This weekend, Sam Altman got fired from OpenAI, and is in the midst of being courted back to run the company as CEO. Working on AI is existential in several ways and in a world that, by many definitions, continues to get crazier, offers a mechanism towards real progress as a human society. The response over the weekend has made me a little envious of the work environment that was created at the company. OpenAI holds a strong density of technical talent, that also thinks deeply about their role in the world. The pace of research output and societal impact of the company has by all metrics been incredible and has prompted reflection about my own work and goals. In this post, I’m going to talk about progress and success in the life sciences.

Progress

In some ways I think the ‘energy’ of OpenAI is unfortunately missing from the biomedical sciences. Many have previously complained that our industry does not have an Elon Musk type figure that catalyzes the recruitment of talent and investment into the sector and as a result, our industry certainly doesn’t have a Sam Altman or a comparable ecosystem to YC.

The reasons I think are mainly cultural, but could also stem from intrinsic business dynamics. There are fewer billion dollar exits in biotech and no equivalents to Google, Amazon, and Apple. Returns are flatter; far fewer millionaires and billionaires have been created by biotech and biopharma than tech. Maybe the biotech industry is missing the capital philanthropy that seeds Silicon Valley. Are people in the life sciences just intrinsically less motivated and worldly than people in tech? I don’t think any of these explanations are true. In any case, there are deep systemic issues that will be difficult to change.

Product development is costly

Converting talent and hard work into revenue is far more inefficient in the life sciences than in tech. If you don’t have already derisked foundational IP, you probably aren’t even NPV positive. When you start making money through licensing deals or milestone payments, you likely needed to sell a substantial stake in the future revenues to get there. Even executing on such a deal requires you to find a partner that A: believes in your technology/product enough to be in it for the long haul, and B: has synergy in some aspect with the product you are selling. In therapeutics, this can be difficult to find, as you are essentially in a market place with few buyers (large pharmas) and many sellers (startups with ‘exciting’ platforms). What you build will never be immediately useful, so as a founder you have no leverage.

When was the last time you saw a deal where the terms were favorable to the startup? Money needs to be cheap. The stakes for failing need to be low. The cost of even knowing whether something works is tens of millions of dollars; proof of concept comes unusually late in the timeline. Technologies that make things orders of magnitude cheaper aren’t necessarily enough to win. That alone buys you a lottery ticket. In tech, tickets are ideas and in the hard sciences, tickets are defensible and useful IP. The issue is that the expected outcome values are the same, you just get less return on your investment in the sciences because your upfront cost is way higher.

If for whatever reason your IP position isn’t airtight, you can and will get sued to death. The recent NanoString — 10x Genomics patent litigation is an illustrative example of this. NanoString is a company whose products (GeoMx and CosMx) are more technically advanced than their counterparts (Visium and Xenium) and whose scientists made major R&D strides, just for them to get screwed by a patent issue. The complaint from 10x, essentially that NanoString is infringing on in situ hybridization is highly debatable and a case unsuited for America’s jury based court system. The 10x sales team owes their bonus to their legal team.

Bias towards credentialism

The deal cycles in R&D based industries are long and require trust. When there is cultural inertia that values grey hair, it makes it harder for younger and less experienced company builders to win. If you don’t have a terminal degree (PhD/MD), it will be very difficult for you to gain traction in a room of senior scientists with 200+ combined years of experience. Unlike in tech where you can get by with an idea and some big name investors, you need a track record in biopharma. You need to have led BD or R&D before. The stakes are too high to waste time and money on an inexperienced first time founder.

Furthermore, there is no ‘sales’ in biotech. Your technology either works or it doesn’t and there is no engineering you can do to quickly make changes to make a customer happy. Biotech is an industry where you need to do all your sales and marketing (ie. showing everyone that it works) prior to selling your product. Before a product is ever sold, you have years of technical details described in papers. The intricate details of whatever is being sold is intricately described and protected in patents. If you have a product with spectacular claims and you haven’t published or otherwise publicly disclosed any data, this is a big red flag.

Is this the right thing for the industry? I think so. This isn’t an industry where you can get an intuitive feel and preference for a product like in software. Lessons are expensive. Making the wrong choice without the totality of the available data for the scientific community to scrutinize should be limited as much as possible and publication costs are a reasonable price to pay.

Lack of role models

Academia is full of role models. So is tech. It was quite surprising to me that there is no one that people really look up to in the life sciences industry. Where is the equivalent of Greg Brockman or Andrej Karpathy in the biotech industry? Its not that there are no impressive people. Luke Timmerman and now Brad Loncar interviews some impressive people in the life sciences, but these are typically C suite executives who can be deetached from the science and often speak in generalities. I want to get to know the senior chemists at Nuvalent or the biologists at Vertex. Give me a window into their personalities and habits so I can emulate and learn.

People in biotech and biopharma are bland. This definitely can and should change.

Perfection over speed

Biotech is a decisions based industry. It is far better to be right and slow, than fast and wrong. There is no move fast and break things.

As a result,the sense of urgency is de-emphasized. Iteration cycles are long, and the industry is less likely to attract people who strive to move at rapid pace. If you think about how people learn how to code versus how people learn how to do genetic engineering, the costs and timelines are vastly different. Unless your high school had an iGEM team, your first experience in a lab was probably highly structured (if in a school setting), or under supervision of a graduate student or postdoc who may or may not have had the bandwidth to train and teach you to think independently. Running a miniprep, designing primers, or running gels are not difficult activities. However, they are 100-1000x more expensive than playing around on Replit. Its like picking up tennis versus endurance running. One is far more accessible.

In tech, raising money is used to scale and move faster. In the life sciences, you will die without money, and that gives investors powerful leverage. You need to be careful and not waste peoples time and capital. The industry has no room for Ryan Breslows and Domm Hollands.

Towards a more ambitious biotechnology industry

Are any of these ‘fixable’? Are they features or bugs? Product development will forever be costly (and perhaps even more so in the future), there will probably also also be a bias towards credentialism and perfection over speed. However, I am optimistic that biotech will eventually find better role models. People are more keen to write on Twitter (ex. Keith Hornberger) and the younger generation of new students in biotech appears set on a more exciting path (Nucleate, Petri, etc).

What is the best way to incentivize breakthrough science? Scientists are choosing between the apparent freedom of academia and the resources of industry. The advantages of industry include salary, clear objectives, competent colleagues, and potentially exciting and resourced projects. This is quite attractive if you are an investigator that can ‘sell’ projects and competently lead a group of translational researchers. This is especially true if you can do so without damaging your reputation from leading projects that stall or were found to be poorly motivated. Think about how you might perform as a research scientist at DeepMind, OpenAI, or DE Shaw research. The downside of course is that these positions probably don’t exist for bench work. In the life sciences, being an individual contributor at the bench is much harder than if you are simply responsible for writing code.

In academia, the main advantage I see is the relative freedom to explore and pivot, opporunity to collaborate. However, an unfortunate part of the job is using most of your time training people and filling out paperwork. It is harder to fire people and alignment of incentives can be poor due to the diversity of stakeholders. The grad students wants to graduate while the PI wants to publish in CNS.

So how do we get to a sort of Bell labs for the life sciences? I think a pretty clean answer is foused research organizations with clear stated objectives and stopping rules. These should be filled with professional PhD+ scientists taking on career defining ambitious projects. These people should be less motivated by citations (the notion of an H-index sounds crazy to anyone not in the sciences), and tightly focused on performing work that is NPV negative and outside the scale and scope of the academic environment. Similar to Janiella farm, you could envision creating an institute that houses many of such FROs, professional scientists only and an environment that recycles talent and ideas once projects have run their course.

Success

With these stated constraints, the question is: where do you try to make an impact? A push towards progress inevitably comes from the personal success of individuals.

The traditional advice for young people is to work for great people and work hard, usually in that order. Is there a PayPal mafia equivalent in the life sciences? In this aspect, I think the life sciences industry excels. There are certainly pockets of excellence both in industry and academia. From Karl Deisseroths lab, Feng Zhang and Ed Boyden. The great scientists behind Array and Genentech have recycled back into the ecosystem with great effect.

However, you can pick up lots of behaviors that help you succeed but eventually you hit a point where the rate of improvement stalls. You become more reliant on your internal drive to improve and instincts to seek out new information. Highly productive people realistically have 60-70 hours of work each week on average, depending on how you want to balance things. The objective in this second stage of learning is to injest novelty and to develop a personal style. You need to think clearly about what exactly you want personally and design a path forward to get there.

Defining success

Science is full of side quests. Science is diversified enough that everybody is not collectively working towards one big hairy audacious goal. As a result, you need to develop your own internal metrics for ‘success’, tailed to circumstance. You could optimize for awards or the admiration of your peers. Some optimize for hard outcomes like drugs approved or patient lives saved. There are surrogate markers like papers published or citations. Sometimes these metrics are correlated, and sometimes really not. But you have to appreciate that these metrics don’t always reflect how hard you work or how smart or capable you are.

‘Trying hard’ simply results in more volume. Is volume what you want? Is taking on more and more projects useful for your career or for anyone else?

You need to understand what your end goal is. What is the measurable metric or objective you want to reach. Do you really want to do research the rest of your life, or do you see your research work as a bridge? What is your timeline? How old are people when they make their major contributions and who are your role models? What can you do now to get to that age where you are consistently releasing great new work? In the sciences, few things actually require grunt work. Most projects if you knew what was going on, could be completed an order of magnitude faster than they play out naturally. Read widely and triangulate your way to a solution. You don’t need to produce all the data yourself. Much of it has already been collected.

Systematically identify your weaknesses and whether they matter. If they don’t matter, ignore them.

Picking a problem to work on

Science is a calling, no one is forcing or even asking you to learn about the world. If you want to go join a hypercompetitive lab, you can. If you want independence, flexibility, and balance in your personal life, you can get this as well. What you will eventually work on probably matters a lot to you in particular but to most people, no one cares. No one knows how drugs work just like no one knows the details for how your iPhone works. You need to develop an intuition for what the important problems are and when you reach a certain level of maturity forget about what other people think. This intuition you can’t get from anyone else; it has to be your own style of thinking. It is what makes a contribution yours.

Science progresses in stages of rapid expansion and consolidation. “Novelty” is often a consequence of expansion while inference is a consequence of consolidation. You can try to pick your battles carefully to ensure you are on the right side of history — that is, working on consolidation right after expansion and expansion in fields that have already consolidated. However, if you keep chasing what is ‘hot’, the contribution will never be yours; it will be the field’s.

Almost by definition, science is an exploration of the unknown. You can’t know what the result of an experiment will be and as a result, I don’t think you can predict or industrialize discovery effectively. You really need to do is maximize the number of spins on the wheel of fortune. Pick something where there are many avenues left unexplored.

Success in tech versus science

A few years ago, Sam Altman published a blog post describing the lessons and traits that are exhibited by successful people.

In list order, these are:

1. Compound yourself
2. Have almost too much self-belief
3. Learn to think independently
4. Get good at “sales”
5. Make it easy to take risks
6. Focus
7. Work hard
8. Be bold
9. Be willful
10. Be hard to compete with
11. Build a network
12. You get rich by owning things
13. Be internally driven

These are all good and well supported but do these apply to scientists? Scientists also publish papers (in academic journals even) describing how to succeed as a scientist (1)(2). One of the best I’ve read is a transcript from a talk of Richard Hamming’s titled “You and Your Research”.

There is a remarkable similarity between the themes discussed there and the list above. Compounding. Working hard. Working on important albeit risky problems. Being willful and focused and bold. Hamming’s scientist and Altman’s founder are similar. They preside over impossible problems and have ownership of a part of the world that they create.

Science can be depressing. In the end, it is more zero sum than most industries. There are ~50k PhD graduates each year. Thousands more come to America from abroad. There are a set number of facts to discover and an expanding set of people and resources dedicated to uncovering them. The literature is impossibly dense and can often feel like an exercise of finding needles in a haystack. Most articles are trash or fake. These are all obstacles that don’t exist in tech.

Accelerating science means finding more people who both put in the time to understand and develop their own perspective. Ones who are unafraid of the fact that science gets harder and still are curious to find a way forward. Most of what people work on will be inconsequential, which is fine. But I think the world could benefit from a few more focused and impact driven scientists. It requires a special ability to triangulate; a holistic understanding of what is needed and what gets people excited. The core competencies that make the best scientists great are similar to those that make the best founders great. Clear vision, a path forward, and the tenacity to make things happen.