JEFF BERMAN: This week on Masters of Scale, Reid Hoffman is here to tell us all about the new company he’s launched, Manas AI. To say that Reid has formed brilliant team would be a big understatement. To say that they have ambitious goals would be an even bigger one. Reid has put some of the world’s most powerful AI tools in the hands of some of the world’s best researchers. The mission: to discover new ways to treat disease and to actually finally cure cancer.

[THEME MUSIC]

BERMAN: This is Masters of Scale. I’m Jeff Berman, your host. We’re checking in with our very own Reid Hoffman this week whose 2025 has had a very busy start. In addition to publishing a book, Superagency, in January, he also co-founded a new company. We thought it would be interesting to hear how Reid tackles starting and blitzscaling a brand new business from scratch. Reid, welcome back to your show.

HOFFMAN: It’s great to be here, and I think, at this point, it’s appropriately our show, but yes.

The motivations behind forming Manas AI

BERMAN: I’m as always thrilled to get to see you and chat with you, and especially so today because we’re going to talk about Manas. And will you just tell us what Manas is, what you’re setting out to do?

HOFFMAN: Manas is a new AI-driven drug discovery company targeting primarily cancer that says, we have this AI revolution which creates these enormously new better cognitive artifacts. And how can we deploy that in various ways to get drug discovery super-powered? And it came about because I had gone to my partners at Greylock, and I’d said, “Look, I think there’s going to be a lot of great investments and agents and productivity tools and work automation and a bunch of those things, and we should do all those, and we’ll help. But I’m really interested in this area of, there’s a bunch of things that this AI technology could really make a huge difference in the world, for societies and industries. And currently most of the AI people are all looking at all the software stuff.” And he said, “Well, what do you have in mind?” I said, “Well, drug discovery.” And so I went out and started talking to some of my smart friends, one of whom is Siddhartha Mukherjee, and—

BERMAN: He’s a very smart friend.

HOFFMAN: Yes, very smart friend, professor at Columbia, world renowned oncology researcher, Pulitzer Prize-winning author of multiple books, including the Emperor of All Maladies, just massively talented guy. And I’m talking to him, and he said, “Well, I’m an entrepreneur too.” And I was like, “No, I didn’t know that.” He’s like, “Well, no, I’ve actually brought cancer drugs to market. I’m not just casual about this. I actually know this stuff.” And I said, “Oh, that’s interesting.” And he said, “And this sounds like this could be amazing. Let’s talk about it.” That’s how Manas came about, and that’s what Manas is.

Inside Reid’s involvement with Manas

BERMAN: You’ve invested in, I’m sure, dozens — I’m guessing hundreds — of companies at this point. It’s very rare that you choose to be a co-founder, but you chose to be a co-founder here. What’s your filter for deciding when to invest and when to co-found, and why be a co-founder here?

HOFFMAN: Well, co-founder in this case is not quite the same thing it meant when co-founding LinkedIn, which is Saturday morning, I’m in the office. And so it is still ‘co-founder’ because it’s at least a day a week. But it’s when you have a CEO co-founder who is exactly the right kind of person that I know how to add in the way that a co-founder adds in versus, for example, a board member, which is different in these cases. I mean, there’s an overlap of the board member. I’m obviously a board member of Manas as well. And Sid and I spent a couple of months basically white-boarding out the entire drug discovery process from the, “I have an idea,” to a drug entering the market, and slicing it as thinly as possible to understanding what each of these stages were.

And then what we do is we talk about my current understanding of what AI is, what I would predict AI is going to go to in one to three years naturally, what things might be possible with AI that are not part of the natural trajectory that we could create or stimulate uniquely as business. And then going through all those, that’s when Sid was like, “Look, I think you need to co-found this with me, because in a sense, I’m the person who’s responsible for making sure the various kinds of parts of the AI talent.” I mean, he’s enormously talent guy, learns things lightning fast and has a very deep understanding and also has that sideways understanding you want entrepreneurs to have, which is, well, why is it this way, and why couldn’t it be some way that’s a lot better? Which is one of the things that’s fundamental to the entrepreneurial impulse.

And it was like, okay, great. Because A), it’s a huge mission. Cancer is a huge killer. It kills children. It kills healthy adults. It kills old people. It kills people in every culture, every society. So it’s like, okay, this is a huge thing. And by the way, there’s not just one cancer. There’s lots of cancers. That’s one of the reasons why it’s a world problem. It’s like, “Hey, we figured out this cancer.” And so it’s like, okay, what is the way that we make a dent on this whole problem? And that really matters — and then also, of course, when you transform an industry.

And so part of this is to say that there’s obviously a lot of things that the drug pharma industry is really, really good at. It’s been doing this for a long time. It’s added a whole bunch to society, but it’s very classic industries. It’s very rooted in the way they’ve been doing things for the last X decades. And it’s like, okay, so you could create a new one like the other giants because you’re bringing in new things. Those combination of Sid, the important target, possibility to transforming an industry and elevation of humanity, that all gets me into co-founding.

Leveraging existing AI technology vs building your own

BERMAN: I mean, that deep mission alignment is so clear here. And I’m curious, as you and Sid were on that whiteboard and you were slicing into the step by step by step by step by step, what did you see that got you fired up about how AI is such a difference maker?

HOFFMAN: When Sid and I went through the whole area, we abandoned any interest in AI things. There wasn’t a minimum of 10x and frequently it was like, “Okay, no, no, no, it’s much greater than 10x, if we can make this work.” And then focus on those areas. And then we didn’t go, “We’re going to invent all the technology.” We said, “Okay, which technology should we be building, and then which technology can we use?”

And that’s part of what led us into partnering with Microsoft because both Sid and I from two different angles had some understanding of where Microsoft had been working on some pretty unique technology that hasn’t really been fully advanced to the market and say, “Hey, this could be a good way for Microsoft to get in this as well, not just Azure, but some of the stuff that the excellent Microsoft research has been doing as baseline.”

Let’s do some areas where we’re building, do some areas where we’re deploying. Some of the deployment as open source. Some of the deployment as Microsoft stuff. Some of the deployment is other things that we’ve learned and discovered as you’re along the way. And then the way you put them all together, including the things you’re building as the angle for having the technology strategy and doing that, that’s much more my network than Sid’s network. Sid’s networks is the world-class scientists. And other folks in mine is the tech people.

BERMAN: I want to drill down into this for a minute because I think this is a really important part of this next phase of how people are solving problems with AI as a, to borrow Microsoft’s term, a copilot. We’re all by now familiar with ChatGPT or Pi or other AI companions effectively helping us figure out what’s for dinner. And most of us by now are familiar with something like Sierra for customer service or Harvey for legal tech, for legal AI. The blend of where you are using someone else’s AI, where you are customizing for yourself, where you’re building for yourself, help us understand how you make those decisions and how business leaders should be approaching these problems as they’re making these decisions for their own companies.

HOFFMAN: Great question, and a very good one for as to scale too, because generalizing out of the problems we’re solving to problems that a lot of people are solving. So for us, it comes down to a couple things. So one is anticipating where there’s going to be a really good continuing workflow where some other entity could be an open source thing. It could be us contributing to an open source thing too. An open source thing could be a proprietary model that we say, well actually, in fact that investment thesis of going forward on that, we will benefit from that, at least call it 70%, 80%-plus of that ongoing technological development, is going to benefit us and can be in the slot. And if we were to start doing it ourselves, we wouldn’t get something that was at least 10x better than what’s happening, then we should use the existing one.

So that’s one decision when you do the open source thing or when you deploy someone else’s commercial library or other direction. Another one might be, well, we’re going to use the commercial library right now or the open source library right now. But that’s a time-saving measure of us getting to market. We think we’re ultimately going to build something here, but we don’t need to build something here in order to get to market and get it going. So we’ll deploy it for now. And by the way, you might have a much cruder, not perfect. And then there’s the question of, okay, which things will be ongoing where we’re learning our ground truths or we’re getting the data where this gives us a ongoing substantive competitive advantage about things that we have as unique assets other than just the prospective molecules that we’re creating that could be cures for various kinds of cancer.

Then on that list, which of those things do we start right now? Which of the things do we do later? That gets to a high priority. Now, part of a course, what we have at Manas that is advantageous that I recommend to try to do to all of our listeners, is that I have a very good sense and a very good network of both where AI is now, where new techniques are coming out, where the current trajectory is going. And those kinds of things play in that kind of strategic decisioning. Then that’s one of the things that you need to do when you’re making these decisions yourself. Now, what happens with everyone in technology, especially in software, is it gets very non-invented here. So they go, “Well, I should just control my own destiny. I should just do it.” The problem is with tech, software tech is that it’s not build once and forget it.

Software tech has to be constantly rebuilt, reinvented, rebuilt, reinvented. And if you’re not in that theme, your thing will out-mode very quickly. It’s one of the reasons why most governments around the world have really broken RFP solutions for technology because they go, “Oh, I will specify the 150 requirements.” And then when whoever provider, usually pretty incompetent people, prime contractors, et cetera, when it comes to software, deliver something to you, not only is it not very good at the beginning, but it starts aging exceptionally fast from the beginning because it’s not part of this stream. So one of the things we think about at Manas, but also one of the things our audience should be thinking about is to say, “Okay, do I really need something specific for me, or can I be benefiting on the weekly, monthly, yearly reinvestment that company X or project group Y or something else is doing on this? Because then I don’t have to be using some of my relatively few resources for cycling in the future.”

And that’s part of how we look through it. That’s part of the reason why some of the stuff we said, “Hey, we’re just going to do that the standard way an AI person does,” or, “Hey, we’re going to just do that the same way a drug researcher does or a drug developer does. We’re just going to do that the same way.” Because doing that the same way gives us the ability to be deploying that not having risk, not having to invest in it, and you have to be very choiceful even as large companies about which things you’re investing in for proprietary reasons.

BERMAN: More with Reid on how he’s using a lifetime of scale lessons to supercharge his new company in just a minute.

[AD BREAK]

Welcome back to Masters of Scale. You can find this conversation and more on our YouTube channel.

Manas’ competitive advantage

One of the things that struck me, Reid, when you first told me about Manas was this beautifully complimentary meeting of you and Sid and your skills and your relationships and your access, your knowledge. You were also entering a category where there are number of companies already using AI to do drug discovery and to solve critical medical problems. And you all raised a lot less money than most of these companies have raised. How are you able to do that? What’s the competitive advantage that you have? And I appreciate it’s mission -riven, but still you’re entering a market where there are already folks who are steps ahead here.

HOFFMAN: Well, we don’t think they’re steps ahead.

BERMAN: Okay.

HOFFMAN: That would be a different question.

BERMAN: Fair.

HOFFMAN: Part of it is, so some of these projects tend to raise a whole bunch of money and throw the money at either the classic AI stuff or the classic drug discovery stuff. So part of the thing that we’re doing with the financing of Manas is to start in a classic Silicon Valley way that I find to be most successful how it drives, which is: start with a really focused project, a Series A investment and move to raising more money from your Series B and then from your Series E. But it’s as you accomplish things, you go.

One of the things that is frequently challenging and frequently predictive for successful projects in my experience has been when you raise way too much capital initially. And that then says, well, it’s “We’re going to try everything and go really bold, and do the whole thing from the start.” And that makes you — even with a lot more capital — much less likely be successful. So we will need to raise more money. We will need to raise it for clinical trials and bunch of other things. We’ll really need to raise a Series B for some of the technology, but we will do it based on having achieved some interesting outcomes.

Navigating regulation in the pharmaceutical industry

BERMAN: Yeah, it’s one of those interesting parallels between art and start-ups where constraints actually produce better results often because it requires focus and discipline and choice making. Reid, you’re entering a space with Manas that is extremely highly regulated. In the case of drug discovery narrowly and healthcare more broadly, do you see regulation as a constraint? Is this an area of concern? Can you work within the existing world? How does that factor into your analysis here?

HOFFMAN: Well, one of the reasons why I very rarely invest in regulated businesses and would be normally very cautious about co-founding them is that regulation always massively slows down innovation, and regulators like to say, “Well, in certain cases it accelerates.” Which is true, and they like to say, “Well, but if you do it really smart, it doesn’t really.” And actually no, it isn’t because in the natural thing for regulatory agencies to say, “I get penalized every time an error happens, and I get no upside for things working more efficiently or working more on target or anything else.”

So I just throw in every possible thing that could be answering a negative, and that’s how it works over time, whether it’s in the FDA, whether it works in the SEC, whether it works in the banking industry, et cetera. Now, that’s all part of the reason why I tend to have this maxim and say, “Look, you should do regulation when bad regulation is better than no regulation.” Whether it’s an industry, whether it’s a specific kind of thing, but you shouldn’t delude yourself that you’re super smart and a technocrat, and that you know how to do this regulation in a way that’s so great for the industry that you’re going to detail this all out.

And it’s one of the reasons why, for example, the Europeans have a super large problem with regulation, not just within technology, but also everything that goes to labor and a bunch of other things because they go, “No, no, no, we’re going to have it as a multi-page detailed thing.” This is not at all saying society’s better off without regulation. This is not all saying that parts of the FDA are absolutely critical in anything from food and drugs, and that there are things to do. It’s just that you need to have a certain epistemic humility and going, “Hey, we should be very focused versus trying to be, we are the genius technocrats.”

And so that’s the reason I generally stay away from regulation, and I only get into the areas where there’s regulation when I think, hey, this could be so great. This could be like industry transformer. This could be something that changes many thousands of human lives, and you could see its impact on a society basis of helping elevate society. Then, okay, regulation is a huge risk like other risks, but that huge risk plays out against that, and that’s part of the case in Manas.

BERMAN: Right, right. Good regulation is really hard, but it’s not an oxymoron.

HOFFMAN: Yes.

How Manas aims to move faster than its competition

BERMAN: How fast can Manas go? Curing cancer has been a moonshot concept since we had a concept of a moonshot. How quickly can this happen?

HOFFMAN: Well, I mean we’re early days. Obviously, we did this because we think we can accelerate greatly. We think we can get a lot more targets into the early process than the standard process would have. We think we have ways of evaluating those targets and molecules at a much faster rate than traditionally happens. We think we can be getting into clinical ground truth things at a speed that is on larger number and much faster. How fast it is is still TBD, and I think this is one of the things where Sid said, “Hey, welcome to regulated industry. Don’t quote numbers when you’re talking to outsiders.” So here I am following Sid’s sage advice and not quoting any numbers.

BERMAN: Most of the time when we’re having conversations on Masters of Scale, it’s about a company that has already scaled. We’re having this conversation because you are the OG Master of Scale, and this is as important as anything that we could be talking about in terms of the effect on humanity. When we come back a year from now and have a follow-up conversation, what are the markers that will determine whether you feel like Manas is on the right track?

HOFFMAN: We already have some of those markers internally. We’ve already had some prospective molecules. We’ve already had some of our technology development show things that we think help build secrets that we think no one else knows. So we’re already on track on some of that. But I think what we would be saying to the world or we’re on track will probably be a couple of the pieces of technology saying, “Hey, these are delivering these things.” We might be saying, “We have a set of interesting molecules for triple negative breast cancer, other kinds of things we’re looking at.” And our process has in fact gotten us there much, much faster than the traditional drug pharma start-up company. That would be some anticipated positive signs.

Using AI in day-to-day operations

BERMAN: You’re obviously using AI to assist in the discovery process and the testing process, et cetera. I would be stunned if you’re also not using AI in the day-to-day operations of the company. What’s different about how you and Sid and the team are building the company with AI that is applicable to other companies and other categories from the other companies that you founded or been a very active investor in?

HOFFMAN: Well, this will obviously scale as we get to it. I mean, right now we’re in raw tech development. So it’s tech development and also scientific development. So I would say today is probably a little bit more like some other companies, which is: it’s a research assistant for various kinds of things we’re looking at. It’s a communication and productivity assistant for generating flow of information and decisioning between us, and it’s a coding assistant for building certain kinds of code.

If you’re not in those basic things as a company right now, you’re well behind the AI curve and therefore well behind the curve. So all of those things, I think, are things that every company should be doing. Now, there may not be a science research assistant in your particular company, but there should be research assistants of some sort, global supply chains or competitive analyses on products or whatever that thing may be. And so those things, I think, are not new and unique things. Now, obviously as you get to questions around, say, for example, “Hey, we’ve got a drug, and part of the issue is really maintaining complex compliance to the drug,” then you can see how AI might actually help with complex compliance to a drug, and there may be some things that get to specific things that are still significantly in the future.

The evolving role of AI in society

BERMAN: Reid, before we wrap, I want to ask you a broader question about what’s happening in the world of artificial intelligence. In your book, Superagency, you talk about the doomers, the gloomers, the zoomers, and the bloomers in terms of the categories of how dystopian and utopian people are in their vision of AI. When you lean into the bloomer side, into the more optimistic vision of what AI can be, what are you seeing today that we may not know about that has you more optimistic, more excited, more leaning into the AI utopian vision of the future?

HOFFMAN: I’ll use an example of, I was hanging out with Atul Gawande. We were talking about his new book that he’s working on, and I was like, “Well, have you used Deep Research?” He’s like, “No.” I was like, “Okay, let’s go.” We went to my account in ChatGPT, and we did the Deep Research prompt on surgeons applying anesthesiology and improving their practice, and it was a really interesting nuance. It’s one of the reasons I’m going into the depth of it, because it produced exactly what he wanted, which was 10 quotations from different surgeons, and he looked at this and said, “Oh my God, this is gold.” And he sent it off to his research assistant. His research assistant came back and said, “Well, one problem. Nine of the 10 quotes are incorrect.” And you go, “Oh, hallucination. Terrible, terrible thing.” Basically, not working.

But then with the research assistant, what she did was she went and looked at the areas it was pointing to, and she said, “Oh my God, this saved me tens of hours of finding the right treasure troves that are the precise things that will be very helpful to this book.” And so even in this kind of hallucination case, because of the research analyst, it was actually still an accelerant. It just is an accelerant in a way that proves the point from my earlier book, Impromptu, and also part of what I’m trying to say in Superagency, which is human amplification with the research assistant helped make her massively productive very, very quickly. And of course, you should cross-check it. You’re trying to produce something. I do the same thing when I’m producing my own books, Superagency, Impromptu, et cetera. I of course cross-check things.

I don’t say what the GPT-4 output was. I just cut and paste it in because I am testifying to these words being accurate as a way of doing it. The hallucinations are improving a lot and doing stuff, but even the hallucinations are interesting, even in potentially importance of high truth fact-checking environments. AI is improving every month, and it’s part of the reason why we tell people “Go try it. Go play with it.” Because it’s not like, “Oh, I’ll wait until it stabilizes.” It’s not going to stabilize soon, and it’s already amazing in some regards. So go start leaning into it. Start seeking super agency.

Investment insights in the era of AI

BERMAN: Reid, as one of the great investors of really American history, certainly modern American history, and is someone who’s deep, deep, deep in AI, for the investors in our audience, probably 90-something percent of their pitches that they’re hearing right now have AI something. How do you separate the wheat from the chaff here if you’re an investor? How do you know what’s real, what’s meaningful, and someone just slapped AI on to be of the moment?

HOFFMAN: Well, when you get a pitch for an AI juice machine, perhaps pass on that. Although watch, after making this prognition, there’s going to be some amazing AI juice machine that I was a complete idiot in saying this. But it’s a particular bit of Silicon Valley history lore that’s a fun gesture. It’s almost like saying 42 or, “Santa Diva’s high school football rules.” It’s in the little gestures. Now that being said, there’s simple mistakes and complex mistakes. Simple mistakes are that does actually not in fact have anything to do with AI. “Oh, I’m going to use buzzword bingo, it’s AI, quantum fusion,” and be cognizant of that. The fact that AI is not a panacea for everything right now. It’s not a solve every problem. And even when they say it’s accelerating a lot and we’ll solve it next year, it’s unclear on that stuff.

So be smart about that. And the number of teams that are really accelerating the raw intelligence of new capabilities in AI is not a large number of N, so be careful about that. Those are all the simple things. Now, the more complex things is that, of course, part of the reason why we’re in the cognitive industrial revolution here is that adding intelligence to everything, whether it’s your PC or your phone, but also your speaker and your lights and your car and everything is going to be revolutionary to certainly human-lived experience. And so there’re going to be a bunch of new products and services. But a new product and service does not an equity make. And so for example, among other things is what will rapidly happen as a whole bunch of these AI models and capabilities, I wouldn’t say is a commodity, but is broadly available to many players. So that you say, “Hey, I can build an AI thing that can be a good tutor.”

By the way, I can do that today. I literally put in a little meta prompt into something like GPT-4 or Pi or others, and I put it in and say, “Don’t give the answer. Work the person towards the answer.” And then I’ve got a mini tutor right now, not having done anything. So they say, well, AI, I can create that. And I say, “Well, I’m going to work harder, and I’m going to make it better.” It’s like, okay, that’s not nothing. But you’re looking for the kinds of structural advantages that you would typically have in a business, it’s a systems integration into an organization, a school, a business. It’s a network effect.

It’s some set of things that you go, “No, no, no. This is a product or service that’s going to go the distance and compound in value and therefore create equity value.” And that’s the kind of thing to think about. And one of the things I think people mistake in this revolution right now is that they go, “Just because it’s AI, and just because it’s moving first that makes a great equity.” And it’s like, well, look, that’s better than no idea. But if you’re really investing intelligently and professionally, you should have incremental and better ideas than that, not just that idea. Other ideas as well. And I think that’s part of how to think about AI and investing.

BERMAN: Reid, I’m super grateful for you joining us to talk about Manas today. I’m very excited to follow up as the Manas journey evolves. Thank you for attacking this problem that has affected probably the family of everyone who’s in our audience, and can’t wait to talk to you about this again soon.

HOFFMAN: Me too, and Jeff, always a pleasure.

BERMAN: Thanks to Reid for sitting down to talk about his exciting new company this week. I know we all hope the scale lessons he’s learned over his career can help make Manas AI’s drug discovery research a monumental success. It has the potential not just to change lives, but to save them. To hear more about his new book Superagency, make sure to check out the conversation he had with our own Bob Safian. We’ll put a link in the show notes. I’m Jeff Berman. Thank you for listening.