia, you could measure prevention earlier in the cycle. But the reality is right now we're not particularly good at it because the MRI, the brain MRI, doesn't really show up until the disease is really bad. You could look at a reasonably healthy brain and not know essentially. It's more complexity than that, but that's the basics. That's always the challenge: what are we measuring that shows there's actually something that's truly going to work? A lot of it is pseudoscience in mice. A lot of it is like, oh, we did this thing in mice and they lived longer. Yeah, a lot of things work in mice. The stupid joke at Flatiron is the best time to get cancer is if you're a mouse, because we'll cure you. We can cure anything in a mouse. But then you stick it in a person, and it's a little more complicated of a biological system. So it's going to be tough. That's why you see a lot of these longevity companies pivot to a more actionable disease: they work on diabetes, maybe they work in neuro.

I have a strong opinion. I think this self-measurement craze — fitness tracker, sleep tracker, whatever — I think it's all basically telling healthy people who already know exactly what you're supposed to do to just do it. I'm fine if it's a behavioral nudge, and that's really what's going on here. What do these things tell you? Let me take a guess: don't drink because it messes with your sleep, don't smoke, try not to eat really late, try not to eat sugar, white breads are not good for you because they become sugar, run, exercise a little bit, eat yogurt, get your colonoscopy when you're supposed to, get your mammogram when you're supposed to, if you smoked before make sure you do lung cancer screening. That's all we know. I guarantee you the result of almost any of these tracker things is basically one of those recommendations, maybe dressed up to be a little prettier, but that's basically what they're going to tell you. Most of the people who are using these already know the answer. Now, if it works for you, if it's the thing that gets you to not eat the donut at 1 AM, I'm all for it. Sounds great. To me, that's just a behavioral nudge, and it sounds good. But I don't think it's really health. I don't think that's healthcare. Healthcare is really about people who are sick or very likely to be sick — think pre-diabetic and beyond — and how do you either prevent serious disease, or manage early disease, or manage serious disease. It's not people in their 20s or 30s or really in their 40s. It tends to be the older segment. Every once in a while, someone will come up with a tool or diagnostic that actually works and is material. There's obviously been a lot of improvements in diabetes, just the actual physical hardware to measure glucose has gotten significantly better. That kind of stuff to me is really interesting because it makes somebody with real healthcare challenges' life a lot better. The over-the-top direct consumer stuff, I just don't buy it. Not that I think it's going to be zero from a financial perspective, I just don't think it matters. I don't think it's important. But as with all these things, there'll be a thousand things that are tried and most suck and don't work, and then one or two of them will actually be important. It's hard to know which ones. We were talking before, you can make the argument maybe in testosterone replacement therapy for men in their 40s and 50s, the data looks interesting. Again, what are you measuring? How do you know? They measured things like vitality and pseudo measurements of health. But hey, people feel better. One of the biggest lessons I had from working in oncology for 10 years was: oh yeah, that's healthcare. That's actually what we're talking about. We're talking about the cancer patient who has CHF and is in the hospital once a month. That's really what healthcare is. That's where all the money goes, that's where all the costs go, that's where all the inefficiencies are, that's where all the drugs are for. So when I think of healthcare, I think of Medicare, actually. The first thing that comes to mind is the Medicare population, 65 and older. There are other sub-segments obviously, but I don't think it's the Silicon Valley version of health: go to the gym.

In general, we look for founders. We do seed investing, so there's only so much you can evaluate about the actual business. There's not a lot of spreadsheets. So I look for founders who I call 'stretching up and down.' It's basically this idea that the same person can talk through the strategy, the customer of the business, the market, the money, the TAM, the margin structure, whatever, and then they can also talk to you in excruciating detail about the product. It's the same person. Because it's that little loop: I talked to this customer, I learned this thing, and that's going to influence the product. You've got to run that loop a few times. If it's three people, the quality of that feedback deteriorates as you pass the information along. So we look for founders that can stretch up and down. Then we look for founders who are likely to attract talent. That doesn't mean they have to be gregarious or the Adam Neumann show type. It's just some founders have a thing where you talk to them and you're like, yeah, I would work with you. You're reasonable, humble but opinionated, you seem to know your stuff, you ask good questions, but you don't just give up when someone pushes back. We look for that because the hardest part is to build the team. If you can't attract good people, you don't really have a great shot of building something big. You can build something small. Then on the idea, we look for somebody that really understands the details. They can talk about the technology, the market, tell you why they made certain decisions. For most startups, that's fine. The challenge with healthcare is you could do all of that and just have the wrong idea, and it still doesn't work because healthcare doesn't allow you to make mistakes in the strategy. It beats you up immediately because there's no pivoting in healthcare. You kind of pick your spot and you're there, and it's really hard to switch the industry, the disease, the customer segment. So for those, we're a little pickier. I want to make sure this person actually understands what they're getting themselves into. Who are you selling to? Why would they buy it? How are they going to pay for it? What does reimbursement look like? We really think about the business model. If it's just a SaaS thing for developers or some simple business model, we're not going to scrutinize it as much. I know how software works: you build it once, you sell it a thousand times, the margins are really good, done, it's a great business. It's the ones where you're like, how do you make money exactly? Those we try and tease out. We've done a good job in healthcare in particular because I know it well. Software I think is a little harder sometimes. If you ask me what the latest and greatest in the language models look like today, I don't really know. I know they're interesting and important and better, but could I tell you the frameworks people are using to do it? I've lost that skill.

I don't know what it means. Please feel free to define it, but I don't know what it means.

Very good. I'm cool with crowdfunding. I mean, that's basically what it sounds like to me: hey, we're just going to crowdfund something. I too have used Kickstarter, and that sounds great. If a bunch of people want to pull their money and fund a science project and vote on which one to fund, sounds good. You don't need the blockchain for that, you can just do it. And then the data sharing side of things, I think what almost every entrepreneur quickly learns is that it's not a technology problem. People don't share data not because they can't figure out how to — GitHub exists. We have unlimited ways to share data. The issues are incentives and what you share and why. If you think of academia, academia rewards scientific publishing in most cases. So you create this awesome new fancy data set, and now we're showing up and saying, hey, I know your career, you're probably in debt, is relying on publishing on this unique data set. Could you just share it with the rest of the world? Most people aren't going to do that. Reality sets in: people have money needs, they have a job, they have kids, they need to survive. So I view this — I think the goals are great, I think the implementations are just a fancy brand on something we've tried before, which is crowdfunding science. I wish there were — the reality is science is really expensive. A few million dollars from a few people here and there is going to do nothing. It's just not big enough. What's the NIH budget? I feel like $100 billion? I don't know, it's big, and everyone already complains about how small it is. Pharma R&D budgets are a few hundred billion. I think it's just a hammer searching for a nail. We have this tool, where can we apply it? But again, if people want to pool their money to fund an interesting lab, sounds great.

The patent that matters usually in biotech, in therapeutics, is the drug patent. It's the patent on the end product, on the molecule that you put in a human, because that's what protects your ability to sell it at really high prices, and that's the incentive to develop it. In our model, we're not licensing existing drugs, we don't have to think about whether this drug patent is an issue because we're not at the drug yet, we're working towards the drug. So the question is, is there something — eventually once you get to the drug, you patent it because it's new and novel and it's yours. Then the next question is, is there something in the process that is particularly unique and protectable that you don't own? We do more defensive patent checking, make sure you're not doing something that somebody else is going to sue you for. We have a law firm that does that check. We don't really care so much about whether your process is patented because that's not where the value sits. The value sits in the drug. The other thing is, when you file a patent, you have to reveal quite a bit about what you do. I actually think there are some good arguments that you're better off thinking about it as a trade secret that you just kind of guard rather than patent. It's very hard to win on process patents. So I think it's a common misconception in biotech of what the patent does and doesn't do. The patent, again, you can think of it as at the end of the conveyor belt, that's the patent, the thing at the end, not the front. I agree with you, by the way, there's a lot of benefits of open source in software. That's why it's cheaper. Bio's still expensive. We're going to write kind of $5 to $7 million checks instead of $1 to $2 million checks, so there's your delta right there. It's still pretty big. But the vendors — you buy the hardware from company A, the reagents from company B, the pipettes from company C, and you kind of stick it all together. Just because it's not open source doesn't mean it's not available, it just means you have to pay for it. So it's just a little bit more expensive. It'd be great if every piece of the stack was free, but you also have to pay for AWS, you have to pay for the database, you have to pay for all the layers on top. So software is still better in the sense that it's easier, but we're chipping at it.

Well, what I feel like — at the end, I'm going to give away all the tips and tricks, and then there's my edge. People like to learn stuff in a conversation. They don't like to be just on the receiving end of questions. They want to learn something. Actually, if they learn something from you, they're much more likely to share something they know as well because they feel like they're talking to someone useful to them. So I always tried to make sure I had something to teach, something that would engage this person and make me feel credible in some way. For me, it was about software. I would always talk about software development and that I worked at Google. You hear people like, oh, Google, that's really cool. I'd have one or two Google stories, and then you'd get some credibility, and then you could start to ask. So I think of it as: what can I share that gets you to believe that I'm not an idiot? We worked on that, we practiced it. The other thing was, I think you can use not being the expert to your advantage because there isn't the 'I'm trying to one-up you, who's smarter' type of thing. I still do this in biotech: I always try and couch my questions with, 'Look, I'm just a software guy, last time I took chemistry was in 11th grade, so let me just anchor you there. Half these words I don't understand.' Then you can go and ask the question, and you find that the person on the other side is less defensive, they're more willing to teach, they're more willing to get — now you have to have some credibility, you have to know something. So you just have to find the balance between the two. This is the human element of being a founder: convincing people who have no business talking to you to talk to you, whether they're employees, experts, advisors, or investors. So we just look for tactics like that. I still think that 'I have something to teach you' is by far and away the most effective tool I've ever had. I would just repeat that one over and over. I have a little war chest of stories I can tell that are tight, two to three minutes, because you only have that long to get someone's attention. And then the imposter syndrome thing, I still have that, but I just have to put it away.

First of all, Flagship doesn't put $25 to $50 million into every idea. They put them eventually into the idea. My belief is that's the equivalent of going and raising a Series A. We're not telling you the $5 or $7 million that we're going to give you is going to get you into the clinic. It's just going to get you to the A, and then at the A you go and raise that $25, $50, $75 million. The difference is that you get to raise it from the market at a price the market sets, meaning you get to retain ownership of your own thing. Whereas if you're an employee of Flagship, you are raising money from your boss at any price they want, and you own 1%, maybe half a percent. So there's just a total economic shift between whose company is this. Is it my company or is it their company? With that said, there are certain ideas that are just so risky and so unique that the market wouldn't fund them anyway. If Flagship wants to fund all those, that sounds great. This is an 'and,' not an 'or.' There are some ideas that work in this model, some ideas that work in that model. It's more about who is in control of the business. I view it as we would apply to everybody, whereas Flagship only applies to its employees where the ideas come from. That creative spark — 'I wonder if that would work' — in an incubating model, that spark has to come from one of however many employees they have, 260, whatever it depends on the firm. In a market model, that can come from one of a few billion people, or a few hundred thousand people of the skill. So I'm going to take that market of ideas over the tight venture firm market. But not every idea can be done in this way. Some ideas you need $50 million to buy the hardware because you're searching for new biology, and that model is needed. So I think of it as an 'and.' The best heuristic I can give you is: imagine for a second you are a reasonably mid-career person, probably in your 30s or late 30s in biotech, you work at a venture firm like Flagship or Third Rock. If you're going to play, and you've got an idea and you think it's good, do you want to start it at the venture firm you work for, or do you want to start it and own it? Today, your option is primarily just to start it at the venture firm, and then you are essentially a co-founder in title, maybe you get a job, maybe you own 1% of this thing. The company gets formed and started, it looks really good on your resume, but financially it kind of sucks. What's your other choice? Our thesis is that other choice sucks too. There's just not a lot of great places that you can go to do this idea on your own because Flagship has a lot of really good resources, incredibly smart people, it's an unbelievable place. You have all these resources, it's easy. Our job as a firm is to recreate that feeling but just outside of the firm. The reason I don't like the Coursera example is I view it as Coursera with the Harvard professors. We poached all the Harvard professors and said, now we're going to bring it out to the masses, but you're going to get the exact same quality people in a different world. Whereas Coursera is kind of lower tier theoretically. That's the idea. It's a really good question because the nuance here is hard.

A lot. They changed as I pitched them. In that same way of seeking expert feedback to find the idea, the same process of validation: go pitch it and see what the reactions are, see what the questions are, go back to the expert. The Curie model didn't exactly look like this in the beginning. We weren't 100% sure of the structure. We kind of discovered it as we went along by just having conversations. That's just the natural evolution. By the way, that question is exactly why in founders we look for people who can do that with just one person. Go have the idea, go have the conversation with the investor or customer, hear something like, 'Ah, this isn't going to work, something's off,' and then have the next idea of, 'What if I change it this way?' and then go back out and pitch it again. Run that little loop. If you have to go back, type up your notes, send your notes to the other person who's doing the sales pitch, have them go out and do the pitch, they type their notes — you see how inefficient that is for finding what is a good idea. That's why we look for it in that one person. Sometimes it's two because they do the call together. But they've evolved quite a bit. Each one has evolved. Flatiron a lot more because we didn't know anything when we started. This one a little bit less because we had a little bit of a head start on the idea.

We did everything: investor, customer, expert. In our world, it was investors just because that's what the business was. But I don't just mean investor, I mean anybody knowledgeable about the industry. The hard part about all of it is sometimes you get a really smart person who tells you something that's wrong. It happens more than you think because maybe they're really good at this but not good at that, and they don't realize it, and they'll say it with really high confidence. There are four co-founders of Curie. Two of us are the business people, and then there's the real deal, the actual people who do the work. We just do the business structuring. He taught me — I love this question, I'm still learning, I'm pretty good at this, I still find like, oh my God, I wish I knew this tactic. Ask somebody how they know. 'Okay, that's not going to work because XYZ. Really interesting, that's cool. How do you know that?' If they can point you to a paper, an article, a fact, then you go, ah, okay, that person actually knows something. Most of the time you will find they are just repeating something they heard and they don't actually know it. It doesn't mean the fact is wrong, by the way, just means they don't know that the fact is right or wrong, they just repeated it. You just have to do that in a very gentle way because people get very defensive. Sometimes it's very uncomfortable. You're in some of these calls and you can tell that person is pissed that you asked them that question. Usually what I've seen is the people who get defensive don't know what they're talking about, and the people who do are like, 'Oh yeah, I read it in this paper, you want me to send you the link?' It's so easy and clean because they actually know where they found that fact. You just have to use that tool kind of sparingly. It can be decisive. I just learned that one like a year ago because I was always trying to tease it out. He was just like, just ask how they know it. That's a way easier tactic than I was using. Only took me 16 years to figure that out. So yeah, some people are very good at this. Learn from them.

Well, someone offered us $2 billion, so it was pretty easy. I was like, well, I own a decent — yeah, that's pretty good. That's the actual answer. Why? We did a model financial model of what the company looks like in five years, what the growth rates would look like, what the margin structure would look like. We basically had our CFO model what this could be if it were public, and we didn't like what we saw. Without going into all the details, you saw this weird revenue S-curve where it was going to grow and then it was going to grow slower. It had to do with the dynamics of the market: there's not enough pharma companies to sell this to, you kind of run out of customers. Very weird. We looked at that and we're like, anyone have any genius ideas of how to make the curve not do that? Debated it for a while. I had a bunch of ideas that I had no confidence in — maybe this thing could work, this could work — but they weren't products we had built yet, they weren't products we had tested yet. So I just was like, we have to sell this because we're not going to like it when growth slows. Then we ran a clever process to sell it. Didn't start at $2 billion, it was a lot lower. We found the right moment in that curve to go out, I think, because it was still growing and there was still a lot of hype. We're like, let's go take advantage of this. We got very lucky in the timing. If we had sold it today, there's no way, it's probably a quarter of that, is my guess. Now we're seeing people who were overly optimistic about their own thing pay the price because the world has reset. It's the hardest thing I think about being a founder — not the hardest, a hard thing — is you spend all your time talking about how awesome everything is because you have to recruit and sell, but then in the back of your head you have to have a reasonably objective view of how good this is, especially if you get offers to sell it along the way. So that's why we sold it. It was a good decision. Knowing what we know now, we made the right decision for sure. The first one, I look back on it and I'm like, yeah, we nailed it, and we didn't realize it. We were at the beginning. I thought the programmatic ad tech stuff was going to grow, and then it just took off. We underestimated the curve. Part of it was I was 22, I didn't know how to do anything. Someone offers you $80 million, which is what the offer was at the time, you're like, I'm going to take this. But we had a competitor who started after us that's worth $25 billion: The Trade Desk. Now I look back on it and I'm like, we were in the lead, that would have been — we had a shot. I don't think we would have been able to pull it off. I don't think we knew how to run a company like that, and the guy who started The Trade Desk is a very seasoned operator, a really good entrepreneur. That one I definitely think about.

I didn't want to do enterprise sales again. I did it, it sucked. It's hard, it's a lot of get on a plane, put on a suit, pitch to a room, hope they like it, then pitch to the room again, do it for like seven years. It's tiring. I definitely wouldn't have started another company if I thought I was going to have to do enterprise sales again. I didn't not do it because I thought it was a bad idea, I just didn't want to do it. I was done with that stuff. Now ironically, I end up pitching quite a bit, but it's mostly to investors. Definitely, it's tough. It's a lot of work, a lot of hotels and stuff like that. So I didn't want to do that again. But if I were starting from scratch and I had a good idea, I wouldn't say selling to biotech and pharma is a bad thing. I would say selling just to pharma is a bad thing because you really limit how many pharmaceutical companies there are. 20, 30 that matter. So if your market is companies with approved drugs, it's going to be tight, it's not that many. And they consolidate. You get one customer gets bought by the other customer, they look at the contracts, they're like, oh, I'm going to combine that contract. Versus selling to discovery groups is much bigger, much broader, much wider customer base. Part of why we're there in discovery. Pharma stuff is really tough.

It's up. The percentage of Medicare spend on therapeutics is a good one to look at. Therapeutics broadly as a percentage of total spend is trending up. There are a few interesting things about that statement. One is, do I think the drugs are getting better faster than the care? Yes. I think primarily because you're rewarded heavily for better drugs than you are for better care. The better surgeon doesn't make $50 billion, you know what I mean? It just doesn't happen. So there are just more people trying to innovate in therapeutics than there would be in care delivery. To the point where sometimes we hurt ourselves because we only view innovation as a thing regulated by the FDA, where all the reward comes from: pill, biologic, whatever. This is why some of the digital health stuff is a real challenge because there's no payment. Just because it's an app or a behavioral modification, we're like, ah, that's not a drug, so we don't pay for that in the same way. It doesn't have its own J-code. So all of a sudden, these actually pretty interesting digital therapeutics that actually work, getting them paid for is such a disaster because it's not a drug. That's a weird history of rules and regs that should be changed. My opinion is, if you run a clinical trial, a randomized clinical trial, that shows your thing is better than the standard of care on an outcome measure that matters — survival, whatever, or your A1C level depending on the disease — you should get paid for it. Why do we care if it's an app or a device or a drug? We shouldn't care. It just works. But that's not how it works, which is odd. So you see a lot of innovation in therapeutics because you get paid, and everything else is a little slower. Do I think it's going to continue? Yes, for a while. We're just getting better at making drugs, so there's more stuff. It's going to cost more. This idea that we keep people healthy longer and that's cheaper? No, it doesn't make any sense. They're alive for longer, there are more people you're paying for. Mathematically, it doesn't make any sense. Then you have this other issue: you have this population boom in Medicare coming, but we don't have the replacement tax base for it because people stopped having as many kids. So you're keeping the older people, who are the most expensive and the sickest, alive longer to use more healthcare, but it's paid for by basically this crew. That's going to be an issue. It's not a Democrat or Republican thing, I just view that as the math. Somehow we're going to have to figure out how we handle this because the demographics don't look great. I think we might see an Alzheimer's dementia drug that really works. If that happens, that alone will pop the budget. Think of the number of people who are going to suffer from it — tens of millions of people. Just one of those drugs could pop the budget. We have these new weight loss drugs coming out from Eli Lilly and another one. They're good. The history of weight loss drugs, none of them work and they have really bad side effects, so they don't really do much. This new one — I can't remember the name, Mounjaro? — they work, which is shocking to most people. There will be a second generation and a third generation that are even better because that's how drug discovery works: you tweak and tweak and tweak. That category is tens of billions. Weight loss affects everything. So I see this just doing that for a while, and at some point, something's going to break. The thing is, if they continue to regulate price — some of the drug prices are crazy obviously, you pay a lot for stuff that doesn't work, but you also pay a lot for stuff that does work. If we don't distinguish between those two, eventually the incentive to build the stuff that does work goes away. The U.S. political system doesn't allow for reasonable debate and conversation about maybe we should try and think about the efficacy of these drugs before we — my thesis has always been the right approach is: I will pay you more, way more, for the good ones. You want a million dollars a year for something that works, that cures? Go for it. We'll pay you $2 million a year. We're not paying for the shitty one. The problem is we don't agree on what works. What does 'works' mean? How much is this worth? Then you get into all these conversations that are more philosophical and ethical in nature. It's really hard to get anybody to agree on this stuff. And pharma has a good lobby, so that hurts. I feel like that was a very negative take, but that's kind of what I think.

In general, you know stuff is working or can work when smart people are telling you it can work. It should be more obvious. You're pitching people and they're like, yeah, that actually could make sense. You find a bunch of the skeptics who normally would say this is not going to happen, and they're actually interested. You find the experts. That's what started to happen with Flatiron eventually. We started pitching a bunch of healthcare people, and they were like, actually yeah, that could work. I don't know how you're going to do this, this, and this, and I'm like, I think I have an idea for those. But does the business work? Can we take the data from here, clean it up this way, and sell it over here? Does that work? We started to get a consistent 'the math makes sense' from doing napkin math. But we were hearing it from smart people who knew what they were talking about, not your friends. You want to hear it from the expert in that area. Same thing happened here with Curie. I would pitch it to a bunch of these old school VCs who had been in biotech for a while. At the beginning, they'd be like, oh yeah, sure, you have this new model for biotech, all right, little tech boy. Eventually, I started to see them go like, okay, yeah, actually this could — no, that makes sense, that structure could work. You see them physically lean in. They would do this. That was when I knew it had real legs. But it took a while to get there, a year, a year and a half.

You have to practice the story, practice the flow. I would actually suggest for most people you should write it down. When I go out now, I don't have to use the deck because I've done it so many times, but in the beginning I had a deck, I had a story, I had an outline. I still have my notes. When I get on a call today, I have the Zoom thing on here and I have my notes over here in case I forget something. I want to make sure I hit these points. I've literally done this pitch probably 500 times, same story, but I still keep the notes up. So it's a skill. You have to practice anything to be good at it. You just practice it. Learn your fact, what's the thing you're going to share that's interesting to them. I think people underuse expert networks. There are groups — you have to pay for it — like Guidepoint, GLG. You can be like, find me an expert in this thing who's done this and is willing to talk to me in the next five days. They will do that pretty well. It's going to cost you $1,000 an hour, so you better use the time wisely and be good at it. But you can get to a lot of people that way. Sometimes they'll recommend someone else. Even at Curie, we use it. We see some idea, we're like, I don't know if we know this area. Guidepoint, find the person, do the call, take the notes. As with anything, it's a skill you have to practice, tweak and tweak and tweak. Even with the Curie story, it took us a while. We didn't have a hook. It took us a while to find that pithy hook. My co-founder was like, 'You can't start a therapeutics company in your garage.' I was like, yes, that's it, because I can stick that on a slide and someone will go, oh, that's an interesting statement. It pulls you in a little bit. That took two months to come up with. I don't remember exactly how we did it before, but it wasn't working as well. We put that on the slide, and it started to work. There's a lot more failure along the way than people realize. It's just like you're trying to fail forward. By the time we were done, I didn't use the deck because I could tell the story, but for the first 100 calls or so, I used PowerPoint.