New Science had six brilliant new fellows in 2025:

1. Maggie Li (tissue engineering, biophysics), an undergraduate @ University of Toronto.

2. Kiran Kling (climate engineering), an independent researcher.

3. Lev Chizhov (neurotech), an undergraduate @ École Polytechnique.

4. Kyrylo Kalashnikov (bio, physics, AI), an independent researcher.

5. Ryan Hassan (philosophy of science & technology), an independent researcher.

6. Suntzoogway (applied moral philosophy), an independent researcher.

I’m extremely excited about all of them, and it’s been a great pleasure to see the progress of the previous years’ fellows and grantees through the years, e.g.:

• Adam Green: independent researcher -> Founder & CEO @ Markov Bio.

• Diana Leung: independent researcher -> Researcher @ Arc Institute.

• Isaak Freeman: independent researcher -> undergrad @ UC Berkeley -> grad student @ MIT -> Founder & CEO @ Axonic.

• Adam Strandberg: independent researcher -> grad student @ Harvard

• Julie Chen: undergrad @ Stanford -> grad student @ Rockefeller University + Hertz Fellow.

Looking forward to an even more exciting year ahead!

Cheers,
Alexey

Transcript on New Science’s website.

Thank you to Graham Bessellieu for editing and to Ishaan Koratkar for the transcript.

In this conversation, I talk with Jacob Trefethen, who oversees science funding at Open Philanthropy, allocating about $100m/year. The discussion reveals how philanthropy makes decisions about which scientific research to fund, with Open Philanthropy's unique "hits-based giving" approach aiming for asymmetric upside.

Jacob shares insights on neglected research areas like Strep A (which causes half a million deaths annually yet receives minimal funding), the limitations of AI in accelerating scientific progress, and the challenges of balancing trade-offs when deciding between funding immediate applications versus more speculative basic research.

The interview offers a window into how Open Philanthropy allocates resources to potentially transformative science, from funding researchers rejected by traditional systems to debating whether certain scientific approaches deserve long-term, substantial funding. Throughout, Jacob provides a thoughtful perspective on what effective science funding looks like and how scientific institutions might evolve.

Topics discussed:

1. On Open Philanthropy’s decision process

2. Is it possible to make expected value calculations for basic scientific research?

3. Non-profit funding vs for-profit funding

4. Why the NIH rejects grants

5. Jacob’s “PhD” working side-by-side with two biochemists for 5 years

6. Today’s system of federal funding are not supporting generalists

7. How will AI affect the pace of scientific progress?

8. AI Baumol Disease

9. Can AI speed up aging research?

10. The wisest people Jacob & Alexey know

11. The future of science in academia

12. How Open Philanthropy uses neglectedness to make funding trade-offs

13. Jacob’s advice and views on purpose

14. Jacob’s favorite joke

Introduction

ALEXEY GUZEY: Good to see you, Jacob.

JACOB TREFETHEN: Good to see you, Alexey.

TREFETHEN: How have you been?

GUZEY: I've been good, good. Traveling around. Good to visit San Francisco, as always.

TREFETHEN: How long are you here for?

GUZEY: For a few days, I think. Flying to the East Coast afterwards. One of the main reasons I'm here is, in fact, to talk to you.

TREFETHEN: I'm honored. Well, thank you for coming by. Thanks for coming to make it easy for me.

On Open Philanthropy's decision process

GUZEY: You run science funding at Open Philanthropy. You spend a lot of time thinking about global health R&D and more basic science.

TREFETHEN: And we have a process where program officers and I have to agree at least 50% of the time. So, in fact, we end up agreeing quite a bit more. Probably more like 75, 80% of the time, currently.

GUZEY: That's kind of a brilliant way to structure this, actually.

TREFETHEN: That's– yeah. I think it might get the best of both worlds, because we are trying to take bets with these grants that have asymmetric upside, hopefully.

Hits-based giving is a framing that we'll often use. We care about the 5, 10% percent of grants that are really, really epic at the end of the day. And it's totally fine if the rest of the grants that we fund don't end up leading so much. So within science, especially, where you can't really predict the future very well, that model, I think, is going to outperform from a social impact point of view something that is more structured.

We have a highly autonomous system. You could argue that it's always about the margins of how far you go. NIH or NSF or Wellcome Trust – or pick your other funder – will often have peer review that does maybe give you more signal. We steal the program manager-type role from offers a little bit – of you have one person who has a budget and they have a lot of autonomy within it – to get the benefits of "you can't get vetoed". POs [Program Officers] cannot veto each other's grants with certain very limited asterisks. And that's wonderful because you don't have to persuade people.

GUZEY: Yeah.

TREFETHEN: You know, of course, PO performance gets assessed over time. So, at a yearly level, you can't disagree more than 50%, and then at a longer term level, if you're making lots of grants that look way above the bar set by global health charities, then Open Philanthropy should allocate more money to that program. It's a slightly more complicated long run system than I'm painting. But anyway, we think it gets the best of both worlds. Overall, my opinion is that science funding does involve too much… a little bit too much peer review, probably. And we're in one direction of like really little peer review. Program officers can decide "I would like to get external peer review" on something. And they often do because they cannot be experts in everything they fund.

GUZEY: This is interesting to me because I've been thinking about this a lot. And I also had experience like running a science organization, funding science projects, right? And I used to talk about how terrible peer review is beforehand. And then in retrospect, the first thing that I ended up instituting was peer review because I was so scared of making decisions. And it's like, well, I'm not a specialist, and then I had always tried to make sure that everyone agrees on what we're doing. In retrospect, something like 50% is a great idea.

I also immediately start to think about like, okay, what would be the dynamics of the system where, like so you say we have a program officer and they don't have to commission peer review, but they can, and they want to, and – I'm like, okay, if you just have the system, then over time, probably what's going to happen is that it's going to tend towards like commissioning peer review becoming the norm. And then not asking for peer review will be kind of weird, even if technically you're not like you don't need to. And just because like, again, people are pressured to minimize risks. And then you can always blame others if you commission peer review versus not. But they also like, again, this gets me back to this 50% benchmark. If you notice that you have like 80%, 90% agreement, then this is a signal to you that maybe you're doing too much peer review, which is like a signal that would be absent if you didn't have this benchmark for how much we should be agreeing.

TREFETHEN: That's interesting, because that's actually not the implication that I take.

GUZEY: [laughs]

TREFETHEN: I think I would correlate my opinion more with program officers at Open Philanthropy than peer review scores from external experts, would be my guess. Which means that the implication... the implication would not be right. Another downside of having such an autonomous system with – at the moment, the only real checkers being me, and in some cases, Emily Olsen and Alexander Berger at Open Philanthropy – is that you end up fitting too much to "am I [points at self] going to give the stamp of 'I agree' or not?" So you should do less peer review, you should spend more time with Jacob. So I worry about that. I don't worry about the peer review one.

We also hire people who are quite… who are not shrinking violets. We're hiring people who want to change the world. They're going out there and they believe in stuff.

Are there other dynamic parts of the system you think could go wrong and how I've described it, though?

GUZEY: People trying to please their boss is always a huge issue. But yeah, there is probably not that much you can do.

Is it possible to make expected value calculations for basic scientific research?

I think for me, like, my immediate thought here is it seems really hard, if not impossible, to do any kind of expected value calculations if you're funding science. And so you mentioned that – I think I'm just really confused about how this actually works.

TREFETHEN: Yeah.

GUZEY: Because like, I myself don't trust the calculations for things like deworming or like malaria nets, even. They seem like... Very, very fishy to me. But then if we're thinking science and basic science and —

TREFETHEN: It's interesting you chose those two examples, where people have looked at them more. I wonder if that's an availability reason you chose those two.

GUZEY: I mean, I saw the tables that GiveWell was providing for deworming impact estimates. And there is like, a hundred effects. And then, these are only short-term effects. And then, like, for malaria as well.

TREFETHEN: Deworming looks at long-term, but yeah, I agree with the point you're making. I'm just wondering if there's, the point you're making is downstream of it being very difficult to know anything about the world.

GUZEY: I mean, yeah, yeah, basically it is, but the point I'm making is that even for things that seem much simpler than assessing expected value of basic science, I'm still pretty skeptical about them.

TREFETHEN: Got it, got it. I'm with you.

GUZEY: And then we get to basic science, and to me, it's like, well, I mean, this has got to be impossible to do here. And yet somehow…

TREFETHEN: Yeah.

GUZEY: Like, how does that work?

TREFETHEN: So, it gets way easier the closer things get to application to put rough quantifications on them. So, we're funding an efficacy trial of – we were just talking about TB vaccines. We're funding an efficacy trial of a TB vaccine in adults, and we can really constrain a bit better the probabilities of "do we think that it will detect efficacy in a given range, given the design of the study, given previous trials in TB?". We have a little bit of a base rate to go off of. If so, do we think that it would become a GAVI vaccine? So, it would be GAVI-supported in lower-income countries? Do we think the Indian government would—

GUZEY: But this is also, like, very applied, like, so...

TREFETHEN: Totally, I'm going to look around.

GUZEY: One of your grantees just got Nobel, right, in Chemistry: David Baker. And you were very proud of starting to fund him back in his lab, back in 2017.

TREFETHEN: Not me personally, my colleagues.

GUZEY: Your esteemed scientists – biochemists – Chris Somerville and Heather Youngs, who I'm going to ask you about afterwards as well.

TREFETHEN: Oh, okay.

GUZEY: Also, sorry if I'm pushing too much on this, but, like, is there an Excel table somewhere in the annals of Open Philanthropy where Chris and Heather, like, calculated the expected value of giving a grant to David Baker and it came out to be larger than distributing malaria nets in Africa? Or was it just – he seems like a really brilliant person working on something very exciting, we should support him?

TREFETHEN: We almost always... So, nowadays, well, I'm trying to figure out the way to answer this. We almost always will have a spreadsheet of some form.

GUZEY: [laughs]

TREFETHEN: Now, what weight we put on that spreadsheet, versus qualitative factors of other forms?

GUZEY: Yeah, because I remember, Holden has this blog post that I keep returning to a lot where he talks about the expected value calculations and their weight in decision-making of when you're doing hits-based giving, right? And the inherent limitations in any kind of expected value calculations.

TREFETHEN: The calculations are always an input. They're always a tool that we try and use. The benefits they hold... If you can make a believable one, is that you can try and set consistently budgets across cause areas that are very different.

So, for example, Open Philanthropy does not do all basic science. You know, we could give all the money away to basic science, but nowhere near that.

GUZEY: Yeah.

TREFETHEN: And to do that, we would have to be persuaded of some quantitative reason to do it. I think it's actually quite hard to make the case that Open Philanthropy should do... I think we might do about 700 million of grant making this year. We should do 700 million in basic science, given the cases you can make for the, you know, global aid policy program we have. The cases you can make for lead reduction that's nothing to do with vaccines. You know, the cases you can make for factory farming is something that is a sort of current really messed up part of our society that you need to get rid of. So, you need some way to do a comparison between causes. The more quantitative you can make it, in my opinion, is great if the numbers reflect anything in reality. I think the further that you go back towards basic science with no intended application, the harder that gets.

You can rely on some things, though. You know, like, we exist within a context. And that context includes the existence of the NIH. That context includes the existence of HHMI. That context includes... The existence of, you know, the Allen Institute. So, it's not like that if Open Philanthropy doesn't exist, basic science won't exist. We're playing at a margin. And we can try and look at what are other actors doing and can we get any sense of how valuable that is in comparison. And I think you can make educated guesses about what a good basic science grant looks like from working backwards from potential applications. And this is a... You know, you've got to be careful. Because you never know what the true theory of change will end up being. But I think it's usually productive to have a theory of change. And, you know, if you look at —

GUZEY: Right. Like, what you're saying is that you don't know how exactly the thing you're funding will end up being useful. But if you literally can't think of a way that it will be useful, then this fact is informative compared to whether you thought there's five different ways something might be useful.

TREFETHEN: Yep, absolutely. And, you know, the reason we wouldn't fund something that's really stand out, that might be cross-cutting, is sometimes simply located in, well, would this get funded by the NIH?

And, you know, you mentioned David Baker and Institute of Protein Design. They had... They... We are not the cause of David Baker being David Baker. You know, David Baker's been supported by– He's a rock star being supported by many people, by many funders, I mean. The grant that we made at that time came from... So this is maybe apocryphal, because I was not in the room. I'm going off of secondhand. But, you know, Chris and Heather seeing Baker present these ideas, assuming, since it's so cool, that he would get funded by conventional federal actors, coming back around later, he still haven't got the money. And it's like, what the heck? And engaging in more detail of, okay, what would it look like if we made a pretty large grant to the Institute for Protein Design? And I believe in that case, again, you should ask David Baker and Neil King as well – get them on the show – but my colleagues, Chris and Heather, were pitched a certain thing and said, "that's a great thing." It was to do with flu vaccines and vaccinology, and sort of applying frontier techniques of protein design to that. But then asked, "okay, wonderful, David, by the way, like, what do you really want to do? That think you can't get funded?" And I believe David Baker was then like, "well, actually, really, I want to do this methods development. So that I can sort of work on protein design at a fundamental –" So deep learning wasn't quite the paradigm back then, probably. But, you know, various machine learning techniques to work on protein design. But, you know, he was – He didn't fit the box of – His background did not fit the box of NIH would give him $10 million or whatever. that... And then Chris and Heather were like, amazing, do both. Like, here's $6 million for that, $6 million for this. And, you know, that's not something that we do with every grantee, for sure. That's not our... But it's something that philanthropy can do. And the reason that looks so good, even quantitatively, relates to the fact that... That he couldn't get funded elsewhere.

And I get more nervous when we're just like, okay, look at this really hot shot person, give them loads of money, – this doesn't always counterfactually look that hot to me. Because the hot shot person often has access to lots of money. So you got to have a more specific story about, their background doesn't quite match what they want to do. Or, you know, their funders have not come to...

GUZEY: If someone wants to get science funding from Open Philanthropy, it's easy for me to go and try to get funding from a bunch of other different places and get turned down by all of them.

TREFETHEN: I mean –

GUZEY: Because otherwise Open Philanthropy will ask you, "well, why are you coming to us?"

Non-profit funding vs for-profit funding

TREFETHEN: It's kind of a horrible aspect of nonprofit funding, that–

GUZEY: Yeah.

TREFETHEN: That it's the exact opposite of if you start a startup and you try and raise money. If anyone hears on the grapevine, if you're a VC and you hear "they're about to get funded," you're racing to try and get your money in.

GUZEY: [laughing] Yeah.

TREFETHEN: Because you're like, "oh my God, there's no financing risk. Incredible." Whereas nonprofit dollars are so scarce, and when you care about counterfactuals, you can't— We don't care if our name is on something. We care about the impact. So it's a huge win. If someone else can fund something, it gives us more firepower to fund the next great thing we want to. But it is way worse from a PI's point of view and from an executive of a nonprofit's point of view. It's pretty... It's really... The dynamics are kind of messed up in that way.

Why the NIH rejects grants

My colleague, Matt Clancy, who works on innovation policy, sort of made a grant and set up a collaboration with the UK government on meta-science grants, which they're now evaluating right now. And I think you can get benefits of having one pipeline, but multiple funders with different perspectives. And sure enough, with the NIH program, there's just a lot of stuff that... I mean, there are many familiar reasons why NIH grants get rejected that you or I or another funder might not agree with. And some of them are that a given scientist doesn't have enough experience in a given area. That's like a classic reason to get rejected. And I don't think that's a great reason in lots of cases. Another is that there's not initial data that shows that what they're trying to do will probably work, and it's basically too ambitious and too risky. Very common. And not the kind of thing that we would rule out a grant for. And in fact, some of our best grants look like that.

So if you can use existing grant pipelines and look at the work that people have put in to write those proposals, I think that can be a great thing to do. It can't get you all of the best grants. Because sometimes you have to... You know, scientists often are trained by the system to break up problems into NIH-fundable units. And if you talk to a professor and say, you actually don't have to work within that... And you really try and think together through, like, "what could it look like if you worked on this?" You can end up with quite a different grant proposal. So it's not a be-all, end-all. Just use the existing pipelines. But it's a pretty good start. And there's stuff... You know, if I was a new science funder sort of trying to get my feet wet, I think it's a great way to start. Because you see, well, before I try and revolutionize the system, what is the system currently doing? I need to sort of know that before I actually know what I'm doing. So you can get a more grounded sense of your disagreements with the system by doing it.

Jacob's "PhD" working side-by-side with two biochemists for 5 years

GUZEY: So what did you learn from Chris and Heather?

TREFETHEN: I mean, it's... I don't know how to condense that into an answer. Because we've worked together for the length of a PhD for me. So I've had, you know, weekly meetings with them—

GUZEY: But I guess there is maybe a little bit more context to make it somewhat easier. This person, like, doesn't have a science background, right, and they are scientists with long careers in biochemistry. You ended up, through the years, you went from not really knowing anything about scientific grant-making and not having a background in science to overseeing a big science portfolio. I wonder if there are some key mental things that you learned. Or is it... Or is it less that, and more you just spent years reading grant applications and talking to them? That is just the accumulation of all of this is what enables you to do your work now? Or...

TREFETHEN: The accumulation of talking to Chris and Heather for many years, it does enable me to do my work now. That's also true outside of some scientific perspective and talking to other people at Open Philanthropy, which is a kind of hub for thinking about effective philanthropy. So I don't want to short shrift other people, too. But about scientific grant making, yes. I mean, we've just gone over a hundred plus grants that have been made I know really well. And that's... I haven't made them. And I haven't been qualified to make most of those. But I've discussed them. I've tried to come up with objections to why we shouldn't do this. Try and do that all the time to be generative or to... Some... Occasionally it will be a useful input into changing direction of something or expanding a budget.

I think the sort of high-level things— I'm just reflecting now on what high-level things I've learned from Chris and Heather. I mean, one thing that's really... They have such a unique relationship with each other. And they disagree a lot. And that is very valuable to watch as someone coming from the outside, that you can have quite different perspectives, even with very similar shared information state. And that cuts against the sort of spreadsheet approach to things. You really can be right there in the trenches with each other and still disagree.

Now, I don't want to overstate that because we've tried to quantify the disagreement, of course. And so we've done exercises like, even if you can't say how good an expected value something would be, give an ordinal ranking of your grants. So in a given year, put your grants into buckets. Tier 1, Tier 2, Tier 3, Tier 4, Tier 5. And we've done exercises where I've done that for science grants. Chris has done that for science grants. Heather's done that for science grants. Katherine's done that for science grants. Ray's done that for science grants. And... Sorry, for their own grants in those cases. I'm going to recall the numbers wrong. But the ranking between Chris and Heather is more correlated. It's fairly correlated. It's not random. They kind of agree at the end of the day on quite a lot. The standard deviation of tier is something like 0.6. I have to look back. I disagree a bit more, interestingly. But overall, don't disagree that much. And, you know, what is that sample? Well, that sample excludes anything that we didn't make a grant to. So there might be more disagreement located in the things that never get a tier. That's very possible.

But, yeah, it's great to see... I've seen a deliberative, reflective process play out every week in an hour and a half or hour-long meeting with agreement and convergence and diversion. That's probably been... I mean, that's been very fun, for one thing. And we – I mean, Chris and Heather, but other colleagues too – cover such a broad area. That's another really interesting part of what you can do as a philanthropist. I mean, you could as other financing entities. You could as a VC as well. But, you know, their background is not in biomedicine. They're biochemists. But that background means that they have a kind of generalism across lots of forms of basic research that would end up being relevant to vaccines, drug diagnostics, you know, vector control tools. But also, you know, medical papers, also structural biology, also this, that. The background is enough to get you in the door to then read all those papers. But... the generalism is quite deep in a sense. You know, there's a huge amount out there. And you can't be an expert in all of it. So if you are trying to be opportunistic, do the highest social value things, and, you know, jump at this gap here, this gap here, you almost necessarily have to be a scientific generalist. Yeah. I don't know if I answered your question.

GUZEY: [thinks]

TREFETHEN: Well, by the way, if you're going to pause, I'll say more. [Chuckles] That's the danger.

Today's system of federal funding are not supporting generalists

The systems of federal funding are really not set up for generalism. Usually – I mean, there's some cases you can pick – but, you know, peer review from your peers in your field is the absolute opposite of generalism. That's notable, when you're able to do organizational design from the ground up, that that is where you end up going is, I think, partly in reaction to the current system.

GUZEY: Yeah... I have this crackpot, semi-crackpot thesis for a nonprofit or for a science grant-maker where the thesis of the org is zero peer review. Just take one person and they just decide whether to fund, like, science projects or not to fund. And that's it. And that's... And they're not allowed to seek external peer review. [laughs] It's a countermeasure. And they're probably going to end up funding a bunch of crackpots and a lot of things that don't make sense, but—

TREFETHEN: I'm generally— one of the sort of, I don't know, almost like meta-level organizational design principles I'm quite a big fan of is – and I think that is inherited from Alexander, the CEO of Open Philanthropy, who talked a lot about this – is flexibility. So the reason I don't like that is why ban someone? Don't ban someone. The reason to ban someone is if you need to jostle them out of the system they're otherwise going to get drawn into. But... Yeah. Just give people freedom. Give a peer freedom to... If they want to get external peer review, they can. And, you know, their manager might get mad at them if they make fewer grants because they're doing too much...

GUZEY: Well, I mean, the reason I'm thinking about that is because, again, I think I'm just thinking about, like, what happened to me. I never intended to be... Doing grants by committee. And that's what I somehow, like, found myself doing because I was just, like, too scared to...

TREFETHEN: Did you have anyone in your life, a board member, who could say, Alexey, "why are you doing this, why are you doing this?" Because I think that's a nice role. You know, I have weekly one-on-ones. That's a nice role. If you have someone in your life saying, "why are you doing this," then it can give you permission. That might be a cheaper way than banning. Why are you trying to ban yourself and stuff, Alexey? You just have, you know, a one-on-one every week. Yeah. Say, "I found that I'm doing this." And then in your one-on-one, whoever you're talking to will say, like, "why are you doing that?" And then you'll say, "because I'm scared." And then they'll be like, "then don't do that." Problem solved. We did it in a week.

GUZEY: Yeah. That has not been my experience.

TREFETHEN: Oh, really?

GUZEY: Yeah.

TREFETHEN: Fair enough.

GUZEY: Actually, I mean, this part probably should really not be included, but I think, like, when I'm thinking back on New Science and, like, large parts of what went wrong is that I didn't really have Heather and Chris figures who I could just go and talk to about what I'm doing. Like, I just didn't have anyone to go to and figure out and talk to about what I was doing.

How will AI affect the pace of scientific progress?

GUZEY: AI FOMO?

TREFETHEN: AI FOMO? I have the opposite. I think, if I know what you mean... I feel like every time someone tries to draw me into AI, I'm like, ooh.

GUZEY: Okay. So you have the opposite of AI FOMO?

TREFETHEN: I'm sure that's been true or false in different ways at different times. So I don't need to claim that's always been true. But...

GUZEY: But at the moment?

TREFETHEN: Yeah.

GUZEY: I guess it's, like, especially in San Francisco, I feel like whenever I come to San Francisco, all of the people I know here, well, except for you, perhaps, are doing something related to AI. You know, there is a couple miles from here, there is Anthropic, another few miles away from here, there is OpenAI and – you know, the Intelligence Age is almost upon us – all of these things, and so you're not... not drawn at all to, like, go and become a technical manager at Anthropic or at OpenAI or something like that? So feel free to... We can skip this question.

TREFETHEN: No, no, no. It's all fair game.

GUZEY: [laughs]

TREFETHEN: I am not, as a descriptive fact, drawn by that.

GUZEY: You don't want to be in the room where AGI is developed?

TREFETHEN: I... Not in that framing. No. That sounds horrible. I mean, that sounds... I think that urge that people have... I think that the urge that people have is often trying to meet a personal emotional need rather than a societal need. And I would love if people could meet that personal need through things that don't affect other people's lives quite so extravagantly. So I would see that – if I were following that urge, I would see that as maybe not the best way to make career decisions.

But, you know, I do think there's a lot of incredibly intellectually interesting things happening in all forms of alignment research and also developing new models and also biological foundation models and all this. There's many different frontiers that are being pushed incredibly rapidly. And that is something that I do think...

Well, you know, I haven't left San Francisco in a few months now. But the – I think San Francisco and the rest of the world, they're really in different head spaces, really very different head spaces about what the next five years are going to look like. And I'm sure that my social circle – I hang out with a lot of friends who work in AI or who have strong views one way or the other on different political problems with AI. So I'm sure I'm influenced by all of that in the way that I think about problems. And, you know, I do think in the case of medical progress, I've been thinking more recently about trying to...

Should we – you won't like this – I'm just going to go back to talking about my current job instead of "should I leave and do whatever you want me to do?" But how should we adapt our portfolio now in the context of potential changes to medical research that are driven by AI progress, and in the context of whenever we make a grant, the impact that actually happens from the grant is many years down the line. So you actually should be thinking ahead rather than thinking the most important thing this year is [gestures] this, so let's do this. You know, you actually have to backchain from years in the future. And I think that I probably do have different views about that that are informed by being in San Francisco than if I – you know, if we were a biomedical research grant-making organization based in...

GUZEY: How does this affect your grant-making?

TREFETHEN: So far, not much. And so I wonder, is that a mistake? But the thing that I often turn to is I think you can make educated guesses about...

Let me phrase it differently. Let's say that AI progress in a broad sense of language models, but also foundational biology models – but also this, that, and the other – continue to make progress at a clip. Biology and biomedicine as a science does indeed drive medical progress. So do you assume that medical progress will move forward at the same clip? And if not, why not?

I think one way to break down the question is, which parts of the system of medical research will become bottlenecks in a world that AI continues to progress really fast? I think you can make educated guesses, certainly at the five year level about that and maybe the 10-year level about that. I can give you my guesses, but I think basically the inference is that those are, if you care about the societal impact of whatever you're trying to make in biomedicine, you should relatively speaking focus more on those bottlenecks potentially, rather than just diving into the hot AI stuff that everyone's sort of already pushing ahead really fast.

GUZEY: Right... So, what are those things?

TREFETHEN: Well, I'm of course not sure. I don't know. But my educated guesses are that—

GUZEY: Because I've been, I mean – this is also for me, very interesting, right? Because I spent a lot of time thinking about biology, a lot of time thinking about AI, and I am – I'm not even sure... how exactly will the kind of science being done change?

Because in a way we have already – especially in biology – we went through this really dramatic transformation where people would spend their entire PhD sequencing a single gene, and now we can sequence an entire genome for a few hundred dollars. In a lot of ways, the biomedical research productivity has already gone, like, just thousands, millions of times over.

In a way everything changed about the way we do biology, but also nothing really changed, and I feel like I'm personally very confused about this and we'll just, like, see the same thing – everything will change in 10 years, but also, kind of, nothing will really change and...

TREFETHEN: I have the urge to get a bit more specific about what changed and what didn't because I don't think that all fields of biology changed at the same rate, and it depends what productivity means because I agree with you that over the course of one generation of a scientist's life – let's take it back to Chris and Heather.

Chris Somerville, my colleague, was one of the main guys behind Arabidopsis and, you know, the senior author on the main paper, this was a lot of his career of the model plant organism that gets used a lot. The first time they sequenced the genome of Arabidopsis, it costs – I think they had like a $100 million grant. So in a way, maybe it costs that. I don't know literally how you'd attribute the sequencing. I don't know how much you got, but, like – an absurd amount of money.

And now you can sequence Arabidopsis for, say, $100, probably less, probably way less, I don't know. But you're talking about real orders of magnitude. The question for me then on, and so it must be absolutely just such a trippy experience for him of, you know, it wasn't that long ago we did the first, like, that was 25 years ago, like, not long.

Similarly, with, more recently protein folding, you used to maybe spend a whole PhD trying to work on one protein, and now you can at least take sequence to structure and do really well very quickly. Yeah. Or you can't necessarily—

GUZEY: Yeah, like validate things in a few weeks, right?

TREFETHEN: Yeah.

GUZEY: Also, we went from, whatever, like, 200 weeks to 2 weeks, something like that.

TREFETHEN: Depends specifically on—

GUZEY: Okay, sorry, I don't know enough biology, but also more than an order of magnitude.

TREFETHEN: Right, I agree. We're not there on lots of things. You know, that's proteins, and that was built off the protein databank. Are we there for carbohydrates? Are we there for lipids? I haven't seen it. And are we there for protein function? Well, not yet, no. Function's harder to define than structure. Like, in some contexts, a protein acts this way, and in some contexts, it acts that way. Function's kind of what we care about. There's a long way still to go.

When we talk about productivity improvements, you can point to these things that have three orders of magnitude differences. The thing that I always try and backchain from in my seat now – I'm not saying this is the only reason to do science – but in my seat right now, that I care about science for health impact, for people, starting from the premise of everyone matters equally, and everyone's life matters equally. I care about productivity that's harder to benchmark. I'm like, "show me the new life-saving drugs at the end of a pipeline", and that's a really lagging indicator. It's a cruel one in a way, but I'm extremely confident that we don't have a 1000x improvement in approval of life-saving drugs.

You're looking more at the level, okay, what does someone actually do now? What does this structural biologist do now, versus what they did five years ago? Like, what are they actually doing day to day? And are they producing discoveries of similar magnitude, different magnitude? It gets a bit more complicated. I don't think that AI will sweep through as quickly.

If you talk to biologists about this, then the classic thing that I'm sure you've heard all the time is like, well, "Guess what? Alpha Folder is built on $10 billion of previous data collection in the protein data bank." So, you're looking around, you're like, "where do I have a really high quality data set that could get a similar kind of AI-driven, sudden improvement." And everyone's doing that now.

But then there's an even deeper question, which people in San Francisco talk about – which hasn't probably [laughs] made it very far beyond San Francisco yet – which is more around synthetic data or more around designing reasoning agents or scientists that you can then instantiate in a virtual university with a million of them, where they debate and publish papers and all that stuff. So that, um, I don't feel able to make good predictions about that world, because I don't think we've even got anywhere near that, but I think in the next few years, probably we could well do.

I feel like we can, you know, like here's one thing that I think will be a block in five years, uh, good targets. Like, we've got really good at designing something that binds to a target, but everyone's going off to the same targets in cancer. There aren't that many known validated good targets, so you can chuck as much money as you want, but if you don't understand how the system works... Similarly, take a brain disease – Alzheimer's – we have such poor measurement of what's going on in the brain that, do I think we're really going to get good targets in the next five or 10 years, even with magical AI from the sky? Well, no, because no one's going to open up their brain and be able to take the relevant measurements probably.

I can tell the story the other way, if you want me to, but I've rambled for a while... that's – [smiles] ah, I'll do it.

The story the other way is that, "oh, you don't need high quality data." It turns out that if you have a genius from the sky, you can actually learn so much from YouTube videos and learn about how physics works. You can have perfect laws of physics. And then when you have perfect laws of physics and you've just only read PubMed or only ingested such and such dataset of human biology, you can figure out what's probably going on up there, even if you never take any measurements and therefore you can make a perfect Alzheimer's drug. I can't refute that. It's possible that you can have wonderful simulations with our current level of data and you don't need to collect much more. That's not where I would currently place my bets. So I don't think we will have a great cure for Alzheimer's, even with magical AI before it goes through other real world data collection. But I could, I mean – what do you think about that?

AI Baumol Disease

GUZEY: I think I agree with you pretty much on all of this. I think one... I guess I think about Baumol's disease and indeed, AI will make sub-problems of problems we care about dramatically easier to do. Like, we care about [gestures] human health and for that we need, say, drugs. And part of getting a drug that works is predicting protein structure at some point to help with – I don't know enough details – but it's like part of a bigger problem. And there is like 20 sub-problems, and AI, maybe for five of them, AI will just make them free, essentially, but then you're still left with 15 problems, which probably cannot really be sped up, and that end up just occupying the rest of the time. So I do wonder whether there is some kind of an upper bound on how much AI can speed things up, because I know even at the end of the diseases just progress at a certain rate.

TREFETHEN: Totally, totally.

GUZEY: And there is this rationalist and empiricist frame, and rationalists say, "yeah, we can just think very hard, predict," and I think I'm just like, especially for complicated systems, now, you need to actually go and do the experiment. Until you run the experiment, you just genuinely don't know.

TREFETHEN: Yeah.

GUZEY: This makes me very sad, because it's...

TREFETHEN: It doesn't make you happy? We have so much to learn.

GUZEY: Uh, it makes it—

TREFETHEN: We're alive way before the beginning of biology.

GUZEY: Yeah, yeah. I mean, this makes me worried. I think I used to...

I kind of bounced around from thinking that – I mean, the original thesis behind New Science, in a lot of ways, was "AGI is not near." And everything will take decades. We need to really figure out how to make sure that science works well, that basic institutions of science are well functioning for the decades ahead.

Then I, like, went through a phase where like, oh wait, yeah it [AI] will change everything in the next few years, so none of this basic stuff of science and institution stuff makes sense, and now I'm back to this point of "oh, shit the world is going really slow." And like, yeah, AI will be strong, but like, it won't change the world. Again, every – in some ways, everything will change for some problems for which AI is really good at, and for other things, nothing will change.

This makes me very... I mean, this is why I'm thinking about institutions of science again, right? [laughs] because I'm like, "oh, wait, we need to plan for decades ahead and make sure that science works well decades ahead" and that all these, like, Alzheimer's – indeed my impressions are that we just we don't really understand, like, biology. We don't really know how to measure things. The brain is still – it's just like I'm... I feel mortality very acutely now, again.

TREFETHEN: What do you mean?

GUZEY: I mean, literally, you know, if I don't expect singularity or anything like that happening anytime soon, and if I really think that at the end of the day, you need to run clinical trials for things you really care about and they take years, then you don't really have that many shots on the goal. We have 50 years, 60 years? How much can you do in 50, 60 years?

[laughs]

TREFETHEN: Well, I mean, the...

[stands up and grabs Her-2:

The Making of Herceptin, a Revolutionary Treatment for Breast Cancer]

I haven't read this book, but I think it's gonna cheer you up.

[tosses the book to Alexey]

Here's my gift to you. I just bought it for me, but you can have it.

So I think, oh, no... am I thinking of the right, is that Imatinib? Which is that?

Yeah, I don't think that [Imatinib] had a phase three trial. I think it was so good in the phase two – but someone will correct me in the comments – so it was so good in the phase two, that they just approved it. It would be unethical not to.

And so the happy world for you would be, we have such magical AI systems that an AI system predicts, "by the way, if you run this in a clinical trial, then there's a 84% chance it will hit endpoint XYZ in 10 years, and we're highly confident it's not gonna cause neuroinflammation and it's not gonna harm your brain for XYZ reasons." And then maybe you'll be happy taking it to slow down progression of Alzheimer's. Now, there's a whole other set of questions of, will the policy world adapt to that and in what ways? I think in the case of prevention, you probably will be slower. In the case of treating people who will die of a disease, I think the policy response may be quite quick of "okay, this AI system spat out a prediction that it's probably gonna work. Do you wanna be in the trial where you might be in the control or just take it?" I think people would be like, "okay."

Anyway, there's my optimistic case. That said, I'm actually more in your...

GUZEY: [laughs]

TREFETHEN: What do I currently believe? Yeah, I think some things still take – you know, Alzheimer's, if you're taking a daily drug to prevent progression of it, which no one is basically currently, but if we had a future drug – that totally could have safety problems, you know? So do you want to take it before you've seen, how someone in a clinical trial did after five years, after 10 years? I'm like, not sure. It's sort of pretty slow moving disease. So we might just have to wait for that trial. And it's, that takes time. I think the thing I, if I would cheer you up in a different direction, it's that, the—

GUZEY: [laughs] I like how this interview started from me asking you about your work at Open Philanthropy into you just cheering me up, so I would continue to pray for the future.

TREFETHEN: [laughs] Well, I, uh...

GUZEY: [laughing] Please continue, I appreciate it.

TREFETHEN: Of course, one of the, um—

GUZEY: [laughs] I hope you're doing this for free...

TREFETHEN: [smiles] Oh, I assume I'm getting paid for this interview. No, one of the, you know, I was saying earlier about how it's frustrating that there are so few good targets to go after and everyone goes after the same ones. Well, the analog, the disanalog of that in the case of products that help people in lower and middle income country settings, but not so much in high income country settings, is that there are plenty of probably good targets that haven't been explored that much. And, you know, if you're... it's way easier to study a bacteria in a lab and to learn a lot about it as well than to study a human system. You might be able to find quite a lot out about that bacteria, and have quite a few potential antigens that you could go after with a vaccine, for example.

If I may squeeze in Strep A into this interview, because any time I'm allowed to, I will. Strep A is a bacteria that you've probably been infected with many times. I have been. It causes strep throat. Do you remember the last time you got strep throat?

GUZEY: I used to get strep throat a lot.

TREFETHEN: It's very annoying, so mostly it's annoying but sometimes it—

GUZEY: I actually was in the hospital twice when I was younger for it.

TREFETHEN: Totally, yeah, it can – it's real... kids end up in hospitals somewhat frequently for it, and can get an invasive disease where the strep bacteria spreads around your body, and it can cause lots of problems, and it can cause many deaths, actually, in toddlers.

In addition, in lots of countries, if you get infected many times, there's an autoimmunity, there's a molecular mimicry issue. Your body's trying to attack the pathogen, but actually ends up attacking your heart valves. So you can get acute rheumatic fever or long-term rheumatic heart disease, which leads to early cardiac problems and death. And if you add up all the problems that strep A causes downstream, the best estimate – which isn't a great one – the best estimate that we've got is that it leads to about half a million people dying every year.

GUZEY: Oh.

TREFETHEN: And you know, when we loop back around to can you even make any comparison between basic science of [gestures] this form or this form? And how will you know what's better than the other? Well, you can look at neglectedness as a pretty good lens and very few people are working on strep A, very, very few.

If you look at the total amount of R&D funding dedicated to trying to make a vaccine, for example, or include if you want some more microbiology and basic stuff, and epidemiology of where's rheumatic heart disease happening? The best estimate I've managed to pull together is that there's maybe $1 or $2 of R&D funding per year of life lost due to infections globally, okay? So you shouldn't have an intuition for what that means, but you're gonna get a comparative one in a second, which is that for cancers, it's probably about two orders of magnitude higher. So more like $200.

So that doesn't tell you anything immediately because maybe it's impossible to work on strep A, and so there shouldn't be any money spent on it, but I'll give you a hint. It's not impossible to work on strep A! There's quite a lot you could do. And if there were more scientists really laser focused on the problem, you know, there's loads of unanswered questions at the level of what makes a good antigen. There are maybe seven leading hypotheses.

GUZEY: This is so weird.

TREFETHEN: [nodding] It's very weird, very weird.

GUZEY: How come?

TREFETHEN: There's a couple drivers. I mean, the deaths are mostly happening in lower middle income countries. So not exclusively though, invasive disease can get you anywhere.

GUZEY: Yeah.

TREFETHEN: That's a big driver. So another reason is the FDA banned trials for 30, 40 years, and that ban got lifted in 2006, I think. So that sort of meant that it was hard to build a career in the field or lots of vaccine companies don't have programs where they've looked at.

Then in addition, because of the auto-immunity issue, you want to be sure that you're not creating problems with a vaccine. With modern tools, I think you can be pretty sure you're not, but there's a question of what data package will I have to get to prove that there are no safety issues?

Then there's a question for vaccine developers of efficacy. It's pretty easy to run a trial against pharyngitis strep throat, because people get that all the time. You and I are at risk of getting that in the next year, so we could enroll in a trial, let alone kids who get it all the time. So you only need maybe a thousand people to run a good trial with that, but if you care about preventing rheumatic heart disease, if you care about preventing necrotizing fasciitis—

GUZEY: Yeah, that takes so long.

TREFETHEN: Proving the link is, you know, in a trial, very difficult because the incidence is too low. You won't accrue enough cases of necrotizing fasciitis where it's in your legs and eating your flesh and every hour it moves another inch kind of thing because it's rare enough.

Can AI speed up aging research?

GUZEY: So I should be pessimistic about the future again?

TREFETHEN: No, you should be optimistic because- well, sorry, sorry, you should be pessimistic unless you want to get involved in the fight against strep A.

GUZEY: [laughs]

TREFETHEN: I think if you and your viewers got involved, we could really take a swing at it because you could test a bunch of vaccines for sure, you definitely could.

GUZEY: Unfortunately most of our viewers work on things like biosecurity and monitoring, things like that. Instead of taking real problems.

TREFETHEN: Oohhhfff, you're insulting your viewers? Wowww, wow. That's interesting. Interesting. Those are real prob—

GUZEY: No pandering to viewers.

TREFETHEN: [laughs] Any other swings you want to get in at them? What else are they working on?

GUZEY: [laughs] All kind of stupid shit.

TREFETHEN: I mean, basically what's the limitation on this problem? There's scientific limitations, human capital limitations and money limitations.

GUZEY: Yeah. No, but it is striking to me – but like the thing that I keep returning to is these things that I do not see a way of speeding up. Like at the end of the day, will we be able to use AI to predict that, or shorten the length of clinical trials from 10 years or 20 years to down to a year?

This makes me think that maybe we'll solve, in some basic way, aging, but we'll solve it before all of these diseases. Like we'll understand how the body falls apart, but not...

TREFETHEN: So, I don't think we will solve aging before we solve strep A. I think that...

GUZEY: Also, sorry, sorry, but since I asked you about this, an immediate follow up is – since you mentioned that health impact is what you really care about at Open Philanthropy—

TREFETHEN: Yes.

GUZEY: Are you guys then trying to solve aging?

TREFETHEN: We have given grants many years ago to Irina Conboy and Steve Horvath. We've tilted out of that grant-making for neglectedness reasons. There are now lots of companies popping up funded by wealthy people trying to work on aging, so the neglectedness of the field looks way less.

In fact, a lot of the – you know, what do we have? Money. What can we do? Give people money. That's sort of our job. A lot of the money flowing into the field is now bottlenecked on people, that it's just pumping out salaries as far as I can tell. So for aging right now, we've pulled back. But we've done some in the past, yeah. It's interesting that as a sector [aging] is so company focused relative to other sectors, nowadays, but I think that's for various sociological reasons.

What is aging? Infectious disease is a part of aging. Inflammation is a part of aging. Heart disease kills however many people, and if you could solve heart disease, you'd solve aging in some way. There's these pendulums that swing in discourse of like, broader society is often focused on specific diseases, and people rightly, 5 or 7 years ago, were like "hold on, in the case of aging, are there cross-cutting functions and cross-cutting drivers that if you could solve those would help with all sorts of diseases?". I think that's a really valuable lens shift. Sometimes, I wonder if we've gone too far in that direction now, of like, well actually you could cure cancer and you could cure heart diseases. That'd be pretty cool. So, I'm not sure at the margin, where I'd go there.

I expect that the stuff we're funding because it's so much more neglected, if we weren't giving support to those scientists, then they're not getting money elsewhere. So, that makes the case much easier from an impact point of view.

GUZEY: Right.

TREFETHEN: You don't want to make a strep A vaccine together? I think we could bash that one out. I think we could do that, I mean, not you and me —

GUZEY: [feigning disappointment] oh

TREFETHEN: But we could help scientists who are doing that.

GUZEY: I thought we'd... [looks down]

TREFETHEN: Okay fiiiiine, you and me.

GUZEY: [laughs]

TREFETHEN: You twist my arm. We can work on strep A. You know, I didn't want to work on strep A until this interview, but if you're going to keep pushing me like this.

GUZEY: I mean you keep bringing it up. Every time you talk to someone, you keep talking strep A, you know—

TREFETHEN: I'm a real freak like that.

GUZEY: Maybe it's time to put your money where your mouth is.

TREFETHEN: Yeah.

The wisest people Jacob & Alexey know

GUZEY: Who's the wisest person you know?

TREFETHEN: Oh, my gosh. Who I know?

I don't know. Maybe – can I give boring answers? Probably, like, my friend Sarah back at home. And my mom. I don't know.

GUZEY: [laughing] Wonderful. Tell us about Sarah. That’s such a fun answer.

TREFETHEN: I haven't discussed before this interview if she would be...

I don't know. I just feel like when you, what is wisdom? I feel like I'm often seeking a – what would someone with good character do in situations that I'm in? And sometimes I think of Sarah.

But what is wisdom? Let's see. There are many people who I either know or have a more parasocial relationship with that I view them as bringing out the best in me, but I don't know if that's always wisdom. Well, which lane do you want me to pick? People I don't know or who I do kno? Or, no, no, no, I'll blur the edges:

[gestures] Seemay Chou, Prachaya Vasti, who are making Arcadia, I find inspiring from up close and from afar. As you know, we sort of hang out in circles that talk about changing scientific institutions and that are trying to be more ambitious and different with how science is done. And they are really living it. They are living it. They've created a whole new institution that is iterating like crazy on doing experiments and how should you rethink publishing, and how should you, uh, rethink stopping investigating something? Like at what stage do you stop? Because of the opportunity costs, you know, you shouldn't investigate everything forever. When you do stop, what can you publish so other people can know? They’re actually living a lot of what I sort of think about. So that’s really cool.

[gestures] Heidi Williams, she is an economist who I think has this presence that brings everyone along with her, and she’s just so prosocial and public goods-focused.

[gestures] A friend, Nan Ransohoff who runs Stripe Climate. She actually has this sort of similar to Heidi description I just gave. She somehow makes the world actually move towards creating whole new technologies to remove carbon dioxide from the atmosphere. I'm sort of like, what? How?

I feel like it's dangerous to give out one's idols because then they might know. And then they might – I have to – okay, I'll hold back on some of them. Virginia Woolf though. She can know.

GUZEY: She's coming for you now.

TREFETHEN: Who do you think is the most wise?

GUZEY: Tyler Cowen.

TREFETHEN: Tell me more.

GUZEY: When I look back over the years, who are the people who I wish I listened more to? He is like at the top, just like consistently somehow tells me things that I think I should have thought about before. And then I ignore him. And then a few years later, I'm like, oh wait, Tyler told me about this a few years before.

TREFETHEN: It's a good way.

GUZEY: Yeah, I noticed that whenever I think about things that are very difficult, I ask myself “how would Tyler think about this?” more often than for any other person. He’s kind of a scary person. He's too wise.

TREFETHEN: [laughs] Good thing to aim for, I suppose.

GUZEY: So how will science be different in the 21st century?

TREFETHEN: Too big, too big.

GUZEY: Role of academia?

TREFETHEN: Role of academia?

GUZEY: You mentioned that for aging, for example, for sociological reasons, it's a problem that often for-profit companies are funded by people who really care about the problem directly affecting them, perhaps. And then right then I know, I guess I think a lot about context for this in my head. I can't stop thinking about this graph of NIH funding distribution. We're like, ages 35 and less like this, and just 65 more like this.

And I think at this point, it's like primary investigators ages 65 and older are getting funded at 7x the rate of those 35 and younger, I believe, something like that. And it keeps getting worse. So this makes me—

TREFETHEN: [nods] That is crazy.

GUZEY: Yeah, this is just like bonkers. I think a lot about what Tyler Cowen said, with Emergent Ventures, where you just fund really young, talented people. And Tyler Cowen once said that it seems that young people are getting accelerated actually, because the internet just allows people to learn much more, to get up to speed a lot more.

Tyler pointed out an example of chess where Magnus was pretty young when he became world champion and the person who Tyler expects to be the new world champion is like 17 or something crazy like that. And so the top performance keeps getting younger.

I was recently talking to a professor of physics at MIT and was asking him about the preparedness for research of undergrads and grad students these days. And he told me the same thing that because of the internet sometimes the best undergrads are better prepared today than like 10 or 20 years ago. So in my head, there is this thing about on one hand, it seems that naturally young people are getting accelerated and like they should be getting more funding and more support and like, should be getting up to speed faster. And like in software, obviously. Or, like even you [Jacob], you are now, whatever, like 30 or so. Maybe you look younger [laughs], but you're now running this big thing at Open Philanthropy, right? And on the other hand, it seems that academia is not reacting to that at all. In fact, it's like the professors are getting older and then there is this AI thing and I think it also pulls away the most talented.

For example, a bunch of the most talented grad students I know in biology who I would have expected them to go and become professors and they all – well, not all, but the most talented ones, a bunch of them – just started biotech companies immediately after graduating PhD, even though I would have been sure they're the kind of people who I think would like, for them, the perfect career path would be to become a professor around the lab.

And so I have all of these things in my head and I am famously a person who likes to criticize academia a lot, but I do wonder if there is going to be some kind of a trend where more and more research on the basic one will just be done by other sources, maybe it's going to be, you know, Tyler funds the 20-year-old and Open Philanthropy funds the person when they're 23 and then they're just now running their own lab doing whatever it is that they want to do, sidestepping the entire system. And then with the help of AI also like one of the things that I feel like AI will do is just amplify the top performers. Like in biology, if you don't need to... if the experiments are done by robots, intellectually intensive work is going to be more and more outsourced and individuals get more leverage as well. And so there is, I feel like this aspect. Anyway, sort of been rambling.

TREFETHEN: There's a lot of different parts of this picture, I think. Maybe a place to start is just me personally, and that's what I know best is my own life. And, you know, I view biology as one of the consolations of aging. There’s an unlimited nexus of different things you could know about cell biology, about different organs and diseases and all that, so that every year it feels better and better. It's so wonderful. The accumulation of knowledge. And I know so little, but next year I know a little more and then I know a little more. And that's sort of, you know, as I realize I will get older, that's been a nice field to go into because I think biology does actually sort of inherently, no, inherent isn't right, but it does reward accumulation of knowledge and connections rather than just raw “think my way through a problem” talent. So I think I dispute some of the premise in that I think some forms or some fields will have top performers of different ages. And then within any given field, you will get heterogeneity anyway. I totally agree with the punchline for what it's worth though, of even within biology, I don't think even the NIH wants the – who is the NIH? – But I think a lot of people at NIH probably feel quite awkward about how high the median age is of PI's getting grants.

And I mean, I view that as one of the big contributions of the work you've done, also of the ARC Institute, also of the meta-science community that's been building over the last few years as a kind of a broad group of people interested in trying to make science go better. Is this just highlighting this fact more and is the solution to that fact that you build more institutions that empower younger people sooner? Is the solution that you try and reform the NIH? Or is it somewhere in the middle? Like the Art Institute are put sort of in the middle of they're working with the existing universities in the area, Stanford and UCSF, and they work with Berkeley as well, I can't remember, but they're empowering certain people with much more resources and much sooner than you would. And that's great. So that's definitely a direction I agree with – sold, that should change. I don't know how it breaks down into what I would change.

Then... now I'm wondering what other parts of the question to pick up on. What will happen to academia broadly?

GUZEY: Okay. Proportion of revolutionary basic science done within like... Universities versus outside universities.

TREFETHEN: As I think through, well, what do I really believe about that question and what drives it? One thing that is coming to mind is how many of the discoveries of the future will come from big science or big collaborative efforts, like the Human Genome Project, like AlphaFold requiring so much of an institution to... be behind it and the compute and all of that, like telescopes. Pick your examples that are kind of big, big collaboration versus the discoveries that just won the Medicine Nobel Prize this year. I don't know that story well, and I'm curious if you do actually, but microRNAs – that was basically, you know, you can discover that in the 90s in a lab and that's kind of what they won it for. As I think through the big discoveries, the big discoveries... As I think through some discoveries from the last 30 years, a lot of the blockbuster ones do seem to have come from more of these big collaborations.

You could view the Baker Lab that way, even though it's one lab. You know, the Institute of Project Design has tens – or maybe a hundred now – of students and just a couple of professors, but a lot of those students end up getting faculty positions. They've been trained on their Rosetta tools. They end up using them in their new position, feedback in and build a tool. So how do you count that in terms of size?

I do think that purposefully set up organizations with, you know, more regular managerial structures than academia can outperform on some of those big problems. I mean, it's happening right now in fusion. Who's going to end up adding the most value, these new fusion companies or the big government funded projects or national labs? I don't know the answer to that. But I really... I wouldn't be surprised if, frankly, because of wealth accumulation due to tech as well, it was like there's some really rich people now. They can do that, you know, projects they want to do in a private company because we support property rights in America. So where does that wealth go is almost going to drive some of the answer rather than the more bottom up scientist led answers to the question, probably.

The future of science in academia

TREFETHEN: Do you have hope for academia? And if so, what kind of hope?

GUZEY: I have hope for undergrad part of universities. I think people, especially in Silicon Valley, underappreciate the value of a college degree funnily enough. But it's mostly because of the exposure. It's like a very nice track to be on where you can just like meet a lot of people, explore a lot of different things, talk to and meet a lot of professors, meet people your age, take any class you want. Socialize, have fun. But in terms of academia afterwards, I am... I mean I feel like I'm professionally obligated to say no in some way because I'm literally working on building new institutions of basic science.

But in a way there is... The way I think about academia now is that it's like... Professional science. I think academia has always been really good at professional science and paradigmatic science. And it's probably going to remain actually pretty good at this kind of thing, whereas I'm personally just more interested in pre-paradigmatic research and generalists doing things like... Uh, I know it seems that like, this is kind of a stupid example, but like Darwin, right? Like he spent whatever, a couple of decades writing his Origin of Species and probably would not do well at all in academia and I think all of the researchers that I'm most interested in somehow don't quite fit in. They can't get funding. They are just too weird for academia. They work on things that are too long term, too poorly defined—

TREFETHEN: Is that right? Don't you like lots of... I think of the people you've told me you love before, some of them are doing pretty well in academia. I don't know if you want me to out them on the podcast.

GUZEY: Well, we can, we can always cut things out.

TREFETHEN: Ed Boyden is doing pretty well.

GUZEY: That's the most ridiculous example you could have come up with. Because Ed Boyden, the only reason he's still in academia is because of you guys, because he was about to close down his fucking lab at MIT and then you came in, Open Philanthropy and gave him a huge grant and now he's doing well and he recovered and didn't NIH literally reject him like nine times in a row. Like I remember he said that in some of his interviews that his expansion microscopy, now one of the most successful projects coming out of his lab, was literally rejected nine times before it got funded. By who? [motions at Jacob]

TREFETHEN: [laughs]

GUZEY: Open Philanthropy. I think that's a great example.

TREFETHEN: It's interesting that that's your narrative on it. Not that I had anything to do with that decision. So when we made the original grant to him, I wasn't there, but of course I'm willing to take the praise if you want to give it. It would be super interesting to ask him what he thinks would have happened to his career in the absence of that grant.

So I think what would have happened to expansion microscopy is a different question. I mean, expansion microscopy is crazy. I cannot believe how big that's gone. Like jeez, I can't believe it worked and it's gone so big. It's like, what? So yeah, that's one of those where you're if you make one of those grants, you can just pack it up and retire kind of.

But I wonder, I mean, he's so talented that in other ways that wouldn’t he have found something great to do with that? I don't know. Anyway, you know him better than I have, I've met him once in my life, so I don't know.

GUZEY: No, I think that's actually a very good question to explore. Like what, what would he have done? I do think academia is doing... like, he’s restricted as a result.

I mean – I actually – I remember like I used to lobby Open Philanthropy—

TREFETHEN: You once sent me an email saying you need to give Ed Boyden a hundred million dollars. And I said, thank you so much. Or maybe I ghosted you. I probably ghosted you.

GUZEY: I think I was sending this email to Holden, to you, to anyone at Open Philanthropy I had an email for. And you guys never did.

TREFETHEN: Not yet, not yet.

GUZEY: Yeah. And I think it would have been better for the world if Ed Boyden had a hundred million grant to start his own thing instead of being in academia.

I think a lot about the fact that he still lives within that system. Where he has to like constantly churn out papers, constantly apply for grants. He's constrained by the fact that he has students who are with him for five years. I would really love to see if Ed Boyden just had like 20 years of funding for an institute. Like, what would have happened? And I suspect that something way more interesting would have happened.

How Open Philanthropy uses neglectedness to make funding trade-offs

TREFETHEN: When you talk about duration of funding and flexibility of funding, to a funder, what looms large are the trade-offs.

GUZEY: Yeah. Trade-offs, of course.

TREFETHEN: Unfortunately. Yeah.

GUZEY: I'm an Internet writer who just pesters people and sends them emails telling them to fund this and that. And you're the person who has constraints, who needs to make decisions, who faces trade-offs, the famous economics thing that sometimes features in our life.

TREFETHEN: Here's an example of a trade-off. If you give someone a four-year grant versus a 20-year grant, that's five times less money. Example: The UK biobank is so useful already, and we know it's going to get more useful over time as it accumulates more data. Does that mean that you should just endow it? Just give it as much money as an institution that can continue for decades, and don't faff around with all these grants that they have to write. And that's an argument that actually I find a bit more persuasive than pick your favorite scientists in 2024, give them 20 years of funding. Although I'm curious, how many scientists in the world would you think of the 20-year thing? Because, you know, for example, HHMI, how many investigators are there? I'm not sure, actually. But there's quite a few. You know, it's... It's a very limited set of science, but it's probably...

GUZEY: Three thousand?

TREFETHEN: Yeah. Well, hundreds of thousands, say. And that's... Okay. So that's already a lot. Would you halve it and double the period? Would you double it and halve the period? Or are you talking about there are five Ed Boydens in the world, and it ain't about the 500? Oh, okay. Interesting. So list them. Who’re your faves?

GUZEY: Do you have anyone except Ed Boyden? Michael Levin, perhaps. Tufts.

TREFETHEN: What would the annual size of your 20-year grant be?

GUZEY: For Boyden, yeah, probably between 50 and 100 million... a year.

TREFETHEN: [laughs]

GUZEY: [laughs]

TREFETHEN: That's beautiful. Thank you. That's really...

GUZEY: That's like 20% of your endowment, right? Something like that.

TREFETHEN: Yeah. But hey, it's a pitch. How would that get spent? So let's divide it out. How much is a postdoc's salary? How many people are in this lab?

GUZEY: Yeah, that seems...

TREFETHEN: I just think you get management constraints in labs. Like, you know, you...

GUZEY: No, that seems too much. How many people can he... Actually, that probably... No, that's...

Probably closer to the order of $10 million per year. I think if it were more than that it’s not just a Boyden thing.

[laughs]

We'll see if I'm brave enough to leave this in.

TREFETHEN: [laughs]

GUZEY: Hopefully no one who is considering giving money to my non-profit is watching.

TREFETHEN: If I were you, the productive line of inquiry, I would sort of advocate for myself is to just really scrutinize the trade-offs more. Because you can do a lot of damage with $1 million. You can really fund work that won't happen without it. And that work can be towards a goal that you think might...

In our case, we work on biomedical science. So towards a goal that you think might prevent tens of thousands of deaths a year, maybe hundreds of thousands of deaths. And the trade-offs get very acute when you're facing down the barrel of do we fund this marginal Malaria monoclonal antibody project?

We just funded Alassane Dicko in Mali to do a phase one to a trial that's basically never been done before, of a potentially transmission-blocking antibody that you could give maybe in combination with something that would have personal protection, too. And, oh my gosh, if that worked, just imagine. Right now, there's seasonal malaria chemoprevention that kids in many West African countries will get, which is taking preventive drugs that you swallow, you know, a few times, four months in a row. Kids hate the drugs. You find them in the trash. You know, like, not everyone finishes the regimen, all that. And they're not perfect. Even if you finish the regimen, imagine if you had one fucking — oh, sorry — one injection that protected you for a whole season, because you had a really good monoclonal and it involved a transmission-blocking component that reduced malaria spread to other people because there's fewer parasites in your body for every other reason. Then, oh my God, and guess what? That did not cost $10 million a year for 10 years. So the tradeoffs in science can be so intense because that's a great project. I'm very glad we funded it. And there's lots of other... That's not the only amazing project. There's many other ones, too. Some of them are in Mali, and some of them are in Thailand, and some of them, you know, they're not all at MIT. So, you know, if you're...

GUZEY: I think about this just... I'm just so much more interested in basic science and methods development. It just seems... Whenever I think about funding things, I always want to fund methods development, essentially. And so this is why I'm so excited about Boyden, for example.

TREFETHEN: Okay, let me sell you another one, that maybe we can meet in the middle on—

GUZEY: Again, I'm, like, I'm thinking about, like, the reason I'm, like...

TREFETHEN: So do you mean, like, tools as well? Like, microfluidic chips, do they get you going?

GUZEY: Maybe. I don't think I... After I said this $100 million a year thing, now I'm very worried about saying anything because that is very stupid.

TREFETHEN: No, that wasn't stupid. I was not trying to shame you for that being stupid.

GUZEY: I'm trying to think of, like, 50-year time horizon, something like that.

TREFETHEN: Yeah, I don't like the 50-year time horizon. I think the world changes too much. I'm into 10. You can get me at 10.

GUZEY: Yeah, 10, I'm just, like, at 10... Again, if we're thinking about really understanding how the brain works, really understanding how aging works, really understanding Alzheimer's, will we get there in 10 years?

TREFETHEN: I don't think we will, no.

GUZEY: Yeah, so why not think of 20 years? Because it seems like if we think about 10-year horizons, then we'll never find this Alzheimer's stuff.

TREFETHEN: But, you know, the trouble is, play it back. It's, well, let's say...

GUZEY: Are you saying that if it's not tractable within 10 years... Then it's just too early to be finding this problem, essentially?

TREFETHEN: No, no, no no. I don’t mean that. I don't mean that. I mean: if you're trying to attack Alzheimer's and you think it's a 50-year problem, the reason I would still try and think in 10-year chunks is that tools from other parts of biology, from AI, from whatever, will sweep in and hopefully be helpful in 10 years in a way that you can't predict now. So, if you just pick, I'm going to focus on Alzheimer's, I'm going to fund on Alzheimer's right now, but you don't yet have, you know, a clever probe that non-invasively detects XYZ that you end up getting in 10 years, then the work you did might not be that useful.

GUZEY: The way I think about this is if I think about 50 years for Alzheimer's, then this kind of... where the middle ground is, it suggests that pure Alzheimer's – pure biological Alzheimer's research – is just not the right thing to fund right now for the reasons that you outlined, and, indeed, they're like more basic methods development things that will in the next 10 years get developed, and then over the course of coming decades really affect how more object-level Alzheimer's research is done – is the right thing.

So, this is why I'm interested in methods. I think you're right to say – I think you are right. Your argument is entirely correct, but I think in my case the fundamental thing I care about is still that the time horizon is too long, but the fact that there is such a long time horizon means that we need to move up the stack and fund something that will contribute to it over the longer term.

TREFETHEN: Okay. Yeah, I think the higher you move up the stack, the harder it is to know, you might end up not funding science. Maybe you end up funding something completely different, depending how high you go up the stack.

GUZEY: Yeah, yeah, I think this is indeed, but I think a lot of things that are funded there, it's probably not going to be like theoretical physics, but it is going to be somewhere between, you know, applied physics and biometrics development...

TREFETHEN: I think, yeah, that’s cool. If you do end up–

GUZEY: Does this like – you kind of, like, said it's cool and then didn't comment on this but like, does this make sense? [laughs]

TREFETHEN: Well, you know, if I was trying to work on Alzheimer's, which we have done some of the thing that we really try and avoid is working on conventional theories or on conventional approaches to the problem, because NIA – the part of NIH that funds Alzheimer's – the budget, you should fact-check me on this, but my memory is, like five years ago, or maybe seven years ago, when we started doing the funding, the budget was maybe 800 million a year on Alzheimer's, and now it's gone up to like 3 billion or 2.5 billion. It's gone way up.

So they're going to fund a lot of stuff, which is great, but that means that we have to go really contrarian. Within Alzheimer's, we go really contrarian. And so we look for unusual lines of evidence, and one that has been very interesting to watch, and that I was wrong about relative to my colleagues, Chris and Heather, and others – so, I place so much weight on neglectedness, and that I'm like, there's so much money in Alzheimer's. There's so much money in Alzheimer's. How the hell can we do anything in Alzheimer's? And then Chris and Heather were like “yeah, but a lot of that money is absolute trash.” You know, like, who cares about that money? And I go, “no, but if it's there, then if you have a project idea, still, you've got a better chance of getting some of it, right?” You know, we have this kind of debate. And they ended up funding a very unusual line of evidence with a relatively early career scientist to take on the next step of what has been a 30-year project by a more senior scientist in Columbia, Francisco Lopera, who worked with an extended family in Columbia who have familial Alzheimer's. They tend to develop Alzheimer's in their 40s, and there are thousands of people in this family over many years sort of working with them.

Then there's this group in Boston, Yakeel Quiroz and Joe Arboleda-Velasquez, that looked in a huge amount of depth at one member of the family – or actually maybe two members of the family – who did not develop Alzheimer's in their 40s, but in fact went on to develop it in their 60s, so a long time afterwards. And you're like, well, what the hell happened there? This is a genetic form of Alzheimer's, so how the hell did they avoid it? This is the polar opposite of something like our TB vaccine grant, where there are thousands or tens of thousands of people, and you're taking a little bit of knowledge from each sample. This is like, there are two people, and you're going really in-depth and trying to learn as much as friggin' possible.

From those studies, there have come out potentially new targets – which would be wild if they end up being useful targets – of the Christchurch allele and the RELN gene variant that they've mapped really carefully from looking at these brains, like what actually it looks like might have happened biochemically, and okay, maybe it makes sense of this observation in this other part of the literature 15 years ago.

So now maybe RELN is a gene you can try and play with as a drug target, maybe the Christchurch allele is important. And that you would never in a million years get from a conventional medical school or a conventional type of exercise on amyloid beta or whatever. You just wouldn't ever get that. You have to go back and think what is this, how can I get something really weird to play into the story?

Now I forget the original question. I just got excited by that story.

[pause]

Oh, yeah, yeah, yeah. So how up the stack do you go? I don't know. I think you might want to do a mix. I'm not sure I'd go – there's still some low-hanging fruit left around, but there's so much money, I'd probably go up the stack. But we've done some down the stack as well.

We've funded a trial of a sleep drug, where, you know [gestures] if you sleep, that is good for you.

GUZEY: [laughs]

TREFETHEN: Oh, God. I actually didn't, as I started saying this...

GUZEY: I... I sleep.

TREFETHEN: I was not planning this. This was not an attack.

GUZEY: It's okay. I sleep a lot these days.

TREFETHEN: That's great. And if you can sleep half an hour more, which on suvorexant, this drug, you probably can, then what happens if you randomize people, give some people this drug? They sleep for half an hour more. Will that affect their progression of Alzheimer's?

GUZEY: Wow. Did you fund this?

TREFETHEN: We did fund this. I mean, we haven't got the results. I'm very curious.

GUZEY: This is great. I'm excited. This is exactly what I wanted to see. This is exactly the kind of sleep study I wanted to see. There are no RCTs where you just make people sleep more or less.

TREFETHEN: God, I can't believe I didn't tell you this earlier, given... We haven't done much in sleep, but...

GUZEY: Wonderful. Well, I'd be happy to see that my writing is making some impact... probably not, but cool.

TREFETHEN: [laughs]

Jacob’s advice and views on purpose

GUZEY: Advice for a talented 20-year-old watching this?

TREFETHEN: Oh, God. I don't know. Giving advice is so... People gave me advice when I was 20, and it didn't work.

GUZEY: How about advice to your 20-year-old self? Do you have any?

TREFETHEN: It's so hard, Alexey, because life ends up being what it is. And I can now see with more clarity what decisions I made ended up changing my life a lot.

GUZEY: Yeah.

TREFETHEN: What do I do with that, though? You know, I don't see the other side. I didn't notice at the time that much that deciding to move to San Francisco when I was 22 would be a big decision that would have lots of downstream changes in my life. And now I have a group of best friends here. I have no plans to leave anytime soon. My job is here. My value system, I'm sure, is infected by lots of people around me. So, did I know when I, on a whim – roughly on a whim – did that in 2015 after grad school that it would be one of the... I narrativize it as one of the best decisions I've made. So, I guess maybe that's my advice to a 20-year-old Jacob. Just do what you're about to do. It's really good.

Then, there's sort of more personal, like, what are the other really big ones? Whether you decide to break up with certain people or date certain people and all that. So, but not that generalizable. Then—

GUZEY: Any love advice you'd give to a 20-year-old Jacob?

TREFETHEN: Love advice?

Well, when I was 20, I was very happily – [laughs] I was about to say very happily married, but that's not true – I was a very conventional 20-year-old. I dated the same man for all of undergrad and had a wonderful relationship. So maybe my 20-year-old self was on to something, you know, he had it pretty good.

Then, I think I have all sorts, actually, of sort of smaller romantic advice, but probably not for the podcast.

GUZEY: Advice for me?

TREFETHEN: Wait, hold on. But can... I don't want to give it, but do you want to give it? Do you have romantic advice?

[laughs]

That's what your listeners tune in for.

GUZEY: My romantic advice for younger Alexey?

TREFETHEN: Yeah, well, don't go there if you don't want to go there.

GUZEY: My romantic advice to my younger self was when we had this eight-hour-long conversation in May of 2022 when I was very conflicted about a lot of things, I should have followed up with you sometime soon afterwards instead of just kind of sweeping everything we talked under the rug and trying to pretend nothing happened, and then it just blew up months later. Yeah.

But I think it's the same thing, actually. It's very much the same thing as with New Science, where you just didn't have people to talk to about things. And that it’s this weekly thing where I would occasionally talk to someone about things that were happening in New Science. But if I talked once and talked through all the problems, it only happens once, there is no follow-up, so it doesn't actually matter.

I think the biggest thing that I do differently now is I just talk to my friends consistently about what's happening in my life.

TREFETHEN: Oh, cool.

GUZEY: That's, yeah, I think that's universal advice. Talk to your friends who are not conflicted with you, not competitive with you, and who listen to you and who tell you what their thoughts are on whatever is happening.

Yeah, I think that’s the single biggest life lesson, in a way.

TREFETHEN: I co-sign that. You make me realize some of my life lessons are just learning from the inside the things that now sound trite.

Another angle on talking to your friends more about what's going on in your life is, like, who are your friends? The trite thing is completely true. The people you spend the most time with then shape who you are in terms of your values and what you think about.

I now understand that somewhat because I've changed. I can see the ways I've changed. I'm blind to many of them, I'm sure, but I can see some of them. And I can see the influence of particular people. I can see the influence of one particular person in this room, actually, in the way this room is set up. So, yeah, I think the extended mind as located in your closest friends or family or partner or children is very real. And so, that has various implications, yeah.

I'm with you. \

And if you don't like someone that much and you end up spending time with them anyway, you will become close to them and you will become more like them. That's another one.

GUZEY: [laughs]

TREFETHEN: There are different ways that can happen to people. Like, you hang out with your partner's friends a lot, for example. It's one.

GUZEY: What's the meaning of life?

TREFETHEN: At least we got easier than academia in 50 years. Of course, I dispute the question, but... You want one word, one sentence, one?

GUZEY: Whatever suits you?

TREFETHEN: I mean, if it's one word. Love is pretty close to the meaning of life in one word. So, maybe I'll go with that.

GUZEY: Can you expand?

TREFETHEN: That would be more than one word. How pretentious do you want me to go? I can...

GUZEY: I would say, okay, just say whatever you want. And then if you aren’t... we can cut it.

[laughs]

TREFETHEN: [laughs]

GUZEY: But I'm curious.

TREFETHEN: Maybe I'll relate it to our conversation about science and about methods.

I think some of the most confusing parts of being alive are that the questions that I care about most, or that people in general maybe care about most, are not always well matched to the methods we have for producing generalizable knowledge.

So, for example, a lot of the most important questions are in the social realm. And the methods of social science do not produce generalizable knowledge as much, as often as people want. And, you know, there's famous debates about which ones produce knowledge that's generalizable and in which ways. There is progress in those methods. I was just talking to a friend about progress in diff-and-diff and how you change the staggering of how you do diff-and-diff in economics. You know, you can make methodological progress, but...

And then some of the questions are, like, you know, is there a God? And what is the meaning of life? And why do I feel this way when it doesn’t make any sense? Why do I feel so overwhelmed, what does it mean? Or, I feel nothing... The things that matter so much, we have no methods, or we have very weak methods. So we live in a state of knowledge-lessness about the things that touch us most deeply. To try and make that more professional, if the things that matter most often don’t touch the scientific methods, what does that imply for our jobs? I sometimes wonder about that.

To get back to your first question, why am I doing what I’m doing?

I think what I'm doing, I can see value in without having to answer deeper questions about what is valuable. Because if you give people the chance to not die of preventable diseases, I don't have to have a further normative sense of what is good for them, what is good for the world. I can just know that those people will go on to do whatever they end up doing, whatever they want to do. They'll have families. And so it's a philosophically easier area to work in because I don't have to solve impossible questions.

I don't know. That's probably mostly a dodge of your question.

Jacob’s favorite joke

GUZEY: Can you tell me your favorite joke?

TREFETHEN: A joke that my friendship group riffs on commonly, so maybe this counts as my favorite joke, have you seen Airplane!? The movie, Airplane!?

GUZEY: [shakes head]

TREFETHEN: That's a huge one for you. You get to watch it for the first time. It's a kind of old slapstick-ish movie.

And there's a recurring joke that I'll try and remember one instantiation of, which is... So they're on an airplane. It's in the title. And someone's having a medical problem. And so a doctor coming down the aisle grabs the air steward and says, “This woman is in trouble. We need to find a place to land the plane. She's got to go to hospital –” or, “we need a hospital.”

Ah, I fucked up the delivery...

GUZEY: [laughs]

TREFETHEN: Okay, let's start again. Ready?

Wait, how does it go? How does my favorite joke go? Okay.

“We need to land this plane. She needs a hospital.”

“A hospital? What is it?”

“It's a big building with patients in, but that's not important right now.”

GUZEY: [laughs]

TREFETHEN: [laughs]

Then there's another version. It goes like,

“There's a problem in the cockpit.”

“The cockpit? What is it?”

“It's the small room at the front with the pilots in, but that's not important right now.”

GUZEY: [laughs]

TREFETHEN: And so, Alexey, I'm worried about how this recent podcast I did went.

GUZEY: What's a podcast?

TREFETHEN: You have to say, “oh, what is it?”

GUZEY: Oh, what is it?

Oh, sorry. Let's do it together. Let's do it together.

TREFETHEN: Alexey, I'm not feeling great about this. I'm worried about this recent podcast I did.

GUZEY: Oh, what is it?

TREFETHEN: A podcast. It's a recording of two men sitting in front of each other talking about – it's not important right now. It doesn’t have to be two men, but it statistically is.

GUZEY: [laughs] Statistically, that’s great.

TREFETHEN: Cool, thank you.

GUZEY: Thanks for doing this.

We’re incredibly fortunate to have the support of so many minds who share our vision of building new institutions of science. 2024 was a particularly exciting time for us. Here’s a look at what we were up to last year, and some of our intentions for 2025 (and beyond)…

🧠 Fellows

We had 10 new fellows in 2024…

1. Seyone Chithrananda (ML & Proteomics)

Seyone Chithrananda is an undergrad at UC Berkeley studying computer science and bioengineering. His research focuses on computational tools for designing and interpreting biological systems, working on projects like protein evolution and RNA design. He also co-organized the BioML seminar series and led Machine Learning at Berkeley’s research committee, mentoring young researchers and fostering a strong ML community.

2. Katya Osipova (Evolution & Genomics)

Ekaterina Osipova is an evolutionary biologist and postdoc at Harvard University, studying how evolution shapes extreme animal phenotypes. Her research focuses on metabolic and behavioral adaptations, such as how birds evolved to consume large amounts of sugar without adverse health effects, uncovering insights relevant to human diseases.

She combines computational approaches with experiments, exploring topics like positive selection, gene loss, and metabolic adaptations in species like hummingbirds.

3. Samarth Jajoo (ML & Proteomics)

Samarth Jajoo is an undergrad at UC Berkeley studying a mix of computer science, cognitive science, and molecular biology. He works on projects at the intersection of machine learning and biology, developing high-throughput methods for protein engineering and modeling protein conformations.

Samarth has previously worked on tele-operated robotics at Prosper and protein design at Popvax. He’s also contributed to programming education and built tools and side projects blending creativity and utility,

4. Mike Ferguson (Developmental Biology & Tissue Engineering)

Mike Ferguson is an interdisciplinary biologist and engineer dedicated to pioneering advancements in tissue engineering and regenerative medicine. By combining insights from stem cell biology, developmental biology and engineering, he focuses on growing human tissues using developmental biology-inspired approaches rather than assembling them artificially.

His innovative work on microfluidic blood vessel technology represents a step toward creating replacement human tissues—and potentially even organs—in the future.

5. Joshua Bauchner (History of Science)

Joshua Bachner is a writer and editor specializing in the history of psychology and physiology. He consults with science and tech businesses on idea generation and systematization, drawing lessons from the history of science. Joshua has written on topics ranging from the scientific importance of walking to modernist housing, and has extensive editorial experience working with authors, museums, and publishers.

6. Theia Vogel (Multi-agent AI & ML Interpretability). In collaboration with Prime Intellect who generously provided compute credits for her research.

Theia Vogel is a programmer and computational linguist working at SecureDNA, developing a system to screen DNA synthesis orders for hazardous material. They are passionate about various side projects including compilers, natural language processing, fiction, and game development. Notable projects include “sortes alearum” (a simulator for an ancient Anatolian oracle), “GPTed” (semantic spellchecking with GPT-3), and various games on vgel.itch.io.

7. Max Shirokawa (Atomic Physics & Open Science)

Max Shirokawa is a graduate student at MIT, working on open source hardware for atomic physics at OpenQuantum. Previously, he has held roles as a research scientist at NTT Research and lead threat investigator at Dropbox, as well as a research intern at NASA Jet Propulsion Laboratory.

8. Andy Kong (Human Implants)

Andy Kong is an independent researcher focused on human augmentation, computational photography, and personal informatics. Currently engaged in a “Year of Output,” Andy is passionate about exploring new sensations, training his senses, and pushing the boundaries of what is possible with advanced technology. His work combines innovation with a deep interest in challenging existing beliefs and creating transformative technologies.

9. Suspended Reason (Cultural History)

Suspended Reason is an independent researcher and essayist exploring topics like language games, dramaturgy, predictive processing, and cognitive aesthetics. Their work connects philosophy, fashion, and the inexact sciences, with influences from Bourdieu, Schelling, and Goffman. They run the Not Nathing press and the research community Pfeilstorch, focusing on cultural and intellectual intersections.

10. Michael Domarkas (Organoids, Neuroscience, and AI)

Michael Domarkas is a high school student in London researching organoid growth control and computational modeling of membrane electrical simulations. Longer-term, he hopes to combat Alzheimer’s and enhance human intelligence by improving synaptic plasticity.

🗞️ Some 2024 updates on our past fellows and grantees

1. Adam Green (2022 grantee) has started a biotech company — Markov Bio — working on interpretable biological simulators.

2. Isaak Freeman (2022 grantee) has become a graduate student in the Boyden Lab at MIT.

3. Adam Strandberg (2022 grantee) has become a graduate student in the Needleman Lab at Harvard.

4. Julie Chen (2022 fellow) got the prestigious Goldwater Scholarship.

5. Avadhoot Jadhav (2022 fellow) has become a graduate student at Karolinska Institutet.

6. James Heathers (2022 grantee) was featured in Retraction Watch with his work on science integrity.

7. Nora Belrose (2023 grantee) was featured in Vox’s Future Perfect 50 list.

8. Kevin Liu (2023 grantee) has joined OpenAI to work on AGI Preparedness.

Finally, Niko McCarty — our former Head of Media — was featured in Vox’s Future Perfect 50 list, alongside Nora Belrose, got Forbes 30 under 30, and has been driving Asimov Press as its Editor-in-Chief from strength to strength, making it the single best popular biology publication out there.

✍ Publishing

2024 writings on metascience and education from Alexey….

1. Losing The Idea of Progress

2. Thoughts on Science Technology Replication

3. The Ouroboros of Academic Peer Review

📣 Future

It’s thanks to the incredible generosity of our donors that New Science continues to support exciting projects and thinkers. In 2025, we plan to collaborate with even more brilliant researchers, helping them to push the boundaries of modern science, all towards the ultimate goal of building new institutions of basic science.

If you are interested in taking on a philanthropic role at NS, we’d love to hear from you! Please visit our donation page here or reach out to me directly at alexey@newscience.org.

Mike Ferguson is a biologist and engineer (BS Cell and Molecular Biology from Michigan, MS Biomedical Engineering from Boston University) working on growing human blood vessels in the lab to unlock replacement tissue/organ therapies. From a young age, his sole motivation in life has simply been "making people live longer". He believes replacement tissues/organs will be helpful in this endeavor.  Mike spent 3 years working in a stem cell/developmental biology lab growing organoids and studying various ways that people were trying to make replacement tissues/organs. Realizing the need for a way to grow human blood vessels in the lab to unlock this dream, he pursued an engineering degree, believing that engineering tools would be required to solve the blood vessel problem. He has built his own lab to manufacture "microfluidic" devices, which are needed for his work, going so far as making cheap versions of expensive lab equipment to do this. Mike likes when biologists and engineers work together. He believes in a developmental biology approach to tissue engineering - growing tissues, rather than trying to artificially piece them together.

Mike is developing a microfluidic device and associated tools that enable one to grow "real" perfusable and transplantable human blood vessels in the lab. The technology is unique in that the blood vessels are formed using the same process that occurred when you were an embryo. Having previously demonstrated that the microfluidic technology can be used to grow scaled up tissues with perfusable human blood vessels, he is currently working on making the technology more mass producible and user friendly. Assuming he can raise the required funding, he plans on transplanting the blood vessels he is growing into rats to demonstrate the clinical potential of the technology.

Samarth Jajoo is a third-year undergraduate at UC Berkeley studying Computer Science. He is passionate about using machine learning methods to model biophysics and runs the BioML seminar series at Berkeley to create an active community in the field. Previously, Samarth worked on developing software for vaccine design at PopVax, optimizing tele-operated robots at Prosper, and contributing to web-based IDEs at Replit. 

As a New Science fellow, Samarth is working in Sergey Ovchinnikov's lab at MIT, where he is building ML models for protein folding that operate by simulating each amino acid as an agent. These models provide interpretability, and could also offer insights into protein dynamics and the biological process of protein folding, despite being trained solely on static crystal structures. Additionally, Samarth is exploring the use of evolution strategies to sample diversely from existing protein folding models, which could lead to better results with significantly reduced computational requirements compared to current state-of-the-art methods.

Katya Osipova is a postdoc at Harvard University working in the field of evolution and genomics. She is broadly interested in how evolution and natural selection shape phenotypic diversity, bridging the gap between genomic changes and phenotypic variation by combining computational  and experimental approaches. She is especially interested in extreme phenotypes observed in animals, specifically unlikely metabolic and behavioral adaptations that could give insights into mechanisms of human diseases. Katya majored in chemistry at Moscow State  University in Russia and completed her PhD in evolution and bioinformatics at the Max Planck Institute in Germany where she worked with large-scale genomic and transcriptomic data sets to find insights into the evolution of extreme metabolic adaptations in hummingbirds. Outside of research, she does everything that is somehow related to music: from listening to producing and recording trashy covers.

As a New Science fellow, Katya will look into how metabolic systems evolve under thermodynamic constraints in the Informatics  group.  She will start with a comparative analysis of eukaryotic metabolic networks, leveraging publicly available genomic data. Next, she plans to use examples of extreme metabolic adaptations to build predictive models of pathways' energetic outputs. Finally, Katya plans to use genome editing in cell lines and organoids to generate novel living systems with defined energetic and metabolic properties.

In other news, our infra tooling experiments are well underway. Along with forming the backbone of our project tracking for each fellow, we also explored using Linear for several book discussions, and have arranged several meetups among fellows online through Gather.town, virtually, as well as in person where feasible. We’re excited to share more about the results of our fellows’ projects soon!

best,
–Sasha

(Correction from the previous newsletter - Scott Berger was awarded the ARCS Foundation scholarship rather than ASCB graduate student award.)

As of the start of the year, I’m very grateful to continue the work that Alexey has done to get New Science to where it is today, in the role of executive director. 

This month, we: 

• Spent time delineating our deliverables for the quarter and year. One among them is creating a playbook for our fellowship programs and remote-first sharing + project management stack. This will include open resources on how we use Distill, Linear, and Watershed on the infra side. We look forward to feedback on how others use these tools.

• Onboarded several new collaborators + contributors to the New Science team

• Soft-launched our research website (research.newscience.org) — many improvements coming very soon.

Our theme of the year (which arose in consilience between our fellows and internal team) is shipping. Our focus area for the quarter is largely executing on this tooling + collaborations and we expect to have prototypes to share within Q1.

This transition point also seems like a fitting point to review what we’ve done at New Science the past year. Some high points from there —

Our footprint is growing! Collin - Maple Ridge; Dylan, Scott and Julie - Stanford; Diana - Berkeley; Michael - Toronto, Yasmeen - Montreal; Avadhoot - Pune; Adam, Benjamin, Daniel, and Alex - Boston; and Alex - NYC.

In 2023 we expanded beyond the SF Bay Area and Boston science ecosystems and supported fellows located in New York City (Rockefeller), Canada (Toronto, Montreal and Vancouver) and India (Pune). We’re looking forward to hosting more opportunities like our last ‘Demo Party’ (fellow retreat) to gather face-to-face.

Receiving well-deserved accolades, our one year fellow Scott Berger became an ARCS Foundation Scholar, Avadhoot Jadhav was accepted to begin PhD studies at Karolinska, Adam Strandberg was accepted to PhD studies at Harvard, and Benjamin Chang won a Rhodes Scholarship to Oxford. We could not be prouder and wish them the best for their adventures in these new places.

New starting this year, we have a ~quarterly snail mail update as well (which does not fully overlap the above). If you’re interested in receiving it, please fill out the form here (friends.newscience.org/updates) and we will send it to you.

We will have more information on most of the areas above, namely our progress to date, in February + March and are looking forward to sharing + getting your input. Until then! ❄️

—Sasha 

---

Hi everyone,

It's been a huge privilege for me to spend the last several years studying metascience and to run New Science. We've supported many dozens of young scientists with fellowships, grants, and mentorship, published metascience research I'm incredibly proud of, and learned a great deal about how science really works ourselves.

However, having been focused on metascience since 2018, I've come to realize it's time for me to take a break. Next, I’ll be in Boston learning physics — say hi if you’re around!

In light of this, Sasha Targ, New Science's Head of Research is now going to be the Executive Director, assuming the leadership of the organization and its mission. While I am a complete scientific ignoramus, Sasha has a PhD in bioinformatics from UCSF, is a real scientist, and is otherwise better than I am at almost everything. I'm excited to see what she has prepared for science and for New Science in the coming years!

Cheers,

Alexey

• Our one-year fellowship begins. Julie Chen and Scott Berger will spend the next 12 months in Stanford and Boston. They will probe the causal role of methylation in aging and explore the evolution of multicellularity (which has popped up independently at least 25 times)! More details soon.

• Last week, we hosted a Demo Day in Cambridge. Summer Fellows presented their work and got feedback from local scientists. This event marks the official end of this year’s program. 🎻

  • Each fellow wrote an essay and technical report about their work. We’ve made the former publicly available on our Substack and website. Full-length research reports will be published soon.

  • Two summer fellows remain in Cambridge and will continue their projects. Avadhoot Jadhav, in Hidde Ploegh’s lab, will continue building a “universal immunotherapy” using nanobody-peptide fusions. And Diana Leung remains with Xu Zhou; her project explores how cells coordinate and change behaviors when cultured near other cells.

• We made a $10,000 grant to Dylan Husmann at Stanford University to study tunable mutation rates.

• We made a $35,000 grant to James Heathers to develop software that will semi-automate detection of spurious statistics in research papers.

Projects

• We’re building a marketplace to connect scientists with philanthropists. If you have an ambitious idea for a research project but don’t know where to go for funding, please email us. We’ll help you write a 1-page proposal — with only minor requirements — and will consider funding your idea.

• We’ve published several pieces on the Substack lately. We think you’ll enjoy:

  • This history of the Rockefeller Institute’s Natural Sciences division, and how they spearheaded the mid-century revolution in molecular biology.

  • An essay on cell culture and (missing) context in experiments.

  • An essay on scientific ‘style,’ and how it went missing from our modern journals.

  • An upcoming piece on 20th-century pioneers. Keep an eye out! 👀

    • If you have an essay idea about science or metascience, please email us.

Interesting Jobs

• Arcadia Science always has compelling job openings. Applications are still open for their Entrepreneur-in-Residence program.

• Lorentz Bio is looking for a Director of Biology to join their founding team.

• Jakob Voigt’s Open Ephys (short for open-source electrophysiology) is hiring. They want to make the future of scientific instruments equitable and open. 

  • “Join us to work on high-performance open-source tools that read from and write to the brain. We are looking for an electrical engineer for firmware development, hardware design, and integration with custom ASICs. You will have the opportunity to propose and develop complete projects and will be working on tools that push the boundaries of neuroscientific discovery over the next decade. These tools will be used by thousands of scientists around the world, from those at elite institutions to labs that were previously priced out of the scientific conversation. For inquiries contact jpn@open-ephys.org.”

Stay Frosty,
Niko & Alexey

“We are not made up, as we had always supposed, of successively enriched packets of our own parts. We are shared, rented, occupied… the colonial posterity of migrant prokaryocytes, probably primitive bacteria that swam into ancestral precursors of eukaryotic cells and stayed there…

I like to think that they work in my interest, that each breath they draw for me, but perhaps it is they who walk through the local park in the early morning, sensing my senses, listening to my music, thinking my thoughts.”
— Lewis Thomas, The Lives of a Cell

Plants weren’t the first to figure out photosynthesis, that careful dance of chemistry that harnesses sunlight to make our food and air. That honor belongs to microbes that first appeared 3.5 billion years ago, in the earliest known days of life on Earth. Some of them, called cyanobacteria, learned to split water into hydrogen and oxygen, and thus created our oxygen-rich atmosphere.

And then, about 1.5 billion years ago, something peculiar happened. A free-living cyanobacterium was engulfed by a larger eukaryotic cell in a phenomenon known as endosymbiosis. That bacterium escaped destruction and became a permanent resident — a cell living within another cell.

Soon, both parties realized they had stumbled upon a mutually beneficial arrangement. The larger, eukaryotic host siphoned food and energy from its photosynthetic guest; in exchange, it offered it shelter from the outside world. The cyanobacterium eventually lost its ability to live independently, handing off regulatory control to the eukaryote in a series of gene transfer events. It became a chloroplast: a specialized compartment that endowed the host with the ability to photosynthesize. All plants, in turn, descended from this single cell.

This transition from free-living cyanobacterium to integrated organelle was a landmark event in evolutionary history. An earlier event led to the creation of mitochondria.

Yet as momentous as endosymbioses are, the precise details of how they arise are still murky. There is much that we stand to learn by reverse engineering the process — but it won’t be easy. Scientific tools are limited, and deciphering this story, obscured by billions of years of complex coevolution, is akin to untying a Gordian knot.

I’m conducting a series of experiments aimed at reversing and reconstructing the chloroplast’s strange and fascinating history. The results could yield new insights into the evolution of multicellular life.

Modern endosymbiotic theory gained widespread acceptance in the 1980s, thanks to the work of Lynn Margulis, its principal champion and “vindicated heretic”. Her 1967 paper, On the origin of mitosing cells, weathered dozens of rejections before it was finally published. It’s now considered a watershed classic. 

Before Margulis, studies into the origins of chloroplasts and mitochondria were driven principally by the work of a few intrepid scientists who made careful observations — and persisted in the face of ridicule. The earliest known mention of endosymbiosis comes from an 1883 article by Andreas Schimper, a German botanist. In On the development of chlorophyll granules and colored bodies, he acknowledged the similarities between cyanobacteria and chloroplasts, before quietly adding a footnote:

“If it should definitively be confirmed that the plastids in egg cells [of plants] are not formed anew, then their relationship to the organism containing them would be somewhat reminiscent of a symbiosis.” (p. 112, footnote 2, emphasis added)

By 1905, the concept had been formalized in essays by the Russian botanist Konstantin Mereschkowski. In On the nature and origin of [plastids] in the plant kingdom, he outlines an endosymbiotic theory that comes astonishingly close to our modern insights. To his more influential contemporaries, though, his ideas were “too fantastic for present mention in polite biological society.” Ultimately, his work did not reach the mainstream — in part because it was overshadowed by his legacy as a serial rapist of children, who often evaded capture by fleeing to new countries.

American biologist Ivan Wallin argued for an endosymbiotic origin to mitochondria in the 1920s. To support his claim, he carried out a series of audacious experiments aimed at isolating and culturing mitochondria ex vivo. He claimed to have succeeded, but his results later proved irreproducible. Perhaps unsurprisingly, he too found little support among the scientific establishment of his day.

Things began to change as biologists developed new tools in the 1960s. Early genomic analyses crucially showed that chloroplasts contain their own DNA. Electron microscopy yielded the first detailed structural comparisons between chloroplasts and cyanobacteria. Unlike her predecessors, Margulis arguably prevailed thanks to this new experimental evidence.

Today, scientists have a greater palette of technologies to draw from than ever before. We can probe, modify, and dissect living things in ways that were previously impossible. Modern DNA editing, for example, allows us to engineer both the chloroplast and nuclear genomes as we desire; some labs are using these tools in attempts to recreate synthetic endosymbioses. 

All this gives rise to a daring question: Can we restore chloroplasts to a free-living state by re-endowing them with the genes that they’ve lost over 1.5 billion years of evolution? Even incremental progress towards this goal would yield important biological knowledge. As we work to wrestle back control over the organelle, we’ll uncover how the cell cycle, metabolism, and gene regulation are shaped by the interaction of host and endosymbiont. In liberating a chloroplast, we’ll be forced to unravel its complex evolutionary history using active experimentation and tinkering, rather than theoretical musings. To paraphrase Feynman, we should create a free-living chloroplast so that we may understand it.

During my New Science fellowship, I isolated chloroplasts from cells of the green alga Chlamydomonas reinhardtii using established methods. I’m carefully tweaking the chloroplasts’ environment to see which chemicals can support their metabolism in vitro. My first goal is to better understand how to make viable chloroplasts without any genetic engineering.

In the future, I intend to rewind evolutionary history by moving genes that were lost to the host over the last 1.5 billion years back into the chloroplast. In doing so, I aim to provide the first demonstration that chloroplast-derived, nuclear-encoded genes can be successfully repatriated without disrupting their function. I’m currently testing an existing method for inserting DNA into the chloroplast genome. Soon, I’ll compile a list of “high confidence” genes to be relocated. This may require the development of new computational approaches that can screen candidates by mining a combination of sequence data and the published literature.

These efforts will bolster our understanding of chloroplast function and phylogeny. In the future, they may even make it possible to engineer a new, free-living cell using an organelle as a starting point. What will this liberated chloroplast teach us — about its evolutionary history and the meaning of life itself? It’s time we find out.

Edited by Niko McCarty

Subscribe now

Thanks to Lev Tsypin, Adam Strandberg, John McCutcheon, Sasha Targ, and Alexey Guzey for feedback on this essay.

Cite this essay:

Faizi, K. “Reversing 1.5 Billion Years of Evolution.” newscience.org. 2022 September. https://doi.org/10.56416/720qud

Pathogens constantly spread, infect, and evolve. Lower respiratory tract infections and diarrheal diseases remain amongst the world’s top five leading causes of death. More than 6 million people — a conservative estimate — have died from COVID alone and new variants continue to emerge. Despite the threats that pathogens pose, we have limited tools to design drugs to fight them. Indeed, drug discovery remains a lengthy, costly, difficult, and inefficient process. In the COVID19 pandemic, nature is (kind of) winning — vaccines and antibodies help, but it is hard to keep up with an evolving virus.

Mother Nature can serve as a source of inspiration for drug design. The human immune system has its own unique method to keep pace with evolving pathogens. And, no matter how virulent a pathogen is, the human species has so far managed to adapt and (largely) survive.

It is thanks to proteins in our blood, called antibodies, that we can swiftly shut down dangerous invaders. With protein ‘tips’ of a flexible design, antibodies can “see” different structural features in a pathogen’s proteins. Upon binding to a viral invader or other pathogen, the antibodies signal and recruit various immune components, such as effector cells (a type of white blood cell), to eliminate infected cells.

Although we have gained some knowledge of how the immune system operates, the sheer complexity involved in its adaptive behavior is not easy to understand. But imagine if we could replicate this process synthetically. I propose a therapeutic strategy to do so, using protein engineering.

Physicians and scientists have made many attempts to manipulate the immune system. Vaccination and passive immunization are two examples, and both have been wildly successful, with millions of lives saved.

By exposing your body to a weakened version of a pathogen, or just a piece of a pathogen, one can train the immune system to produce antibodies that help fight that pathogen when it is first encountered in the wild. Despite the success of vaccines in tempering the COVID19 pandemic, there is already evidence of enhanced transmissibility and reduced vaccine effectiveness against emerging variants of concern, including Delta and Omicron. Current vaccines also have a narrow target range.

Moreover, a slow vaccination drive tends to increase selection pressure on the pathogen, leading to the emergence of more infectious variants. Current vaccines cannot keep up with new variants, and revamped generations of vaccines are constantly required. There are also psychological and logistical limits to how often people can be vaccinated, and we are still collecting data about these vaccines and immune-related adverse events.

At Caltech, Pamela Bjorkman has developed a mosaic-nanoparticle based vaccine that covers a broad range of targets, including variants of concern. Still, it can only prevent new infections once vaccine-driven immunity has built up, which takes time; it cannot deal with established infections.

Passive immunization is one alternative to vaccines — people are given antibodies that recognize the pathogen in question. But monoclonal antibodies, while efficacious, are expensive. Regeneron’s two-drug cocktail for COVID19 costs around $1,250 per infusion, according to Kaiser Health News. Monoclonal antibodies also must be kept cold, so as not to lose their efficacy. Even minor changes in a target virus’ protein structures can reduce the antibody’s effectiveness. That is why we still need a sustainable, cost-effective therapy with broad target specificity.

Two years ago, I was studying drug resistance in malarial parasites (as a part of the iGEM IISER Pune Team). It was then that I realized how structural changes in a viral protein, caused by mutations from drug-driven selection pressure, can be a root cause of therapies losing their efficacy. Naturally, I wondered: If shifting proteins can evade antibodies and cause this problem, maybe they can also provide a solution. After all, if nature has managed to devise an adaptable approach for keeping up with new pathogens, why can’t medicine do the same?

Proteins are modular in nature and are composed of domains, which are independently folded regions within a protein. Such modularity enables functional protein domains to be joined together synthetically or genetically, with the goal of altering their function to modulate the immune system. The therapeutic efficacy of a protein of such a design can be enhanced by choosing fusion partners that selectively target particular cell types or receptors.

Nanobodies — the smallest immune recognition modules derived from a unique class of antibodies found only in camelids — serve as the perfect functional building blocks to create such fusion constructs. They can recognize antigens, they are easy to produce and — equally important — they are cheap to make. Nanobodies show exquisite specificity, they are structurally stable, and they are capable of modulating immune responses.

At New Science, I am using protein fusions to develop a pan-coronavirus therapy.

Antibodies have two functions: they recognize antigenic proteins, and they recruit immune components. Immune evasion occurs when small structural changes in the antigenic protein render the antibody unable to recognize it. To overcome this problem, I built a computational algorithm that uses crystallographic structures of various protein-binding partners, such as receptors and antibody complexes, to generate a library of peptides, or small proteins, that can bind at many distinct sites on the viral protein’s surface.

My summer plan, from here, is simple. I took my computationally designed peptides — each of which recognize a site on the coronavirus' spike protein — and attached each of them, individually, to nanobodies that recruit different immune components to selectively eliminate infected cells.

Unlike monoclonal antibodies, which “see” a single structural feature on a virus, my approach can theoretically address any emerging variants of a viral pathogen. Changes in the virus’ antigenic protein, acquired by mutation and selection, should not result in resistance or impact the peptide cocktail's function because the designed peptides target several distinct sites across the protein’s surface. While resistance can easily develop against a single peptide or antibody, a range of peptide-nanobody fusions, each recognizing a distinct epitope, would be far more challenging for a virus to evade.

After assembling the peptide-nanobody fusions, I will test them in mice to see if they are effective at preventing or treating SARS-CoV-2 infections.

Are there limitations to my approach? Side effects may occur, as is the case for any other new drug. But I don’t know. This project is in its early stages, and any fusion proteins would go through rigorous, years-long testing, much like other therapies.

If this project works as intended, we will have developed a tool that can harness the immune system in completely new ways. It should be possible to selectively target, label, or destroy any cells we desire (be it SARS-CoV-2-infected cells or cancer cells), based on the antigens that are expressed on the surface of those cells. Most importantly, this type of therapy is inherently modular — built by fusing bits of proteins together — and should thus enable us to deal with any new or emerging pathogens. Only time will tell, and I’ll update soon with my progress.

Edited by Niko McCarty

Subscribe now

Thanks to Alexey Guzey, Sasha Targ, and Hidde Ploegh for reading this.

Cite this essay:

Jadhav, A. “Towards a Universal Immunotherapy.” newscience.org. 2022 September. https://doi.org/10.56416/591plq

If you put me on a desert island, I would die an early and pathetic death, and I wonder if that’s what happens to cells in culture, to some extent. 

Cell culture originated in the 20th century. The initial goal was simple: Make cells viable long enough to study them. And because the cells in our modern labs seem happy enough, the practice hasn't changed much in the last few decades. But a constellation of experiments suggests that a cell’s properties are conditional on context — neighbors and environment — and there is much to explore about how they behave outside of a dish in a lab.

On Change & Identity

Cell culture exerts a slew of selection pressures on living cells. Media and environmental conditions are not neutral. Scientists almost exclusively study a subset of cell types and behaviors that they can reproduce in normal laboratory environments. When you take cells from the body and try to grow them in cell culture, many don’t survive. Those that do survive change with time: Over a number of growth cycles, most primary cells drift from their original identity — properties like stemness, morphology, and gene expression patterns. After a time, they are thrown out, no longer relevant for experiments. Other cells pick up mutations and become immortalized spontaneously. A cell’s identity is fluid, easily deformed by its contextual container, and constantly reacting to the environment. 

It’s telling that the cells which prosper most in culture are not cells that are representative of cells in our bodies. The first human cells to successfully thrive in cell culture were unusually aggressive cancer cells taken from Henrietta Lacks in 1951. They’re so well-suited for the lab environment, in fact, that they are often studied accidentally because they contaminate and outcompete normal cell types in vitro. HeLa has grown so distinct from its human origins that it has been proposed to be categorized as its own species, with varying subspecies by lab or region. The robustness and adaptability of HeLa cells that make it a poor team player in the multicellular context are characteristics that contribute to its success in culture.

In an interesting mirror to HeLa cells and the usual negative framing around cancer, some cancers can adapt to welcoming cellular environments, eventually integrating harmlessly into their hosts. For example, a specific type of cancer can merge with mouse embryos to become viable chimeras. There are also neonatal cancers that don’t require medical intervention because they are expected to spontaneously regress. Given sufficient signals/direction, a ‘bad’ cell can integrate into a functional body. On the other hand, stem cells, for all of their therapeutic potential, also drift and accumulate cancer mutations if they are cultured for an extended length of time. They often develop into tumors, too, for lack of sufficient direction. In all of these contexts, the question is less about a cell’s original identity, and more about how they integrate into the larger system.

For multicellular organisms, we ultimately evaluate functionality at the organismal level. Here, every kind of cell plays a role in a larger ecosystem, and the identity of each cell changes as its context changes. To study these cells in isolation destroys the signals that maintain their functional identity. You can have the right parts, but without the right organization and integration, we will see them as a pathology, a cancer. So how can we understand the factors that inform a cell’s identity from first principles? These are the questions that I’m exploring at New Science.

Understanding Identity

Consider two ways to understand what happens to cells when their neighbors and environments shift: From the top down, or from the bottom up.

Aging is a great example of the former. While working at Yale, I was introduced to epigenetic clocks, which are algorithms that calculate biological age based on how many sites, across the human genome, are covered in methyl groups. This methylation data is used to predict a person’s risks for illness or death, and to devise models that predict age, morbidity, and mortality. Such statistical models are intriguing in their ability to predict high-level phenomena like how people age, when they might die, and whether they’ll get sick. But they are limited in their ability to suggest biological mechanisms.

So I turned to a nascent field called cell systems, which is about understanding how changes in a cell’s multicellular context can shape its behavior; a bottom-up approach. At Boston Children’s Hospital, Xu Zhou’s lab is examining the interactions between macrophages and fibroblasts, two cell types commonly found in mammalian tissues and involved in tissue homeostasis. Zhou previously found that fibroblasts regulate the carrying capacity of macrophages, or how many of these cells can grow in the body. Fibroblasts help to increase macrophage counts during an inflammatory response, and hold it steady at other times. His work offers a template for examining how emergent properties arise from individual parts. When we isolate cells and grow them in culture, we miss these emergent properties entirely.

In Zhou’s lab, I am co-culturing fibroblasts and macrophages and using the tools of RNA-seq and metabolomics to examine the differential expression and phenotype of these cells in monoculture versus coculture. My aim is to study these cells from the bottom up: to explore the emergent properties of mixed cell systems; of how the presence of other cell types influences their behavior. Other scientists, such as Michael Levin and Michael Todhunter, are conducting related experiments to organize normal frog cells into living robots, or to examine the impact of growth media and other culture conditions on cell viability, respectively. Those efforts, too, will significantly improve our ability to predict and manipulate cell dynamics.

A historical, scientific focus on individual cells, and individual genes, has led to impressive accomplishments. We can now induce differentiation of cell types with a specific set of transcription factors, and we can grow organoids that self-organize to recapitulate subsystems. But we do not fully understand the various knobs that we are turning to trigger those behaviors. Expanding the repertoire of cell culture and artificial differentiation, today, more resembles art than science.  

I suspect that greater attention on how parts interact as a whole, as opposed to measuring characteristics of parts themselves, will help us grasp those unique factors that inform a cell’s identity. Getting at this ground truth will improve stem cell and regenerative therapies in vivo, as we direct engineered cells to better interact with their neighboring cells. 

Biological systems have often been analogized to machines, but it can also be useful to cast multicellularity as a system of coordinating individuals, with the interest of aligning local behavior with global goals. This is why cells in the body parallel individuals in a society. But we’ve only scratched the surface. We will surely uncover more surprising connections when we study cells from the bottom-up, and in their natural context.

Edited by Niko McCarty

Subscribe now

Thanks to Oliver Yeung, Brian Toledo, Alexey Guzey, Sasha Targ, Brian Timar, and Steven Elleman for reading.

Cite this essay:

Leung, D. “A Call for Context (in Cell Culture).” newscience.org. 2022 September. https://doi.org/10.56416/021uwn

“There is no cost to getting things wrong. The cost is not getting them published.”

– Brian Nosek

In the 1990s, a small biotechnology company tried to develop food crops with enhanced traits. Taller shoots, fatter fruits, drought resistance — that sort of thing. The company designed experiments and ran field trials for 15 years.

Results were published in peer-reviewed journals. Everything seemed on track. The small company convinced a large, multinational seed company to license the improved strains — and that’s when the problems started.

Faced with inconsistent results, the small company hired a statistician to review the data. The verdict: It was impossible to tell whether the engineered plants were statistically superior to wildtype. The company took this as a null result and said "At least now we know it doesn’t work!" (Perhaps they were mindful of the famous story about Thomas Edison failing in his first 10,000 attempts at inventing the lightbulb.) The statistician replied: “Actually, the data you gathered don't even let you conclude that.” 

Years of effort and $70 million, down the drain. After testing all those lightbulbs, they had forgotten to plug them in.

This same story has played out many times in drug development, too. Most major randomized, controlled trials before the year 2000 showed therapeutic benefits (Figure 1), according to data on NHLBI-funded trials for cardiovascular disease between 1970 and 2012. But after the year 2000, when the agency codified hypothesis pre-registration, trials mostly ended in null results. Useless drugs, with possible side effects, had almost certainly been approved, manufactured, and swallowed by Americans.

Scientists are taught pre-recorded hypotheses and rigorous statistics — those fundamental tools of good research — but apply them inconsistently. In this sense, they can learn from the law.

The vast majority of neuroimaging studies are underpowered and rarely produce results above noise. The odds that an average neuroscience study is true is 50-50 or lower, according to a 2013 review. And an estimated 50% of studies in biomedicine “have statistical power in the 0–10% or 11–20% range, well below the minimum of 80% that is often considered conventional.” P-hacking and hypothesis fishing are rewarded because they make for more remarkable, and therefore more publishable, results (Figure 2).

The legal system, by contrast, is obsessed with questions of proof and evidence. Lawyers have rigid structures to test the hypotheses underlying a legal battle. Lawyers are trained to study and argue both sides of a case, carefully finding holes in each. Evidence unrelated to a case can be thrown out by a judge, who is supposed to be an unbiased referee of competing claims. Science is inconsistent, with ad hoc rules, in comparison. 

The U.S. judicial system is not a paragon, but pivotal breakthroughs in understanding usually come at points of convergence between two seemingly unrelated subjects. Appreciating the epistemological differences between science and law may give researchers and policy experts a new set of tools for thinking through science’s problems.

Consequences

The legal system is built on consequences. The wrong decision can ruin lives. If a judge is not impartial, or a defense attorney is incompetent, an innocent person goes to prison. A ‘remarkable’ result based on biased truth-finding or planted evidence is a wrongful conviction. The stakes are high.

In science, consequences for wrongdoing seem rare. I know this from experience because my first writing gig was at Retraction Watch. 

While there, I reported on many scientists who did ethically dubious things — they published papers with manipulated images, served as a guest reviewer at a journal and only accepted papers that cited their work, or touted (ineffective) COVID drugs while failing to disclose familial or financial ties to the drug’s manufacturer. None of these people, to my knowledge, have been punished. 

Even when editors are presented with clear evidence of data manipulation, they can take years to issue a simple expression of concern. In some cases, blatant data manipulation is punished with a slap on the wrist, or authors are asked to publish a correction to their (undermined) study. The current paradigm for dealing with misconduct is to preserve the integrity of the scientific literature, rather than punish individuals.

A veteran UCLA researcher, Janina Jiang, faked data in 11 different grant applications, 3 of which were funded for $58.7 million, according to reporting by Retraction Watch. The punishment? “Three years of supervision for any federally funded work.”

Jiang still appears to be employed at a UCLA affiliate hospital.

Subscribe now

Evidence

The legal system has a nuanced view of evidence. There are special rules that dictate what and how evidence is presented, or which evidence is sufficient to establish proof in a case. Evidence can be thrown out entirely if it prejudices the jury, or was obtained through unlawful means.

There is transparency in the legal system, and that is important. It's not enough that the right answer is found; the process by which it is found must also be correct. Usually, that means how evidence was collected must be transparent and well-documented — something that the NHLBI only came around to in 2000 (Figure 1).

Most of science, by contrast, still runs on the “honorable gentleman” system. Admissibility of evidence is basically, "You know it when you see it.” Some academics give the pharmaceutical industry a bad rap, but it’s clear that the latter’s standards for evidence are far higher. Much like the law has codified processes and rules for (legal) trials, large biotechnology companies have rigid, formal procedures for their own (clinical) trials.

Biotechnology companies follow rigorous rules because they know their data must be scientifically sound — at some point, the FDA will snoop around. Data must hold up, or the company can’t reap financial rewards.

Academic scientists, by contrast, garner money, prestige, and tenure through short-term impacts and papers. Reviewers rarely have time to fact-check a paper’s data analysis or statistical methods, and most grants are awarded to scientists for short-term, “sure-to-succeed” research proposals.

Many scientists are familiar with the 2005 Ioannidis paper, and its claims that the majority of published “research findings are false.” But the consequences of such falsehoods are less appreciated and lead to billions of dollars in waste each year. A decade ago, the drug company Bayer halted “nearly two-thirds of its target-validation projects because in-house experimental findings [failed] to match up with published literature claims,” according to a news piece in Nature Reviews Drug Discovery. 

Besides fraught standards for collecting evidence, academics are also less likely to report evidence compared with drug companies (at least, when it comes to clinical trials). The Mayo Clinic was compliant with FDA reporting rules for just 21.3% of trials, according to 2020 data from Ben Goldacre. The National Cancer Institute (which is part of the NIH) was compliant for 30.4% of trials. Pfizer? 92.9%. Novartis? 100%.

The U.S. Food and Drug Administration can impose hefty fines on non-compliant sponsors who do not upload results from a clinical trial. But they don’t, and they continue to award grants to researchers who break the rules. The standards between academia and industry are vastly different.

Share New Science

Adversaries

The American, English and Canadian legal systems are inherently adversarial. Lawyers argue for their clients, and the truth is thought to emerge from this struggle dialectically. Disputes are settled by ‘impartial’ judges.

Good lawyers, famously, know how to argue both sides of a case. They are trained in this art form early in their career; usually in the first year of law school. The best lawyers craft steel-man arguments to sharpen their case — they build up the best form of an opponent’s argument to find holes in its logic. A trial lawyer’s sense of truth, then, is often provisional.

An ideal scientist is her own worst critic, and is receptive to oppositional arguments. But debates seem quite rare at the institutional level. Domineering voices are all too common in NIH study sections, and “uniform” ideas often emerge victorious. Such siloed thinking can stifle progress.

Consider the Alzheimer’s cabal, a cadre of scientists who believe that beta-amyloid accumulation in the brain causes Alzheimer’s. To support their argument, they shut down competitors, rejected grant applications, and doled out money to friends from within NIH study sections. 

In 2006, a Nature paper claimed that Aβ*56 is associated with cognitive decline. But recently, a six-month investigation by Science found “strong support” for image tampering by its lead author, Sylvain Lesné. The consequences? Lots of investigations, scant decisions, and many “no comments”. Bureaucracy as usual.

And the ongoing Alzheimer’s story is one fish in a roiling sea. In biomedicine, an estimated “3.8% of published papers contained problematic figures, with at least half exhibiting features suggestive of deliberate manipulation,” according to a 2016 analysis of more than 20,000 studies from 40 journals.

It’s unfortunate that, in study sections, many researchers are overwhelmed and simply don’t have time to consider dozens of proposals and debate their merits. Early-career faculty, perhaps cognizant of career suicide, are often hesitant to oppose decisions by senior faculty. And moving NIH peer reviews to Zoom, due to COVID, caused about 30 percent of scientists to contribute even less to discussions.

Leave a comment

Margins

“Everything runs downstream from culture.” 

– Tyler Cowen

Progress is made by small improvements to culture. Some scientists p-hack, but at least now we talk about it. Publication bias is still a major problem, but at least it’s recognized by editorial boards. Data sharing is still a problem, but at least journals are shifting policies. It seems we are on the right track.

But problems in science are hard to solve because they are cultural, and culture is dominated by incumbents who benefit from the status quo. 

In 1930, legal scholar Karl Llewellyn published a book called, “The Bramble Bush.” In it, he argues that the officials in charge — sheriffs, clerks, judges — do more than just settle disputes; they dictate the law itself. Bloomberg columnist Matt Levine’s thoughts on Llewellyn is prescient for science and its problems, too (emphasis my own):

“...I went to law school. And I took the first-year course in Constitutional Law, and I learned about the fundamental principles that rule the United States. And I learned -- or at least was given the general impression -- that…the Constitution has served as a wise guide and constraint on the power of our rulers, and the foundation of our system of government. But in the back of my mind I thought about Llewellyn. I thought about the fact that those principles can't automatically enact themselves, that they only work if the human actors in the system choose to follow them and to demand that others follow them. They persist because the people constrained by them believe themselves to be constrained by them…Their magic is fragile, and can disappear if people who don't believe in it gain power. 

Improving science takes time, and will demand that individuals who believe in the truth, and want to get at the truth, acquire power. A new generation of ideas must strangle the old. 

Pre-registered hypotheses should be a requirement for just about every experiment. Scientific misconduct should be punished. Domineering voices should be removed from NIH study sections.

In a hundred years, I suspect our grandchildren will look back at our modern, scientific landscape and laugh at academia’s cavalier attitude towards rigorous statistical training, p-hacking, and hypothesis switching. Many studies are irreproducible and poorly designed. The whole system seems Victorian in its regard to prestige and social status, much like the legal system 150 years ago.

But if scientists behaved a bit more like lawyers, at least on the margins, perhaps we’d build a better future for those grandkids all the same.

Thanks to Kian Faizi, Benjamin Reinhardt, Adam Strandberg, Brian Finrow, Alexey Guzey and Sasha Targ for reading this.

Cite this essay:
McCarty, N. "The Laws of Science" newscience.org. 2022 August. https://doi.org/10.56416/721mpl

Without more context, you’d be forgiven for thinking that the organization about to “break in” its staff members was a Silicon Valley tech company, perhaps, or a cult. But you’d be wrong — It’s an excerpt from a 1946 How-To guide for new program officers in the Rockefeller Foundation’s Natural Sciences Division.

During the 1930s-1950s, this division funded the revolution in molecular biology; the very term was first coined by then-director, Warren Weaver, in the Rockefeller Foundation’s 1938 Annual Report. In the dozen years after the discovery of DNA’s structure, in 1953, all but one of the 18 scientists who received a Nobel Prize for genetic molecular biology research had been supported by Weaver’s team at the Rockefeller Foundation (Foreword, Science and Imagination: Selected Papers of Warren Weaver).

If the Rockefeller’s Natural Sciences Division shaped and enabled the midcentury revolution in molecular biology, then studying their work is instructive for contemporary funding organizations aiming to similarly accelerate modern science.

Weaver’s team funded Linus Pauling and his development of a general theory of protein structure. They funded George Beadle and Edward Tatum, who discovered that genes act by regulating definite chemical events while studying the fungus, Neurospora, and for which they shared a Nobel Prize in 1958. And grants made to Salvador Luria and Max Delbrück helped them to demonstrate Darwin’s theory of natural selection in bacteria, for which they shared a Nobel Prize in 1969. Weaver’s team steered the cross-disciplinary application of then-recent advances in physics, chemistry, and mathematics to produce seminal discoveries in biochemistry, molecular genetics, and immunochemistry.

Notably, they didn’t achieve their funding success through the now-familiar approach of decentralized decision-making by a sprawling bureaucracy of study sections, program officers, and peer review. Weaver saw himself, instead, as a “manager of science.” He led his small team of program officers in a highly opinionated grantmaking process that focused on the organizational and social environments of research institutions over the specifics of any individual project. And once his team selected a field to focus on, it funded people over specific ideas.

Somehow, Weaver’s team did all of this with relatively little domain knowledge. Weaver’s background was in physics, and the program officers that he hired — and who crisscrossed the world and recommended scientists for grants — were rarely the top scientists in their field. At one point, Weaver even deflected an employment inquiry from Leo Szilard.

Weaver also avoided the temptation to exert his influence by insourcing funded research, as was commonplace at The Rockefeller Institute for Medical Research (now Rockefeller University). This allowed the Natural Sciences Division to back out of its grantmaking role once new fields and institutions had established themselves in their own right.

In this essay, I draw from the Rockefeller Foundation’s archives, particularly its program officer diaries, and other sources to explain how Weaver’s Division accomplished these remarkable feats, and what modern science philanthropy can learn from their success. I discuss Weaver’s background and role, and dissect his approach for building relationships with the scientific community; special emphasis is given to Linus Pauling and Max Delbrück.

Later, I explain how Weaver scaled his impact by hiring program officers and maintaining a harmonious relationship with the Rockefeller Foundation trustees, who controlled the purse strings. The essay ends with lessons for contemporary philanthropies with similar goals, including Arcadia Science, Convergent Research, and New Science.

But first, I highlight Weaver’s key Funding Principles; those fundamental things that made his team so successful in its day.

Subscribe now

Funding Principles

Several crucial takeaways from the Rockefeller Foundation, and their funding of molecular biology, are directly applicable to modern, life sciences grant-making organizations. Many of these takeaways mitigate the second- and third-order effects of grant-making, such as the ways in which grant recipients might adversely alter their actions in response to funding, or the potential of future funds.

1. Triangulate to excellence: Weaver’s officers lacked domain expertise across the ever-shifting research areas they funded, and so they developed heuristics to find researchers worthy of funding.

   1. This is the most counterintuitive, yet useful, guiding principle. It is still applicable today. Science funding organizations that want to back high-caliber researchers, according to their own criteria, can efficiently locate them by repeatedly asking the right questions.

   2. What to do: Natural Sciences officers would proactively ask independent experts in a given (sub)field, “Who is the best scientist studying X?”

   3. What not to do: Natural Sciences officers were instructed not to reactively (in response to a grant application) inquire, “Is scientist Y a good candidate to receive funding to study X?”

   4. Pros:

      1. Natural Sciences officers could move between subfields while incurring minimal switching costs. This allowed Weaver and his officers to adapt to the Rockefeller trustees’ shifting interests in different areas of research.

      2. Weaver could recruit Natural Sciences officers without respect to their preexisting domain expertise, giving him flexibility to optimize along other axes instead.

   5. Cons:

      1. Researchers who were reclusive or not highly regarded by their peers (for example, due to having non-consensus scientific views) were probably less likely to receive funding compared to their gregarious peers.

      2. If one funds a specific subfield for extended periods of time, it is probably more efficient to develop a technical understanding of the space.

2. Hire foxes, not hedgehogs: If you asked anyone today who Harry Milton Miller, Jr., Gerard Pomerat, or Farnie Loomis were, they wouldn’t be able to tell you. But these are the men who identified and funded so many Nobel Prize-winning scientists.

   1. What to do: Hire scientific foxes as your program officers. They will triangulate to success.

   2. What not to do: Hire successful scientists as your program officers. Leo Szilard was rejected for a Natural Sciences officer position by Weaver in 1952.

   3. Pro: Scientific foxes can be found and recruited from the same academic environment inhabited by grant recipients, thus allowing each program officer trip to double as a recruiting trip.

   4. Con: Whereas the Rockefeller name was sufficient for program officers to stand out amongst scientists in Weaver’s day, a name of such singular prominence does not exist today. Having program officers of some scientific fame may be helpful for modern organizations.

3. Avoid creating long-term dependence: Weaver’s team took great pains to ensure that its grants did not make researchers dependent upon, or come to expect, continued Rockefeller funding.

   1. For any contemporary private science funding organization, this is a critical principle of practical import. Large quantities of government funding are unlocked for fields of scientific inquiry once they become commonly accepted as worthwhile of study. Therefore, driving underfunded areas of inquiry into the mainstream of science is key to alternative funding efficiency.

   2. What to do: Make grants contingent on the researcher’s demonstrated ability to obtain matching funds from third parties. This policy tests whether they will eventually be able to survive on their own.

      1. Embrace relationship-driven grant-making, as long as it helps to eliminate indefinite dependence on Rockefeller Foundation grants. Back scientists and research that would entice other benefactors to provide matching and subsequent funding.

      2. This very principle was included in Natural Sciences Notes on Officers’ Techniques (1946), and was emphasized to newly initiated program officers.

   3. What not to do:

      1. Give grants that cover the majority of a researcher’s salary. It isn’t easy to stop paying someone.

      2. Allow grant recipients to designate junior personnel with the title of “Rockefeller Foundation Assistant” or similar.

4. “Make the peaks higher”: Carrying on a philanthropic principle established by John D. Rockefeller himself (see: The Story of the Rockefeller Foundation), Weaver’s team sought to reinforce naturally existing advantages amongst scientific institutions and investigators, so as to maximize the chance — and speed — that Natural Sciences grants would catalyze important advances.

   1. This principle is a useful corollary to the first principle. Assuming that you have already found the scientists you want to fund, ensuring that their environment is also ideal makes their work more likely to succeed.

   2. What to do: Create concentrations of researchers at specific institutions, rather than encouraging or allowing them to fan out.

      1. Ex: Natural Sciences provided funding to Caltech in the late 1930s to build on Linus Pauling’s success by recruiting Thomas Hunt Morgan (from Columbia) and George Beadle (from Stanford).

   3. What not to do: Natural Sciences did not provide much funding for biology research at Harvard, despite it being favored by the Rockefeller Foundation, because much of biology there sat in the shadow of Harvard’s dominant medical school.

   4. Pro:

      1. By incentivizing collaborations between scientists at a single institution, such as Caltech, Weaver’s team implicitly lowered the barriers to new, out-of-the-box projects. This was particularly important, at the time, because long-distance communication was limited.

   5. Cons:

      1. Grants for individual projects, the most scalable and repeatable vehicle of funding that Natural Sciences possessed, were not very useful for institution-building. Rather than delegating to his officers, Weaver was often personally involved in major grants.

         1. Ex: The $350,000 grant to Caltech in 1936 for a center for molecular biology.

5. Stick to a fairly informal application process: Natural Sciences had no application forms for grants. Applicants were encouraged to organize their applications in whatever format made the most sense for them.

   1. Although informal grant application processes remain unusual in science funding, this approach is widely used in other domains, such as by the tech startup accelerator, Y Combinator.

   2. What to do: Natural Sciences officers often gave situational advice to specific researchers on what grant applications would be most likely to receive funding.

      1. Example: Program officer, Frank B. Hanson, told Linus Pauling in 1934 that a grant application for x-ray crystallography of biological molecules would receive funding, while a similar application for inorganic molecules would not.

   3. What not to do: Natural Sciences never required researchers to adhere to formalities in their applications

      1. Contemporary example: the font size requirements of the NIH.

   4. Pros:

      1. Preordained formats for applications would have constrained creativity and added friction for researchers, giving an advantage to academic bureaucrats over true scientists.

      2. There were fewer opportunities to game the funding rules, particularly when Natural Sciences’ focus shifted from year to year.

      3. In today’s scientific era — which is dominated by academic bureaucrats skilled at responding to hyper-specialized Funding Opportunity Announcements from institutions like the NIH — nimbler organizations have a massive advantage in funding high-value research proposals that inevitably slip through the gaps of government bureaucracy.

   5. Con: This approach may not scale very well, as it is dependent on program officers skillfully reviewing informal ideas.

      1. Natural Sciences funded $1.9M of experimental biology research in 1939, equivalent to $41M today. The NIH today funds 1000x that, or $40B/year.

6. Minimize the influence of salesmanship: Although no specific format or content was required for Natural Sciences grant applications, its officers were warned not to be drawn in by scientific salesmanship.

   1. Although this seems logical, this is the hardest principle to enforce, since it requires constant vigilance and self-awareness.

   2. What to do:

      1. Enforce some form of a written application requirement.

      2. Ignore verbal pleas in lieu of reviewing those applications.

   3. What not to do: 

      1. Get taken in by a charismatic scientist.

         1. This occurred to Weaver with Linus Pauling at times, culminating in Natural Sciences funding in vitro antibody production experiments in the early 1940s that proved to be shoddy science. Ultimately, Natural Sciences officer Frank B. Hanson cut Pauling’s funding.

To understand how these principles came to be, it’s important to understand the history of the Rockefeller Foundation’s Natural Sciences Division; where Weaver came from, what he looked for in program officers, and how he shaped his management philosophy.

Share New Science

Molecular Missionaries: A Brief History

Weaver’s Early Days

That so many breakthroughs in molecular biology were funded by grants emanating from within the four walls of the Rockefeller Foundation’s New York buildings was far from preordained when Weaver was named Natural Sciences director in 1932. 

The numerous Rockefeller charities had just exited a tempestuous period of consolidation. Trustee Raymond Fosdick (future Rockefeller Foundation president and confidante of John D. Rockefeller Jr.) had pushed the organization away from its prior focus on societal initiatives, such as improving public health and promoting education. There had long been a secondary focus of funding academic research at institutions, but only with minimal involvement in the specifics. 

Fosdick reconstituted the Rockefeller Foundation and pushed it to, instead, focus on funding the creation, diffusion, and application of scientific knowledge. Fosdick’s ultimate goal, as a Progressive social reformer, was to solve societal challenges by advancing science (see: Chapter 9 of Partners in Science).

Having spent almost his entire adult life immersed in the land-grant progressivism of the University of Wisconsin, Weaver wholeheartedly embraced Fosdick’s agenda upon his arrival. Weaver prepared a detailed program in natural and medical sciences, together with the Medical Sciences director, Alan Gregg (see Table 10-2). Weaver’s program largely adhered to Fosdick’s Progressive agenda, avoiding “science for science’s sake” (Chapter 10 of Partners in Science). A focal point of Weaver’s plan was “psychobiology” (now known as behavioral neuroscience), a field that attempted to unearth the physiological underpinnings of psychiatric and psychological phenomena.

However, Fosdick’s wrenching changes to Rockefeller philanthropy were not universally accepted by the trustees. Internecine strife persisted well into Weaver’s tenure, and placed his new proposals on shaky ground.

In late 1933, when former Rockefeller trustee, Simon Flexner, was tasked with reviewing the joint Natural Sciences/Medical Sciences program proposal, he took it as an opportunity to attack the very concept of “managing” science, and Weaver in particular. Flexner was a close confidante of John D. Rockefeller, Jr., a respected physician-scientist, and a former professor at the University of Pennsylvania. Weaver, by contrast, was a novice in life science research, and a newcomer to the Rockefeller Foundation and its hegemonic, Northeast social circles.

A Midwestern mathematical physicist who had never stepped foot in New York City before interviewing for the Rockefeller Foundation job, Weaver was now endeavoring to put his thumb on the scale of scientific inquiry by funding specific avenues of research that he deemed most worthy of philanthropic support. In particular, Weaver saw an opportunity to “create a new science of Man” by applying quantitative methods to biology, the study of which had, until then, been largely descriptive (see: Chapter 10 of Partners in Science).

However, Weaver’s vision for, and evident skill in, cross-disciplinary research administration was not hastily formed upon his arrival at the Rockefeller Foundation. In the years immediately preceding his move to New York, Weaver had been involved in an attempt at the University of Wisconsin to create an interdisciplinary geophysics research center that pulled together several departments. This small-scale experience in the management of cross-disciplinary academic research proved fortuitous for Weaver when he received a call from Rockefeller Foundation President Max Mason to join him at the philanthropic giant.

Upon arriving in New York, Weaver worked with Fosdick to put his ideas to paper. But Weaver’s initial reports to the trustees hurt, rather than helped, his case. Weaver wanted the Rockefeller Foundation to play a still-controversial active role in planning and directing scientific research.

Fortunately for Weaver and his proposal, the tide had turned in broader society by the 1930s. The hardships of the Great Depression had highlighted the limits of the laissez-faire thinking of the 1920s, bringing to the fore a generation of New Dealers who believed they could manage society to superior outcomes (see: Scientists in the new deal: A pre-war episode in the relations between science and government in the United States).

When Fosdick consulted David Edsall, dean of Harvard Medical School and a member of the Rockefeller Foundation board’s executive committee, for thoughts on Weaver’s program, Edsall strongly supported both the cross-disciplinary bent and the active scientific management to be performed by Rockefeller Foundation officers. 

Max Mason brought prominent natural scientists, Frank R. Lillie and Walter B. Cannon, onto the Rockefeller Foundation’s Committee of Appraisal. This committee had been formed by Fosdick a year earlier to provide oversight for Foundation officers and their programs on behalf of the trustees, who felt poorly equipped to evaluate the new scientific focus areas, compared to the Foundation’s earlier focus on public health and educational initiatives. 

Both Lillie and Cannon attested to the benefits of programmatic research, allowing Weaver to fend off Flexner’s attack. When Weaver presented his five-year program proposal to the board of trustees in late 1933, he won their approval (see: Chapter 10 of Partners in Science).

Over the two decades that followed, Weaver’s division would provide more than $23M in grants as a part of their experimental biology program, ultimately funding nearly every scientist who would go on to win a Nobel Prize for research in molecular biology during the 1950s and 1960s.

Pauling and Caltech

Weaver wasted no time putting his agenda into practice. He swiftly funded Linus Pauling’s x-ray crystallography work at Caltech, and pushed Pauling to direct his methods toward the biological world. 

Pauling had begun x-ray diffraction studies of crystals in 1922, when he first arrived at Caltech as a PhD student. By 1932, Pauling extended his techniques to deduce the molecular structure of organic compounds, such as urea and carboxylic acids, and to demonstrate the resonance of amide groups. Much of this pioneering work relied on Rockefeller Foundation funding provided before Weaver’s arrival at the philanthropy (Brief Account of Research in Chemistry Supported by Grant from the Rockefeller Foundation).

It was during his visit to Caltech in October 1933 that Weaver first introduced Pauling to the new Natural Sciences agenda. Pauling responded by writing up a grant proposal and sending it to Weaver’s team. In that proposal, Pauling laid out his plans to apply x-ray crystallography to biological molecules, such as hemoglobin and chlorophyll. 

Weaver funded the proposal in the form of $10K for the following academic year. Eschewing a multi-year grant for the time being, the Foundation first wanted Pauling to demonstrate his commitment to Natural Sciences’ new program of cross-disciplinary research (see: The Molecular Vision of Life, Chapter 5).

Giving Pauling only 12 months of runway appears particularly prescient in hindsight. He moved with alacrity to study the structure of hemoglobin, a protein still poorly understood despite its central importance to life and medicine. Within eight months, Pauling had discovered that the chemical formulas reported in the literature for hemoglobin were incorrect.

Early results in hand, Pauling moved in September 1934 to expand the Rockefeller Foundation’s support for his research program, which encompassed not only the x-ray crystallography hemoglobin research, but also parallel pursuits to elucidate the structures of inorganic compounds. However, Weaver’s division was now focused solely on advancing “the science of Man,” and so he pushed Pauling to pursue research that Weaver himself felt to be most important. 

At the same time, Weaver knew it would be disruptive to simply cut funding off for Pauling’s ongoing, inorganic research agenda. The goal, as with all Natural Sciences grants at the time, was to enable preliminary work that could later be expanded into standalone institutes and research fields. Weaver wanted to wean Pauling off of Rockefeller Foundation support, and so rejected Pauling’s request and tasked him with finding an alternative source of funding (Pauling to Weaver, Sept 1934).

Weaver soon found a way to simultaneously achieve all of his goals. Natural Sciences officer Frank B. Hanson visited Pauling in November 1934. The two men came to an agreement: Pauling would try to convince Caltech’s President, Robert Millikan, to provide $5K each year (of Caltech funds) to support his inorganic research agenda. In return for that commitment, Weaver’s team would provide $10K/year for Pauling’s organic research. Moreover, the longer the commitment that Pauling could extract from Caltech, the longer the commitment from Weaver (FBH 1934 diary). 

The following week, Weaver sent a letter to Pauling and notified him that an extended commitment from the Rockefeller Foundation was contingent on the Caltech contribution (Weaver to Pauling, Nov 23 1934).

Pauling soon persuaded Millikan, who agreed to a three-year commitment. With funds in hand, Pauling responded to Weaver’s letter the same day it arrived and told him about Caltech’s support (Pauling to Weaver, Nov 1934). This led to a three-year (1935-1938) Natural Sciences grant renewal for Pauling’s organic research agenda, at the agreed-upon $10K/year.

This longer-term funding enabled Pauling to continue his hemoglobin studies in the laboratory. During this time, he developed magnetic methods capable of measuring free oxygen with greater sensitivity. Using those methods, Pauling empirically demonstrated that hemoglobin oxygenation involved the formation of covalent (not ionic) bonds, in contravention to what might have been expected from theory. The application of physical science techniques to the interrogation of living systems, as Weaver and Fosdick had sought, was beginning to pay off. 

By the summer of 1935, Pauling was applying the Gouy method to ascertain the magnetic susceptibility of hemocyanin obtained from keyhole limpets at Caltech’s marine laboratory in Newport Beach. That work led to Pauling’s landmark 1936 publication on the magnetic susceptibility of hemoglobin, in which he distinguished between arterial (oxygenated) and venous (deoxygenated) blood. Those discoveries would eventually lead to the first ever characterization of the molecular basis of a genetic disease, sickle-cell anemia, in 1949 (Sickle Cell Anemia, a Molecular Disease).

But word of Pauling’s progress reached Natural Sciences long before any of it had been published. Pauling visited Weaver in New York in April 1935 and shared updates on his research progress. While there, he piqued the attention of leading protein chemist, Alfred E. Mirsky (of the Rockefeller Institute for Medical Research), who had previously studied the reversibility of protein denaturation.

A few months after Pauling’s visit to New York, Mirsky took a sabbatical to join Pauling in California for the 1935-1936 academic year. During that time, they jointly produced a general theory of protein structure that included the crucial role of hydrogen bonds in stabilizing the tertiary structure of polypeptides (see: The Molecular Vision of Life, Chapter 5).

Why begin a history of Weaver’s tenure, and the rise of molecular biology, with Pauling? For two reasons. 

Pauling’s work at the intersection of physics, chemistry and biology laid much of the groundwork for later advances in molecular biology, including the discovery of the structure of the DNA helix and proteins. And because Weaver’s investments in Pauling were so immediately impactful, their success amplified contributions to scientific research by sources outside of the Rockefeller Foundation.

Riding Pauling’s string of success, the Caltech chemistry department chair, Arthur Noyes, persuaded local steel magnate, Edward Crellin, to provide $350,000 for the establishment of a center for molecular biology in 1936. What started as a one-time $10K commitment for molecular biology research had quickly expanded by more than an order of magnitude.

Crellin’s funds, and the state-of-the-art facilities they enabled, played a key role in the recruitment of luminaries such as Thomas Hunt Morgan, Max Delbrück, and George Beadle to Caltech. Although each of those scientists was impressive in his own right, it was evident that Rockefeller’s Natural Sciences team was conscious of the broader institutional milieu inhabited by their grantees.

Leave a comment

Delbrück’s Phage

Caltech provided an ideal environment for institution-building. Compared to other leading research institutions, such as Harvard, Columbia, or Stanford, Caltech has no medical school. This meant that life science research at Caltech could stand on its own — basic research for the sake of advancing science — rather than exist in the shadow of medicine. 

Additionally, as a young institution, Caltech benefited from a distinct lack of competing, disciplinary traditions. Its closest competitor, the University of Chicago, also enjoyed robust Natural Sciences support in the early 1930s — biochemist F. C. Koch, geneticist Sewall Wright, and biologist Paul Weiss were all funded by Weaver’s team. However, by the late 1930s, many of Chicago’s prominent life sciences faculty members were approaching retirement, and the university was undergoing administrative reshufflings that would impact new faculty appointments and orient its life sciences research towards the clinic. Weaver’s team soon focused their grants elsewhere.

The lack of traditions at Caltech allowed its faculty to establish new ones. Max Delbrück’s phage group taught summer courses on physics- and math-focused approaches to biology to young scientists at Cold Spring Harbor from 1945 to 1970. Their teachings spread a ‘new’ kind of quantitative biology to institutions around the world (The Cold Spring Harbor Phage Course (1945-1970): a 50th anniversary remembrance).

However, Delbrück had been in much humbler circumstances when the Natural Sciences Division first sought to invest in him. A mediocre theoretical physicist who had failed the oral examination for his PhD at the University of Göttingen on his first try in 1929, he did not have a particularly bright future in physics. As contemporaries like George Gamow, Lev Landau, and Eugene Wigner were publishing notable research, Delbrück bemoaned his lack of success (Ordinary Geniuses, Chapter 6). 

However, good fortune struck when Delbrück received his first Rockefeller Foundation grant in 1931, which he used to study with the famed theoretical physicist, Niels Bohr, in Copenhagen. Although Bohr was a physicist by training, he harbored a deep interest in biology, particularly the relationship of biological phenomena to the physicochemical phenomena that had been unraveled in recent decades. During his fellowship in Copenhagen, Delbrück recognized an opportunity to tread new scientific territory (Creating a Physical Biology).

Upon completing the Rockefeller Foundation fellowship with Bohr, Delbrück faced a pivotal decision: either become famed physicist Wolfgang Pauli’s research assistant in Zürich, and dedicate all of his energy to physics, or return to Berlin and join the Kaiser Wilhelm Institute for Chemistry. Although the latter path would not involve working with a luminary like Pauli, it would present a unique opportunity to freely collaborate with then-leading biologists in Europe, including geneticist Nikolai Timoféeff-Ressovsky and radiation biologist Karl G. Zimmer. 

Delbrück chose the latter path and wrote to Bohr: “I have accepted Lise Meitner’s offer to go to Dahlem [in Berlin] as her ‘family theorist’ largely because of the neighborhood of the very fine Kaiser Wilhelm Institut für Biologie” (Ordinary Geniuses, Chapter 8).

Delbrück was soon making crucial advances in biology. For his seminal 1935 paper, “On the Nature of Gene Mutation and Gene Structure,” Delbrück induced mutations in bacteria and Drosophila with radiation. The paper was co-authored with Timoféeff-Ressovsky and Zimmer. And though its conclusions, largely, did not stand the test of time, it set the precedent for a model of cross-disciplinary collaboration between the physical and life sciences that was essentially unheard of at the time (Creating a Physical Biology).

When program officer Harry Milton Miller, Jr., a globetrotting parasitologist from Baltimore who oversaw Natural Sciences’ international fellowship program for three decades, visited Bohr in January 1935, Bohr requested that Natural Sciences fund a second fellowship for Delbrück to return to Copenhagen. Although Miller deflected the request for funds (Bohr, evidently, could have funded the fellowship for Delbrück himself), Bohr’s admiration towards the younger scientist was clearly well-known to Rockefeller Foundation officers (see: HMM diary, first half of 1935).

Thus, when Miller arrived at the Kaiser Wilhelm Institute in October 1936 to find Delbrück reading a book on biomathematics by the English statistician and biologist, R.A. Fisher, he sensed an opportunity to once again further the Natural Sciences program agenda through Delbrück.

Miller immediately suggested a second Rockefeller Fellowship for Delbrück, this time to study with Fisher in London. Delbrück was receptive to the proposal but, after thinking it over, instead suggested to Miller that the fellowship be with famed geneticist T.H. Morgan, at Caltech. Miller accepted (see: Interview with Max Delbruck, 1978). For Natural Sciences, the specifics didn’t matter so much as drawing Delbrück further into the Rockefeller orbit.

What started as a fellowship at Caltech for Delbrück turned into a close association with the Natural Sciences Division for the rest of his professional career. When Weaver checked in with Morgan in February 1939, Morgan spoke highly of Delbrück, describing him as a physicist who “really understands biological problems” (WW diary, 1939-1940). 

Delbrück, for his part, told Weaver that he had no desire to return to Europe, which was inching ever closer toward war. With the Caltech fellowship expiring later that year, Weaver took responsibility for finding a new academic home in the U.S. for Delbrück. In September, whilst paying a visit to Chancellor O. C. Carmichael at Vanderbilt University, Weaver discovered that a physics faculty vacancy through retirement was only a few years out. Weaver obtained an agreement from Carmichael to make that faculty position immediately available for Delbrück. In return, Weaver’s team gave Vanderbilt University interim funding to support Delbrück’s appointment.

In classic Natural Sciences fashion, Weaver had engineered a desirable outcome without incurring an indefinite financial commitment (WW diary: 1939 - 1940). Sure enough, Delbrück joined the faculty at Vanderbilt the following year. While there, he collaborated with Salvador Luria, leading to their 1969 Nobel Prize in Physiology or Medicine for “discoveries concerning the replication mechanism and the genetic structure of viruses.”

Program Officers

The Rockefeller Foundation’s desire to stay in the background as Natural Sciences grant recipients went on to reshape humanity’s understanding of living systems means that, to this day, surprisingly little is publicly known about the peripatetic officers who made it all possible. A deeper understanding of where they came from and what drove them will be helpful for modern grant-making organizations seeking to emulate facets of the Rockefeller model.

Natural Sciences program officers had an appealing job. Their grants made important contributions to the advancement of science, the Rockefeller Foundation was an extremely prestigious employer, and program officers traveled frequently and flitted amongst leading scientists.

Many of the scientists with whom Weaver came into contact through the course of his work asked about joining him, and this was precisely how Weaver sourced most of his officers. Although it was unrealistic (and unnecessary) to expect an officer to possess domain expertise in any specific research areas, the most important criteria for a new hire were: broad familiarity with the topography of contemporary scientific research, and a deep understanding of the mores of the scientific community (see: How Knowledge Moves, Chapter 9).

Successful program officers required a moderate depth of understanding across a dizzying variety of fields, too. During his first five years in New York, Weaver taught himself genetics, cellular physiology, organic chemistry, biochemistry, and developmental biology (Scene of Change: A Lifetime in American Science by Weaver).

And, much like Weaver, none of the Natural Sciences program officers were well-known or had distinguished scientific careers. The unique combination of scientific and social aptitude possessed by a skilled program officer made them much more like foxes than hedgehogs (Rockefeller Bureaucracy and Circumknowing Science in the Mid-Twentieth Century). As Freeman Dyson aptly pointed out many decades later, scientific progress is dependent on these two types of scientists playing complementary roles (A Many-Colored Glass: Reflections on the Place of Life in the Universe, 2007). 

When prolific physicist, biologist, and inventor Leo Szilard inquired with Weaver about a program officer job in March 1952, the answer was no—other scientists had already been selected (Leo Szilard Papers, Warren Weaver letters). Although they might not have been as successful as Szilard as scientists, they were much better fits for this particular role.

Other Natural Sciences officers included:

• William Farnsworth “Farnie” Loomis, who came from an illustrious, medico-scientific family. He was the great-grandson of Alfred Lebbeus Loomis, who established one of the first clinical labs in the U.S. to diagnose illnesses of microbiological origins, and was the son of Alfred Lee Loomis, who developed the LORAN system for long-range, aerial navigation. Farnie taught himself biochemistry from scraps of textbooks mailed to him by his mother while serving as an OSS physician during WW2, where he parachuted into China behind Japanese lines (Hydra: Research Methods). It was during Farnie’s stint as a research associate in Nobel Prize winner Fritz Lipmann’s lab, in 1949, that Lipmann recommended him to Weaver for a job (see: WW 1949 Officer’s Diary).

• Gerard Roland Pomerat also came from a medico-scientific family. Born in 1901 in Massachusetts, he was the son of a medical doctor, Charles Marius Pomerat, who originally hailed from France. After completing a PhD in cytology at Harvard in 1940, Gerard taught at the university until joining the Office of Scientific Research and Development (OSRD) during WW2 (source: Des savants aux chercheurs). His brother, Charles Marc Pomerat, was a zoologist and Rockefeller Foundation Traveling Fellow in 1937-1938 (source: American Men of Science), during which he worked in Buenos Aires with Nobel Prize-winning physiologist, Bernardo Houssay (source: FBH diary, 1937). As OSRD wound down after WW2, Gerard was recruited by Weaver as a Natural Sciences officer to help revitalize postwar European science. Gerard put the French he had picked up from his parents to good use. As with Loomis, Gerard was not a brilliant scientific mind in his own right, but his background and life experiences helped him to engage with scientists and drive Natural Sciences funding activities.

Trustees & Oversight

While Natural Sciences officers chose specific grants to recommend for funding, final decisions were made by the Rockefeller Foundation trustees, and not by the officers or by Weaver.

Rockefeller Foundation trustees, drawn from the leadership of government and industry, rarely had scientific expertise (Simon Flexner, Weaver’s nemesis, was a notable exception). John Foster Dulles was a typical Foundation trustee. A partner at the white shoe law firm, Sullivan and Cromwell, Dulles served on the Rockefeller Foundation board from 1935 to 1952, and then left to become Secretary of State. Trustees Henry Allen Moe and Raymond Fosdick similarly possessed legal and/or political experience (The Story of the Rockefeller Foundation).

Prior to the late 1920s, the Rockefeller Foundation primarily funded specific social and public health initiatives, which the trustees were well-placed to evaluate. But even after the Foundation’s grant-making focus shifted in the 1920s to scientific research, the trustees’ ideological and economic positions continued to influence Natural Sciences funding policies, sometimes going so far as to impact specific grant-making decisions. As such, trustee approval was always top of mind for Weaver’s program officers (The Molecular Vision of Life: Caltech, the Rockefeller Foundation, and the Rise of the New Biology, p10).

Deep involvement by the trustees had some advantages; a significant number of them held positions in university administration (The Molecular Vision of Life: Caltech, the Rockefeller Foundation, and the Rise of the New Biology, p6). Harold W. Dodds, for instance, was Princeton President from 1933 to 1957 and a Rockefeller Foundation trustee from 1936 to 1955, while Ernest M. Hopkins was Dartmouth President from 1916 to 1945 and a trustee from 1928 to 1942. Such overlaps in responsibility unquestionably assisted Weaver’s program officers in their efforts to triangulate toward scientific excellence.

That triangulation frequently involved independent, third party assessments, as well. In a sense, program officers conducted a sort of “rolling peer review” of their grantees, and focused much more on holistically evaluating scientists rather than any specific publications. In cases of unequivocally negative feedback, officers did not hesitate to pull funding, as they were always cognizant of the trustees’ watchful eyes. 

For example, Natural Sciences funding for Linus Pauling’s immunochemistry research was largely cut in the early 1940s after program officer, Frank Blair Hanson, identified problems with Pauling’s work (FBH diary, 1943; FBH to Pauling, May 1943), which claimed to have resulted in the first ever in vitro manufacture of antibodies, to Pneumococcus polysaccharides (Pauling and Campbell, 1942). 

Hanson solicited the opinions of immunologist Karl Landsteiner and bacteriologist Oswald Avery, while Miller consulted immunologist Michael Heidelberger. Although reluctant to openly disavow Pauling’s work in immunology, all scientists cast doubt on its reproducibility, given Pauling had not published details of controlled experiments (The Molecular Vision of Life: Caltech, the Rockefeller Foundation, and the Rise of the New Biology, p181-182). 

Even a scientific giant such as Pauling, working in a field that was then central to the Natural Sciences research agenda, was not immune to the loss of Rockefeller Foundation funding.

Effectiveness

From its grantmaking patterns, it is clear that the Natural Sciences funding style incentivized and enabled quick progress in specific areas of inquiry where opportunities to advance the scientific frontier had gone unfunded. This approach required that Weaver’s officers adopt an opinionated and ever-shifting focus. Unlike many other scientific research and funding institutions, program officers benefited from being able to simply present themselves to grant recipients as being from “the Rockefeller Foundation”—an appellation that required no further explanation. And by belying an explicit interest in any particular field of science, Weaver’s team retained optionality to shift its focus over the decades, as it did from the late 1930s (biochemistry) to the early 1940s (immunochemistry) to the late 1940s (phage biology) to the 1950s (nucleic acids).

This characteristic distinguishes the Natural Sciences Division from the government funding agencies that mushroomed in size in the second half of the 20th century.

Knowing that Natural Sciences’ support would not last forever, scientists were incentivized to produce rapid results that they could then wield to supplant Rockefeller dollars with funding from other benefactors. Indeed, by the early 1950s, more than half of Natural Sciences dollars were directed to its program in agriculture (which played a key role in the Green Revolution), rather than its program in experimental biology.

Changing focuses and long-term impacts are not exclusive. Weaver’s team sought to ensure that their grants enabled rapid progress and would later lead to independently sustainable centers of scientific excellence. The Rockefeller team preferentially selected scientists (such as Pauling) and institutions (such as Caltech) that were adept at obtaining outside sources of funding. Equally important, scientists who sought to bring their peers, especially junior colleagues, along were rewarded, as in the case of Delbrück’s phage group.

However, these strategies could be vulnerable to exploitation by charismatic personalities, who improperly commandeered third party funds. Pauling’s in vitro antibody manufacturing claims in 1942 were a good example of this failure mode. Natural Sciences officer, W. F. Loomis, was referring to this incident (and others like it) when he later remarked that Pauling had “gone off the deep end in some cases…he has no further worlds to conquer in straight science, so why not shoot at the moon…” (WFL diary, 1951)

Post-War Evolution

Despite the preeminence of the Rockefeller Foundation in mid-century elite circles and the success of its Natural Sciences Division, science funding in America took an entirely different turn after WW2. 

Vannevar Bush, who had led the wartime OSRD efforts to create atomic weapons through the Manhattan Project, sought to apply a similarly heavy-handed approach to the establishment of the postwar scientific order. In his famous July 1945 report to President Truman, “Science, The Endless Frontier,” Bush advocated for the creation of a new federal agency (the National Science Foundation) that would fund the creation of new scientific knowledge and the development of scientific talent—the very role that Weaver and his officers had been playing in the burgeoning field of molecular biology for the past decade.

A few months later, Weaver penned a New York Times editorial, forcefully refuting the foundational premise of Bush’s proposal—that top-down federal government funding should play a primary role in funding basic science research. Weaver drew a distinction between, on the one hand, successful, but temporary, efforts to apply basic science advances to wartime aims and, on the other hand, the process by which those underlying basic science advances were achieved in the first place. It was an ironic twist, given that Weaver had been similarly accused of attempting to exert top-down influence on basic science research, and to point it in a more societally relevant direction, by Simon Flexner just a decade earlier.

But Weaver’s pleas came too late—the wheels of government had already been resolutely set in motion. In subsequent decades, top-down funding for basic science research by the federal government came to dominate American academia, reaching heights of more than 70% of all dollars during the 1960s and 1970s. Although that percentage subsequently receded to 50%, contemporary science funding still largely occurs through NSF/NIH-style decentralized bureaucracies, rather than the shoe-leather methodology of Weaver and his band of Natural Sciences program officers.

Takeaways

As Weaver pointed out in his 1945 op-ed, industrialized science would be unlikely to rapidly yield nonlinear, basic science advances. Indeed, one of the biggest wins of government-funded life science research over the past half-century — the Human Genome Project — required 30 years (from 1990 to Jan 2022) of bureaucratic persistence to push forward a largely symbolic effort that has primarily produced indirect value through tool creation (low-cost sequencing, fluorescent dyes, and computer software for genomic analysis). 

Shining a very bright spotlight on a centrally defined and heavily funded goal, without intrinsic urgency or deadlines, inevitably caused the Human Genome Project to be plagued by political infighting and schisms, such as Craig Venter’s attempts to privately sequence the human genome (The Genome War). 

Although Venter’s efforts did push more scientists to adopt his efficient, shotgun sequencing approach, a less top-down Human Genome Project might well have fostered the intellectual freedom for alternate sequencing methods to be discovered, tested, and adopted in a less acrimonious manner.

A government-directed scientific accomplishment frequently compared to the Manhattan Project, Operation Warp Speed (OWS) was rapidly successful because it sought to achieve a clearly defined set of goals (testing, manufacturing, and distribution of COVID-19 vaccines at population scale) while operating under an extreme level of societally accepted urgency (a severe ongoing death toll from the COVID-19 pandemic). 

Instead of making new basic science advances itself, OWS rightly focused on industrializing those advances (namely mRNA vaccine technology) that had been made in prior decades by a subset of the scientific community and which had faced intense skepticism and, thus, received limited financial support from the government (A Shot to Save the World).

Is there an opportunity to apply the principles that drove the Manhattan Project and OWS to far more initiatives that are a poor fit for government facilities, university research labs, or venture-backed startups? 

The non-profit Convergent Research seems to think so—it’s convinced Schmidt Futures and the Astera Institute to fund a new type of entity, termed a Focused Research Organization (FRO). Each FRO is a scaled-down version of the Manhattan Project/OWS that receives $20-100M over a predetermined time period (usually ~5 years) to engineer a solution to a clearly defined, scientific challenge. The solutions generally take the shape of technique development or platform creation, producing public goods that can then be used by researchers (Unblock research bottlenecks with non-profit start-ups).

As Convergent, Arcadia, New Science, and others work to reshape science funding and innovation, they should remember that:

• Natural Sciences’ principles for science funding were of their time.

  • The social environment of science and science funding at that time played a key role in the heuristics developed and followed by Weaver and his team.

  • The Rockefeller Foundation’s preeminent role in midcentury philanthropy opened doors all over the world for its program officers. This facilitated the intelligence gathering activities that drove funding decisions.

• The scientific landscape has fundamentally changed since Weaver’s time.

  • Federal government funding now has an inexorable cultural influence on life sciences research and researchers (source: New Science's Report on the NIH).

  • New information and communication technologies enable far more seamless scientific communication and information dissemination.

  • A wide variety of private funders are exploring different science funding models.

• Nonetheless, many of Natural Sciences’ principles were based on basic human psychology and incentive design.

  • Although the shape of contemporary science funding organizations will undoubtedly differ from the machine that Weaver created and oversaw, the opportunity to “create a new science of Man” presents itself in much the same way Weaver envisioned when he arrived at RF nearly a century ago.

About the Author

Samir Unni is an NYC-based technologist building products that help improve human health and our understanding of living systems. Follow him at https://twitter.com/SamirUnni.

Edited by Niko McCarty

Thanks to Sasha Targ and Alexey Guzey for reading a draft of this essay.

Cite this essay:

Unni, S. “Molecular Missionaries: How the Rockefeller Foundation revolutionized biology.” newscience.org. 2022 August. https://doi.org/10.56416/480pmz

The advocates of the new science in the seventeenth century so reacted against the excesses of stylistic artistry that a reluctance to use any artistry at all seems to have prevailed ever since.

— Whitburn et al., 1978

Since the founding of the first scientific journal in 1665, there have been calls to do away with stylistic elements in favor of clarity, concision, and precision.

In 1667, Thomas Sprat urged members of the Royal Society to “reject all the amplifications, digressions, and swellings of style; to return back to the primitive purity, and shortness, when men delivered so many things, almost in an equal number of words.” Some 200 years later, Charles Darwin said much the same: “I think too much pains cannot be taken in making the style transparently clear and throwing eloquence to the dogs” (Aaronson, 1977).

Darwin and Sprat eventually got their way. Modern scientific writing is homogenous, cookie-cutter, devoid of style. But scientific papers weren’t always like this.

Writing in The Last Word On Nothing blog, science journalist Roberta Kwok explains how old articles differ from their modern counterparts:

1. Scientists used to admit when they don’t know what the hell is going on.

When philosopher Pierre Gassendi tried to capture observations of Mercury passing in front of the Sun in 1631, he was beset by doubts:

“[T]hrown into confusion, I began to think that an ordinary spot would hardly pass over that full distance in an entire day. And I was undecided indeed… I wondered if perhaps I could not have been wrong in some way about the distance measured earlier.”

2. They get excited and use italics.

In 1892, a gentleman named William Brewster observed a bird called a northern shrike attacking a meadow mouse in Massachusetts. After tussling with its prey, he wrote, “[t]he Shrike now looked up and seeing me jumped on the mouse with both feet and flew off bearing it in its claws.” 

3. They write charming descriptions.

Here’s French scientist Jean-Henri Fabre rhapsodizing about the emperor moth in his book, The Life of the Caterpillar (1916):

Who does not know the magnificent Moth, the largest in Europe, clad in maroon velvet with a necktie of white fur? The wings, with their sprinkling of grey and brown, crossed by a faint zig-zag and edged with smoky white, have in the centre a round patch, a great eye with a black pupil and a variegated iris containing successive black, white, chestnust and purple arcs.

All this to say: Scientists in the pre-modern era wrote freely, despite calls to do away with that freedom. At some point, narrative and literary styles vanished and were replaced with rigid formats and impoverished prose.  The question now is: Have we gone too far in removing artistry from scientific writing?

A Brief History

To answer this question, we must first understand how we got here — when and why did our modern writing style first evolve? To my knowledge, there has not been a rigorous, empirical analysis of this topic. Such an undertaking would be difficult because style is not amenable to a simple definition. The percentage of papers written in IMRAD format (Introduction, Methods, Results, and Discussion) over time, though, may be a useful proxy for stylistic rigidity (Figure 1).

Although these data are far from comprehensive — based on a “randomly selected sample of articles,” nearly 1,300 in total, published between 1935 and 1985 — it seems that the post-WWII era was an inflection point for IMRAD style in scientific writing. Nearly every article published in the leading medical journals was written in IMRAD format by 1975.

The scientific community swelled in size after WWII. Between 1938 and 1953, funding for basic research in the United States increased 25 times over (in inflation-adjusted, constant dollars). With more money came a surge of papers (Kaiser, 2014) (Figure 2).

Increased scientific output strained the publishing infrastructure and created pressure towards greater clarity and concision (and thus homogeneity) in scientific writing. 

“After World War II, science became big business…Money meant research, and research meant papers. And our journals were virtually overwhelmed by manuscripts pouring out of our research laboratories. What could be done in this crisis atmosphere? We look back now, and what was done makes obvious sense. The editors of the journals, themselves and working through their organizations, began insisting on tightly written manuscripts in the IMRAD format.” (Day, 1989)

The rapid rise in scientific papers was decisive in other ways, too; namely, the arrival of ubiquitous, pre-publication peer review. 

In stark contrast to the modern publishing landscape, in which most journal editors simply sit back and wait for submissions to roll in, editors in the pre-WWII era actively solicited articles from personal connections. Editors who failed to source articles took up the pen themselves to fill empty pages (Burnham, 1990). 

Given this constraint, it was difficult for editors to reject submissions for stylistic reasons or to insist on significant revisions — an author might just take their paper elsewhere. Peer review was quite rare through the 1960s. 

Of the nearly 300 papers that Albert Einstein published between 1901 and 1955, only one (a paper on gravitational waves, co-authored with Nathan Rosen and sent to the Physical Review in 1936) seems to have been peer reviewed (Kennefick, 2005). Nature did not require external pre-publication review until 1973; before that time, external referees were consulted only occasionally. Papers submitted or recommended through trusted, personal connections were often published without review. 

The massive increase in public funding that followed WWII also led to greater concerns about credibility in the scientific community. Peer review was the obvious solution to bolster that credibility, and the process soon “became a mighty public symbol of the claim that these powerful and expensive investigators of the natural world had procedures for regulating themselves and for producing consensus” (Csiszar, 2016).

But for everything that peer review did to increase scientific rigor, it also acted as a powerful mechanism to enforce style norms and push papers toward formality, clarity, and concision.

Subscribe now

Origins

It’s not hard to find scientists who think proper scientific writing should be devoid of style (I was a teacher for 6 years, and IMRAD remains the de facto style taught in schools). But there has been an undercurrent of dissenting voices dating back to the very origins of scientific publications. In 1661, four years before the first issue of the Philosophical Transactions of the Royal Society, Robert Boyle wrote: 

“And yet I approve not that dull and insipid way of writing, which is practiced by many...for though a philosopher need not be solicitous that his style should delight his reader with his floridness, yet I think he may very well be allowed to take a care that it disgust not his reader by its flatness…” 

In the last hundred years, these debates shifted in focus, and scientists debated how best to balance standardization and free expression. This is the conflict that played out in the American Psychological Association (APA) in the 1920s and 1930s:

“In the case of psychology, it was only in the late 1920s, a time when the APA was experiencing the “growing pains” of professionalization that a formal document pertaining to publication standards was drafted. This period saw a dramatic increase in membership and an ever-widening definition of the topics and methods that pertained to one calling oneself a psychologist … This led journal editors who previously only had to manage a handful of submissions from individuals they often knew first-hand to flail under the sheer quantity (and variable quality) of manuscripts on topics that the editors might not have been familiar with, and it was finally decided by the editors as a collective that something needed to be done to address these burdens…” (Sigal and Petit, 2012)

Journal editors, under the auspices of the National Research Council (NRC), gathered in 1928 to discuss the problem. They decided to send out surveys to psychologists and editors to figure out what should be done.

The surveys revealed a wide range of opinions. Many respondents recognized a need for some kind of standardization and quality control in order to deal with the large influx of manuscripts, while others vehemently disagreed with standardization of any kind.

The chairman of the New York City Committee, Harry Hollingworth, did not mince words (emphasis mine).

“I want to put myself on record by way of a minority report, to the effect that for the Research Council to bother the scientists of the country with this kind of an effort at standardization and chain-store method should be condemned. Let any one who wants to run a journal do so, in his own way and with his own personality. If we insist upon cramping his style and insisting upon arbitrary form, censorship, and the like, we may make uniform pages, but we kill the life of science. And if the scientists of the country put up much longer with this eternal array of talk meetings and subconferences, we may come to have an elaborate and endowed scientific machinery, but we will have no people of scientific caliber to run the machine. It is my private belief that the best thing the Research Council can do is to make some effort to encourage personal variation, rather than cooperative conversation about trifling topics.” (NRC Proceedings, 1928d, p. 19)

In the end, the standardizers won out. A first draft of the APA’s publication manual was completed in 1928. After years of argument and editing, a first edition was released in 1937. Standardization followed in other scientific disciplines — the Council of Biology Editors released their first style manual in 1960.

The Life of Science

Language is the connective tissue of science. Mathematics and experimentation are the heart, but words are what make it come alive.

— David Lang

What is “the life of science” articulated by Hollingworth, and why is it killed by rigidity of form?

I believe he is alluding to a certain creative vitality that can only shine through when scientists are given full freedom to express themselves. Some scientists, for example, may choose to add a dash of humor to their writing. Aside from making the reading experience more enjoyable (not something that should be underestimated given how much scientists read), humor often juxtaposes ideas or introduces a novel way of looking at something commonplace, two things that are wonderful to do when searching for new ideas. And what can be said about humor can be said of other aesthetic qualities as well; infusion of beauty (either in content or language) or emotion may lead an author to develop a new metaphor or consider some minor aspect of a phenomenon in more detail, either of which could provide the seed of a new idea or observation.

Another way in which the aesthetic deficiency in scientific writing dampens our creativity is that it makes us less likely to use metaphor in our prose. Metaphor is a particularly valuable tool for creativity because it can bridge ideas between the sciences and other domains of culture, allowing virtually any phenomenon to serve as a source of scientific inspiration. Historical examples abound.

In comparing planetary motion to musical harmony, for instance, Johannes Kepler made revolutionary discoveries in astronomy.

Something akin to poetic creativity — the search for hidden likenesses — can be traced to the roots of scientific revolution. Kepler, who believed in the ''music of the spheres'' and tried to work out the notes sounded by each planet, discovered laws that describe how the planets trace elliptical paths about the Sun. (Broad, 1983)

Harmonices Mundi (Harmony of the Worlds), the book in which Kepler derived his laws of planetary motion, devoted three of its five chapters either wholly or partially to musical themes. Even today, computational astronomers at the Simons Foundation and elsewhere use music to describe supernovae.

Darwin developed crucial insights into evolution while studying economics.

One of the most striking examples of metaphor comes from the work of Charles Darwin, who hit upon ''survival of the fittest'' in his theory of evolution… Silvan S. Schweber, a Brandeis University historian, has shown how Darwin immersed himself in the economic works of the period just prior to his discovery, imbibing the belief that society functions best when individuals are free to compete and struggle to their own advantage.

''I think most scientists falsely deny the influence of culture for totally honest reasons,'' Dr. Gould said in a recent interview. ''They are trained that way. Most scientists genuinely believe they are reading nature literally, without inspiration from politics or culture. But it's no accident that you got Darwin in the 19th century, and that it was right out of the political economy of the day.'' (Broad, 1983)

And in 1979, Stephen Jay Gould and his co-author, Richard Lewontin, used metaphors — both architectural (spandrels) and literary (Voltaire's Dr. Pangloss) — to critique prevailing evolutionary ideas. Their essay, “The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme" was described as a “rhetorical masterpiece” by evolutionary biologist, David C. Queller, and remains highly influential (cited 10,202 times as of July 2022). The essay has zero empirical data. Gould himself described it as an opinion piece.

As a final example, the metaphor between biology and machine, once useful in helping scientists to think beyond the ‘specialness’ of life, is now becoming useful precisely because of its inadequacy — see “Is the cell really a machine?” (2019) and “Living Things Are Not (20th Century) Machines: Updating Mechanism Metaphors in Light of the Modern Science of Machine Behavior” for two recent examples.

The “life of science” demands a balance between the instrumental and the aesthetic in the same way that evolution requires a balance between selection and mutation. Selection improves the average fitness of a population, but drains it of diversity and limits its potential for future adaptation. Mutation increases diversity, but lowers the average fitness of the population.

Like mutation, aesthetics are diversifying and generative, but generally harmful to clarity and concision. And like selection, restrictions on style and format may improve the average quality of our writing, but at the cost of creative potential.

Cultivating Style

My pencil and I are cleverer than I.

— Albert Einstein

Despite the importance of style and aesthetics in scientific writing, there are good reasons to think that we shouldn’t do away with norms and conventions entirely. Efficient and accurate communication is paramount, and norms do not evolve for no reason. One could reasonably argue that proper scientific papers and journals are not the place for more stylized writing — that’s the domain of blogs and popular science books.

I’ve also been discussing style and scientific publishing as if they are fully divorced from the challenges and incentives faced by scientists. The truth is that scientists often have no choice but to adopt rigid norms because they compete in a hyper-competitive job market that values papers in prestigious, peer-reviewed journals above all else. It is difficult to justify aesthetic value when it unnecessarily opens an author to criticism.

Despite these challenges, there should be more effort on the margins to encourage style in scientific articles. Existing journals could add a section to their standard article format, for instance, in which authors can write freely and speculate on their work (eLife’s new “Ideas and Speculations” section is very much in this spirit, but a perusal of 50 recent articles suggests that only a small fraction of authors decide to opt-in).

Perhaps expanded length limits would also naturally lead to more stylistic writing. The primary reason for imposing restrictive length limits — namely, the cost of producing physical documents — is no longer a valid concern for journals that are fully or primarily digital.

Excessive brevity, imposed by journal length limits, can also have detrimental effects on clarity, as Dorothy Bishop, professor of Developmental Neuropsychology at Oxford University, argues in her blog: 

“I recently read a paper that reported, all within the space of a single Results section about 2000 words long, (a) a genetic association analysis; (b) replications of the association analysis on five independent samples (c) a study of methylation patterns; (d) a gene expression study in mice; and (e) a gene expression study in human brains. The authors had done their best to squeeze in all essential detail, though some was relegated to supplemental material, but the net result was that I came away feeling as if I had been hit around the head by a baseball bat.

These recommendations, while good in theory, require top-down changes to editorial policies and publication formats, something which seems unlikely unless a broad swath of the scientific community demands it. Change seems most likely to proceed from the bottom up.

As a starting point, scientists might, as Stephen Heard suggests, “suppress the reflex telling us to question any touches of whimsy, humour, or beauty” when reviewing manuscripts, and freely announce admiration for stylized writing to editors, colleagues, and authors. 

Another possibility is to create new publishing platforms (as Arcadia Research has done) or journals. I am the co-founder of Seeds of Science, a scientific journal that explicitly encourages speculative and non-traditional writing.

Conclusion

Maybe because writing is so often the final step in a research project (when we are exasperated and ready to be done with the damn thing), scientists tend to think of it as ornamental, akin to a peacock’s tail, and not as a core metabolic process in the cycle of scientific innovation. 

This is a grave problem. The way that we write is inseparable from the way that we think, and restrictions in one necessarily lead to restrictions in the other. 

The greatest thinkers in science (and business) are often prolific authors. They write books, blogs, and copious emails to sharpen ideas. Richard Lewontin, E.O. Wilson, and Paul Graham are but three examples. Dorothy Hodgkin’s scientific correspondence and papers, stacked together, extend 25.85 meters in length. Great thinkers, in other words, write all the time.

Unfortunately, the bureaucracies of modern science have created an environment in which all research must be justified in purely instrumental terms — “importance” or “impact” relative to expenditure. Researchers are evaluated by simple measures of productivity or influence — number of papers published, citation count, and grant dollars. In such an environment, it has become exceedingly difficult for scientists to take stylistic risks in their academic writing or to devote significant amounts of time to other forms of creative writing.

Individual scientists, then, must deliberately decide to value style. They must fight for their right to be playful, to be beautiful, to be wonderful. The life of science depends on it.

Edited by Niko McCarty & Alexey Guzey.

Cite this essay:
Bacon, R. "Research Papers Used to Have Style. What Happened?" newscience.org. 2022 August. https://doi.org/10.56416/837uwh

About the Author

Roger's Bacon is a NYC-based writer (substack, twitter) and co-founder of Seeds of Science, a scientific journal specializing in speculative and non-traditional writing.

Your rejection may have nothing to do with good or bad ideas. Perhaps the grant was rejected because a figure legend “creeped into the 0.5in margin.” Or because you had “not received previous NIH grants.” (Hilarious.) Or, perhaps on the same day that your grant is rejected, you are awarded a Nobel Prize (Nice).

So chin up. Most extramural NIH grants have acceptance rates slightly higher than 20 percent. And scientists, on average, spend nearly 10 hours per week on fundraising, according to self-report survey results published in 2020. Take solace in the fact that many others are suffering, too.

What are the solutions for your misery? If the goal of grant competitions is to hand out gold medals, then perhaps it’s best to eliminate the fluff, theatrics, and competition — and just dole out money equally, or randomly, to scientists. Surely, this would leave more time for research. Or at least we could minimize application requirements — including shorter, less detailed proposals — to make it easier to submit grants in the first place. 

Both options have been proposed by serious scientists. And both could have unintended consequences. 

That’s according to a new arXiv preprint from Kyle R. Myers, a professor at Harvard Business School. Myers’ preprint explains how rejected grants could generate positive value, because the act of writing a proposal forces scientists to refine their ideas, think through experiments, or explore new directions.

Science moves forward, even for the losers.

The preprint extends prior work by Kevin Gross and Carl Bergstrom, who modeled science funding contests using auction and tournament theory. Their conclusion: As contests become more competitive — or pay lines get lower — scientists collectively spend more time writing, even though the total amount of money stays the same. Gross-Bergstrom assumes that benefits are only garnered by proposals that win money; unfunded proposals are a waste of time.

Empirical evidence supports Myers’ model.

In a 2018 survey, scientists said that “at least one third of the effort spent on applications is scientifically useful,” Myers says. And "for each additional hour scientists report spending on fundraising, they report spending 6 minutes less engaged directly on their research,” Myers says.

In other words, more fundraising does not mean less research.

In contrast, “when scientists report spending an additional hour on teaching or administrative duties, they report spending 24 minutes less on their research." (When it’s time to point fingers, look no further than your department’s dean.)

The data Myers uses to support his model is, unfortunately, not causal. They are mere associations between the amount of time scientists spend on various tasks, as reported by scientists themselves. But this is part of a larger problem in science: There is precious little empirical data about how science is actually done. 

Federal agencies should fund more randomized, controlled trials into how scientists spend their time and manage their work (a topic of a recent blog by Tyler Cowen). Incremental answers in this area could have a major impact, but we honestly have almost no data to support one grant competition structure over another.

The Myers model has obvious limits, too. Not all of the time that scientists spend writing grants is useful. Reference letters, budgets, and spruced up CVs, surely, do not advance research. There is also an upper limit for scientists in terms of time spent writing versus improvement of ideas. After a certain point, grant writing becomes less useful, and ideas stop getting better. Many grant competitions also do not provide in-depth or nuanced feedback; many failed applications are only read by a small number of people, and so the ideas within them are unlikely to spread more broadly.

“The main takeaway of my model is that it’s very difficult to assume anything about how good or bad a grant competition is. When you see raw statistics about success rates and time spent on grant writing, the takeaway is that you shouldn't infer whether that means things are going well or not,” Myers says.

For now, the Myers model is influencing how we at New Science structure our fellowship applications. We want our funding contests to be purposely useful for scientists. We believe grant-making contests should eliminate bureaucracy while encouraging deep thinking from applicants. Scientists probably don't spend enough time thinking about projects at a 10,000 foot view, and competitive grant writing forces them to do that.

Unlike the NIH, we don’t care about font sizes or formats. Our grant writing process is also collaborative; applicants don’t just scream their ideas into a nebulous void, without getting in-depth feedback. If we see promise in a proposal — regardless of its clarity or specifics — we will spend weeks working one-on-one with applicants to sharpen ideas.

And our process seems to be working. About 10 percent of applicants who were rejected from our summer fellowship later sent emails to thank us, and to say that the proposal and interviews were helpful for their science.

New Science is embracing competition and iterative feedback as fundamental tools to accelerate science. We want to identify and fund promising ideas. In-depth research proposals seem like a good way to do that, and damn the complaints.

Thanks to Alexey Guzey, Sasha Targ, Kyle Myers and Andy Halleran for reading drafts of this.

Cite this essay:
McCarty, N. "Rejected Grants Are Good for You." newscience.org. 2022 August. https://doi.org/10.56416/137juw

If you’d prefer to only receive monthly updates about New Science (like this newsletter), please update your subscription preferences on your Substack account page.

2022 Summer Fellows

New Science today announced the recipients of our inaugural summer fellowships, which will run from early June to the end of August. We’ve funded five researchers:

• Avadhoot Jadhav, Boston Children’s Hospital: Designing nanobody therapeutics.

• Diana Leung, Boston Children’s Hospital: Probing how cells share resources in changing environments.

• Katherine Xiang, Broad Institute: Mapping protein binding sites using DNA sequencing.

• Kian Faizi, Harvard Medical School: Creating a free-living chloroplast.

• Riley Stockard, UC Berkeley: Designing cells that engineer other cells with minimal human intervention.

We invest roughly $40,000 per fellow, including stipends, project costs, and computational credits. By program end, fellows will publish “Go/No-Go” proof-of-concept results. Our next program will be a 12-month long fellowship that starts in September. In a few years, we plan to fund entire laboratories outside of academia.

Read the full announcement on our website.

Head of Media

Niko McCarty has joined us as Head of Media. Niko studied synthetic biology at Imperial College London as a Fulbright Scholar and developed tools for CRISPR multiplexing. He started a Ph.D. in bioengineering at Caltech, dropped out with a masters, and then studied science journalism at New York University.

He spent the last year working as a data journalist at the Simons Foundation and edited our recent, 33,000-word NIH report (upon Niko’s first edits to the report, Alexey immediately made a decision to never do any editing whatsoever because of the shame that its quality would bring to his family for the next 3 to 10 generations).

• We are looking for essays on science; how it could be better, how it fails (with viable solutions), and why some people and places seem to be exceptionally productive. Send ideas to niko@newscience.org. 

Science Retreat

Fellows arrived in Cambridge last week. They were promptly shuttled to a lake house in Athol, Massachusetts, where they discussed research ideas and planned projects.

Day One: Fellows present slides on their project — plans, concerns, what they need help with. It took three hours to discuss five projects. Synergies quickly appeared across projects; fellows forged collaborations.

After the talks, we grilled sausages and vegetables, then ate our meals on a floating barge (which seemed stable enough, and was propelled by a solar-powered motor).

At night, we roasted marshmallows over an indoor fire. Techniques to attain optimal marshmallow fluffiness were debated (tentative consensus: 10 seconds in the microwave).

Each fellow raised concerns about their project, no matter how small — I haven’t figured out how to design these peptides. Does anyone know of a computational tool? — while others offered advice and shared resources.

Day Two: Fellows ponder how best to share intermediate results. During the program, they will share preliminary data and seek feedback from Twitter and elsewhere.

Miscellaneous

• To receive updates about future New Science programs — such as our summer and one-year fellowships — please register your interest.

Stay Frosty,
Niko & Alexey

It’s finally the time to announce New Science!!!

New Science aims to build new institutions of basic science, starting with the life sciences.

Over the next several decades, New Science will create a network of new scientific institutes pursuing basic research while not being dependent on universities, the NIH, and the rest of traditional academia and, importantly, not being dominated culturally by academia.

Our goal is not to replace universities, but to develop complementary institutions and to provide the much needed “competitive pressure” on the existing ones and to prevent their further ossification. New Science will do to science what Silicon Valley did to entrepreneurship.

New Science is a research nonprofit incorporated in Massachusetts with 501c3 status pending. The board of directors consists of Alexey Guzey, Mark Lutter, and Adam Marblestone. New Science is advised by Tessa Alexanian, Tyler Cowen, Andrew Gelman, Channabasavaiah Gurumurthy, Konrad Kording, and Tony Kulesa.

In the summer of 2022, New Science will run an in-person research fellowship in Boston for young life scientists, during which will collect preliminary data for an ambitious idea of theirs. This is inspired by Cold Spring Harbor Laboratory, which started as a place where leading molecular biologists came for the summer to hang out and work on random projects together, and which eventually housed 8 Nobel Prize winners.

The plan is to gradually increase the scope of projects and the number of people funded by New Science, eventually reaching the point where there are entire labs operating outside of traditional academia and then an entire network of new scientific institutes.

In the process, we intend to both enable researchers who would’ve been working in traditional academia to work on problems they could not work on in academia and to increase the absolute number of people who work on pushing the frontier of science, by attracting those who want to pursue basic research but would not have chosen to pursue a career in traditional academia.

New Science's 2021 plan is to:

1. Prepare to run the summer in-person research fellowship for young life scientists to work on exploratory research projects they couldn’t work on otherwise in 2022

2. Continue to dig into how exactly the structures of science work and publish the results of that research

3. Get 501c3 status and raise funding for the first year of operation

If you'd like to learn more about New Science's next steps and/or are interested in:

1. Joining New Science and helping to build the new institutions of basic life sciences

2. Supporting New Science financially

3. Taking part in the summer fellowship mentioned above as a student, mentor, organizer or otherwise

4. Or getting involved in some other way

Please reach out.

And let's make science advance one young scientist at a time, not one funeral at a time.

Fore more background thinking, check out https://newscience.org/.

Cheers,
Alexey

Thanks for bearing with me over the last few months!

New Science is a nonprofit I’m announcing in two weeks, working on building better institutions of science.

Now is your last chance to enter history as someone who supported New Science before its public announcement, so, if you are considering donating, reply to this email or schedule a call with me directly at calendly.com/guzey and we can chat about it.

New Science’s (formally, New Science Research, Inc.) 501c3 status is pending and donations - both fiat and crypto - will be retroactively tax-deductible if everything goes right and New Science gets 501c3 status by the end of the year.

Otherwise, see you in two weeks! Thinking about the structures of basic science has been my main thing for almost three years now and I’m super excited that it has finally coalesced into something with a concrete big vision, enabling me to move from just talking to people and writing to building.

Stay frosty,
Alexey

Sign up now so you don’t miss the first issue.

Subscribe now

In the meantime, tell your friends!