Want to join the conversation? Connect with us on Tribe.
Find “Founders in Arms” on Apple Podcasts, Spotify, and YouTube.
Transcript of our conversation with Hannu Rajaniemi:
00:00:00:00 - 00:00:23:17
There, was I guess that's almost almost 12 months ago now. Was a, paper from, a company called Profluent, who, published a paper where they used, one of these, these protein LLMs to basically generate new versions of Crispr that do not exist in nature but still work as genome editors.
00:00:23:19 - 00:00:43:10
So that's been kind of, I think that's something that's only, only kind of still also percolating through the industry that, now we can basically take a protein that has a function that we want and make like a completely synthetic mimic of it, that, that has a different sequence from, what is found in nature, but still does the job we want.
00:00:43:10 - 00:00:45:14
00:00:45:14 - 00:00:47:06
00:00:47:06 - 00:00:52:20
Hi everyone. Welcome to the Founders and Arms podcast with me Immad Akhund co-founder and CEO of Mercury.
00:00:52:20 - 00:01:01:20
And I'm Raj Suri, co-founder of Lima and Tribe. And today we have with us. Hannu Rajaniemi Did I get that right?
00:01:01:20 - 00:01:02:18
yes.
00:01:02:18 - 00:01:20:01
Yes. Okay, great. And, Hannu is a co-founder CEO of a biotech company called Helix Nano and also a prolific author, author of the Quantum Thieves series and many other science fiction novels. Hannu, welcome back to the show.
00:01:20:02 - 00:01:25:08
You were on when we were, the curiosity podcast, too. So actually, you are our first return guest.
00:01:25:08 - 00:01:27:20
Oh, amazing. It's my honor to to be back.
00:01:28:09 - 00:01:31:20
Yeah. We, we love talking to you, so we wanted to have you back.
00:01:31:20 - 00:01:47:23
Actually, before I met. How do I read one of his books? Maybe all of his books, actually, at least all the ones that are on audiobooks. And I love your books. So you, you've launched.com, but it's not an audiobook yet. Is a complicated to get it into an audiobook.
00:01:48:01 - 00:01:48:09
So I could
00:01:48:09 - 00:01:59:20
so it is actually it is actually available on audible in the UK. So, so people have reported people have reported success by it. You can temporarily change your geography on audible. So. So that does
00:01:59:20 - 00:02:02:20
do that. Why is it not in the US there? Is it done
00:02:02:20 - 00:02:03:20
with, like, a British voice
00:02:03:20 - 00:02:05:08
or something?
00:02:05:08 - 00:02:08:14
We're still figuring out a, a, a, us deal.
00:02:08:14 - 00:02:20:20
So, so it's basically, one one slight obstacle and slight slight spoiler is, is it's book one of two. So I think, some of us publishers, want to see their, second part or want to do, like, a two book, two book
00:02:20:20 - 00:02:22:20
Raj, have you. Have you read Hannu's books?
00:02:23:10 - 00:02:25:20
No. I actually bought one of them, but I haven't read, I read it. We
00:02:25:20 - 00:02:30:08
great. It's like super hard sci fi. I love the whole, you
00:02:30:08 - 00:02:31:20
know, sci fi, but, like,
00:02:31:20 - 00:02:42:21
obviously, Hannu is just, like, a deeply technical person. I think, like, really shows through, what have you been up to? Hello. I guess it's probably been almost a year since we.
00:02:42:23 - 00:02:58:20
Since we last talked, like, what's what's been the updates, both for you, the company and also, I guess, like how, you know, the most recent, environment, AI upgrades and all that stuff has, like, changed, what you've been kind of working on.
00:02:59:08 - 00:03:20:08
Yeah. So, quite a lot of stuff going on there. So, so Helix Nano, kind of starting from the company side, like Helix Nano has been been, running a clinical trial, which is now now fully dosed. And we have some exciting human data that, I mean, the trial is still technically ongoing, so I can't, I think, officially, disclose disclose the results, but,
00:03:20:11 - 00:03:21:20
And that's where the FDA.
00:03:21:20 - 00:03:25:16
this this is actually, related to the other the other topic you mentioned.
00:03:25:16 - 00:03:28:08
This is actually with TGA in Australia. So,
00:03:28:08 - 00:03:29:20
Okay. Interesting.
00:03:29:20 - 00:03:47:13
unaffected by the, the recent recent events at least, so far. And, yeah. No, I, I think we are we are very excited by what the outcomes are, are so far. And, yeah, I mean, a lot, a lot of work also on, on oncology.
00:03:47:15 - 00:04:31:12
And, it's kind of there, there is this interesting convergence. What's with with, the ideas are in, in dark, that you might alluded to of, of basically making the immune system digitally updatable and the kinds of things we're working on. It is, of course, also, hard to, to, escape, the, the impact of AI and the single thing I'm probably the most excited about, across like, basically everything, is, combining reasoning models with mRNA, so, so there's like, non-intuitive, but you can't unsee it once you once it sort of clicks, kind of complementarity between those, those two technologies.
00:04:31:14 - 00:04:50:15
We can we can talk more about that. That's something also we're now collaborating on with, with one of the big AI labs. And again, like, limited, limited in what I can say about the exact exact work, but, the kind of overall topic I think I can, I can talk about, so those are and, my wife and, yes.
00:04:50:17 - 00:04:56:20
Go ahead. No, my, my wife and I also had our second child. So that's probably why I'm looking more more haggard, perhaps, than
00:04:56:20 - 00:04:58:08
I was lost last time with
00:04:58:08 - 00:05:00:06
your, the podcast.
00:05:01:03 - 00:05:04:08
Immad would be like that very soon. So it's going to
00:05:04:08 - 00:05:05:13
I know I heard. Yeah.
00:05:05:13 - 00:05:06:08
way here.
00:05:06:10 - 00:05:07:01
Yeah. Just like
00:05:07:01 - 00:05:08:13
two weeks ago. Probably.
00:05:08:13 - 00:05:13:01
Yeah. Yeah. Yeah. You'll be still doing the podcast, right Immad? And you're going to be podcasting
00:05:15:00 - 00:05:15:13
I'll try,
00:05:15:13 - 00:05:17:01
I'll try.
00:05:17:01 - 00:05:18:13
I'm taking, like, a few weeks off
00:05:18:13 - 00:05:21:19
work, so I guess I'll have more time for the podcast.
00:05:21:19 - 00:05:23:01
we'll do it. We'll do a three hour podcasts.
00:05:23:01 - 00:05:24:01
I'll
00:05:24:01 - 00:05:25:13
do the Joe Rogan podcast. Yeah.
00:05:25:13 - 00:05:25:21
Yeah.
00:05:25:21 - 00:05:27:01
What
00:05:27:01 - 00:05:38:13
I want to do is remind our audience that, you know, what you do at Helix Nano, right? So it's very interesting technology. It's, you're using mRNA, vaccine technology to augment the immune system. Did I get that right?
00:05:39:01 - 00:06:12:10
Yeah. That's that's, that's that's pretty much the one liner summary. Summary. So, so mRNA is this very natural API to do human biology or all of, all of biology. And gives us the ability to really talk to the immune system in its own own language. I mean, it can I mean, you can, it can do a lot more than just vaccines, but, and I guess, like I would say, mRNA vaccines are vaccines in the same sense that computers are calculators, but that the, it is we do we do also believe in the power of the immune system as one of the most foundational parts that impacts
00:06:12:10 - 00:06:13:12
our, health.
00:06:13:12 - 00:06:20:15
Super cool. So I guess one question is. Why Australia? Why are you working with Australia? And, what makes that such a great place to work
00:06:21:05 - 00:06:42:12
So Australia does have, a very, like a top tier regulator. So the TGA is, is just as rigorous and, and, and, a disciplined and, so on as the FDA, they are, but they're leaner and foster. So, so there there's, little less red tape. To deal with.
00:06:42:14 - 00:06:52:15
And and it's yeah, it is better for, for the types of studies we're doing it. It's just like just a natural natural geography and also where we, we want to, to deploy the products
00:06:52:15 - 00:07:01:03
you. Can you quantify the difference? Like, how many years would it take you to do the same thing? And with the usfda versus the Australia, TDA, is it?
00:07:01:09 - 00:07:21:10
Yeah. I think our lawyers would probably also tell me that I should not like directly, talk about regulatory arbitrage. But the, the, the, but I mean, for us specifically, we, we probably saved like, like, about nine months, just in terms of, like, getting the trial. Trial started. So it it is quite it is quite significant.
00:07:21:12 - 00:07:36:01
And then the kind of cadence of interactions has, has overall also been being faster. I mean, we certainly certainly intend to also to, to engage, with the FDA as well for, for for this in this in other programs. But, but this was a really great way to, to get started.
00:07:36:01 - 00:07:40:01
Yeah. Let's try to think of other things we can do to get you in trouble with your lawyers.
00:07:40:01 - 00:07:41:13
What,
00:07:41:13 - 00:07:59:13
what are your general thoughts on, on the changes at the FDA? Like, yeah. A do you think there's a lot of change required? And B, do you think the kind of RFK administration will be the change agent we need?
00:08:00:17 - 00:08:04:20
Yeah. I mean, the the the here are we here are we need are the hero we deserve.
00:08:05:04 - 00:08:06:03
Yeah.
00:08:06:03 - 00:08:24:13
the, or or or the villain, you know, the the, I mean, it's obviously, every everyone is, watching very closely what's happening? For, I mean, obviously the the FDA was set up, with a particular purpose.
00:08:24:13 - 00:08:33:13
Really this, like, like, safety above safety above everything in response to the Thalidomide I can never pronounce that, Thalidomide. It's like so like
00:08:33:13 - 00:08:38:13
Thalidomide? That was like a medicine that killed newborn like babies, right?
00:08:38:13 - 00:08:46:23
and of course, horrible birth defects. Like, like really like one of the terrible, most worst disasters of of that kind of kind in history.
00:08:47:01 - 00:09:18:15
And, and that's sort of where the, the current structure of the FDA really, really flowed from. But the FDA, like even prior to the current events, has been working actually quite hard to to figure out how to adapt. And they are some extremely forward looking people, thinking people, inside the FDA and, like one, example, I'd sort of call out who I hope very strongly is kind of survives the current, current, changes is Peter Marks, who heads the cyber, branch of the FDA.
00:09:18:18 - 00:09:45:19
Who, who sort of who was also one of the chief architects of Operation Warp Speed, that developed the, the Covid vaccines. And, and he's been really thinking really hard about, how does one, for example, deal with AI? How does one how does one regulate things like personalized, mRNA vaccines for cancer, where where you might be making like a different product, for, for each individual, which furthermore is actually designed via neural network.
00:09:45:21 - 00:10:07:00
So so they are very much have been thinking about those kinds of things. And and also on the medical devices side, like figuring out frameworks where, if you have, an AI system that learns, like, how do you how do you actually, under what circumstances can you update that algorithm on your device without having to rerun, all the trials?
00:10:07:00 - 00:10:38:16
Because traditionally, the fixed product that you take through the trials is the one that you're allowed to sell. But now they are trying to figure out how to how to be more, more dynamic and adaptable. Now, obviously, there's been organizational inertia. And, and these changes take time. And now now, I think that the, certainly the kind of near-term, immediate practical impact of, like, personnel cuts at the FDA is that people are actually getting bogged down, that they're getting bogged down with, like, too many AI and filings, and, and are sort of at over overcapacity.
00:10:38:16 - 00:10:39:13
So, so,
00:10:39:13 - 00:10:41:01
Too many. What? Filings?
00:10:41:01 - 00:10:47:01
I and investigational new drug application. So that's like the first package you put in to the FDA
00:10:47:01 - 00:10:54:21
that tie into the kind of the. I think, like we're just going to have a lot more drugs and drugs people want to test. Because if I.
00:10:54:21 - 00:10:57:01
I think because of China,
00:10:57:01 - 00:10:58:13
Yeah.
00:10:58:13 - 00:11:13:13
so one one trend in the, pharmaceutical industry is that there's a lot of clones of, of, or like variant variants on a theme, drugs, coming from China that, is now about 30% of the big pharma pipelines.
00:11:13:13 - 00:11:19:01
Wow. So he's saying 30% of all these applications in to the FDA, Chinese drug
00:11:19:01 - 00:11:20:07
manufacturers.
00:11:20:07 - 00:11:38:17
not necessarily to the FDA, but like in terms of, big pharma companies filling their pipelines. And then some of those trials obviously are later stage trials are carried out, also also under the FDA. So, but I mean, I guess like we haven't seen that there's a, there's a whole topic there of, like, why, why haven't we yet seen AI impact?
00:11:38:23 - 00:12:02:15
Drug discovery, productivity. But, that that that hasn't happened yet. I'm sure. I'm sure it will, but, but that has not happened yet. But so, so I think near-term, there's going to be just like, a bit of, a bit of, like a logjam of applications and, and, companies will struggle to, to, you know, execute their, their programs in a timely, timely fashion.
00:12:02:17 - 00:12:29:16
The longer term impact is really difficult to, to foresee. Certainly the sort of R.K. junior presence has caused jitters amongst biotech investors and pharmaceutical companies around vaccines. And undeniably, the maybe the silver lining with the sort of Trump, administration engagement with, the FDA and the health industry is that, the in the previous iteration, they were very pro compassionate use.
00:12:29:17 - 00:12:49:06
So they were they were trying to figure out, how could you use, rapidly like an unapproved drug, for a patient to treat it, treat it, treat a patient who has a life threatening condition. So it's sort of like make that process very effortless and rapid, which which, you know, I think in my view, it's a good thing.
00:12:49:08 - 00:13:09:13
So I would expect, that, those kinds of pathways get accelerated, again, which, I think can also have positive impact. So, so near-term, near-term, certainly like, problematic right now, long term, long term, maybe maybe maybe it'll be useful. Creative destruction. We'll see.
00:13:10:06 - 00:13:20:01
You don't think there's, like, a good path to an an a deregulation slash? Libertarian i's the FDA. I don't know what that looks like, but,
00:13:20:01 - 00:13:21:13
I guess we don't want,
00:13:21:13 - 00:13:32:01
like, anyone just makes a drug and releases it without any sort of, like, steps. But I wonder if there's, like, a medium ground that's like, the FDA is still there, but it's, like less restrictive, goes much
00:13:32:01 - 00:13:36:08
quicker.
00:13:36:10 - 00:13:36:13
Well,
00:13:36:13 - 00:13:38:01
compassionate use is a very small
00:13:38:01 - 00:13:40:21
book, right? Yeah.
00:13:40:21 - 00:13:54:03
so I guess like something that fascinates me and which dark the novel involves heavily is actually, you know, people operating just entirely outside that framework, and doing self experiments.
00:13:54:05 - 00:14:32:00
And, one of the, the sparks for the novel, originally, which I mean, just like in one line, it's, it's it's a, a, you know, your nearest future at 20 years hence where, there is this underground network of biohackers doing experiments on themselves. There's also a giant company who who has built this sort of immune computer interface, a way to update everyone's immune systems digitally because there's just so many, both synthetic and natural pathogens around, but the main character, grows in a community where people do sort of heavy self-experimentation, has a genetic disorder that she and her mom are trying to treat, treat themselves.
00:14:32:02 - 00:14:53:13
So, I think we'll see more of that. I mean, that's already happening. So you have, one of the sparks for the novel or already a few years ago was, the realization that, just like I described these personalized cancer vaccines, people are doing DIY versions of those, so they are like, you can find instructions online, like if you have if you have a cancer,
00:14:53:18 - 00:15:03:03
Where is that? Is it legal in the US? If I wanted to try to, like, cure my own cancer. Like. And assuming I have the skills to do that.
00:15:03:03 - 00:15:06:13
I think no one can stop you from ordering those reagents and putting them together and, and
00:15:06:20 - 00:15:11:01
It could be like. I mean, there's lots of things that are illegal that maybe no one could stop me doing.
00:15:11:19 - 00:15:40:17
Yeah. I mean, I think the very great, great territory. I mean, you're not really necessarily practicing medicine without permission. I mean, I mean, Self-experimentation does does seem to be this, like an interesting, interesting loophole. And, and the, and that has featured actually in, like, Nobel Prize winning discoveries, there's a there's a famous example around gut bacteria that I'm trying to try to remember, but like, like, Self-experimentation has a long, inglorious history in the science.
00:15:40:19 - 00:15:54:13
So, Yeah. So, so I think that's that's maybe an interesting direction as the tools needed to do all this become more and more democratized. People will I think people will just bypass the system entirely.
00:15:54:13 - 00:16:05:01
Honey. I was curious. What do you thought? I saw an interview with Bill gates quite recently where he talked about, like, we're not ready at all for the next pandemic. What do you think about that?
00:16:05:01 - 00:16:06:19
we are not,
00:16:06:22 - 00:16:09:00
So where are we? Where do we need to be?
00:16:09:00 - 00:16:39:07
so I think there's there's sort of a spectrum of, of threats there. So, so there's the natural pandemics and then there's the synthetic ones. We the kind of just to kind of play out one scenario. I mean, the immediate, immediate, immediate situation is, H5n1. So avian flu, in, which which now, now is obviously in birds and causing egg sort shortages and, but but also like more worryingly, in, in, in cows, cats, other mammals.
00:16:39:09 - 00:16:56:00
And now we've seen some human cases, although not yet human to human transmission. So is and clearly I mean the virus isn't yet hasn't yet developed the mutations it needs to jump into humans and human to human to human transmission. But it does seem like a matter of time. And I think we all are unprepared for that.
00:16:56:00 - 00:17:17:23
So, so so we don't have a protests. We don't have, like, I mean, there are a couple of vaccine programs and programs in development, but, those may also be behind the actual mutational curve. And can we scale those those, sufficiently with bird flu specifically? There's a weird like there's a literal chicken and an egg problem because of lot.
00:17:18:01 - 00:17:34:12
No, seriously, because a lot of the traditional traditional, flu vaccines are produced in eggs. So, so and so now, now the virus is sort of wiping out that, that, that sort of, resource, to, to, to produce them at scale. Assuming, assuming they can of course, be updated
00:17:34:18 - 00:17:38:12
My. What do you mean? Like, we get chicken eggs and we make vaccines in them.
00:17:38:12 - 00:17:42:08
They. Yeah. So you basically incubate the virus in the egg.
00:17:42:10 - 00:18:01:11
So, so you make out so, so so like you have and then you inactivate the virus. Virus usually, usually. So these are sort of very old school vaccines. These this is sort like 1950s 1950s vaccine technology that there's obviously Moderna is working on like an mRNA candidate for H5n1. And there's this others others too. But the, the but this is also very politicized.
00:18:01:11 - 00:18:19:12
Like like the, the there's also like the obvious thing to do right now would have been to, to, vaccinate all the cows. And there's, there's, UPenn has a is like a cow H5n1 vaccine in, in development. But, the, the cattle industry actually resists it. So, so there's, there's,
00:18:19:12 - 00:18:21:00
Why do they resist it?
00:18:21:00 - 00:18:27:04
there is just a general so, so, so I think the, it's specifically actually the chicken industry that is really resistant.
00:18:27:04 - 00:19:05:03
So there's, there's like, there's a, situation. The problem is that, at least with the traditional vaccines, you might not be able to tell whether a chicken has been vaccinated or whether it's been infected and they infected one. You obviously need to need to destroy. So so then it creates problems for like, exports and, and, but I mean that that's not the case obviously with more modern vaccines, but that thinking kind of persists throughout, throughout the industry that, that, you might actually, like have to call animals that you that, that you that you vaccinate, if you can't distinguish an infection from vaccination but so, so, so I
00:19:05:03 - 00:19:34:04
think but but I think like more broadly, we, we don't really have like, any kind of global pathogen monitoring system. I think Bill gates specifically has called for this, like, like nucleic acid observatory or like a more, more like international organization whose job is to to deal with and respond to the pandemic's, I think the current situation or the current current sort of political climate, this kind of international cooperation seems to be even more, more challenging.
00:19:34:04 - 00:19:53:13
Like, I mean, one of the key things obviously, that happened during Covid was that this Chinese researcher, just like out of their own volition, sent out the sequence sequence of SARS-CoV-2. And that's that, then immediately allowed everyone to start working on working on a vaccine. So is that going to happen now? Like, are people going to, because, I mean, it didn't end well for for that, that
00:19:55:00 - 00:19:56:12
Why did. What happened to that person?
00:19:56:12 - 00:19:57:08
He Died
00:19:57:08 - 00:19:59:05
He died?
00:19:59:05 - 00:20:01:17
went, yeah. I mean, I think he was imprisoned and then then died
00:20:02:08 - 00:20:03:12
Why?
00:20:03:12 - 00:20:06:16
The Chinese government didn't want him to release this information.
00:20:06:16 - 00:20:11:22
Wow. Well, that's. He's a he's a hero. I think that is really sad.
00:20:11:22 - 00:20:28:04
yeah. I mean, there's definitely like a heroic tell tale there and so sad said one. But, the, the, so, so now and like actually actually similarly like, like, I mean, South Africa was sharing I mean not not not not, but instead of quite the, quite the same context.
00:20:28:04 - 00:20:50:05
But like South Africa during the pandemic was, was like very active sharing, information about new variants. And then then they basically like, like had economic consequences, from, from that and and the, the, so, so yeah, I didn't get the spirit of cooperation actually maybe was damaged by, by by Covid and made made the concerted response, the response harder.
00:20:50:07 - 00:21:10:12
So yeah, I, I, I'm, I'm, I mean, I'm obviously always fundamentally optimistic that we can, we can figure these, these things out and if like we have compelling enough reasons, we will. But but yeah, we certainly don't I don't feel like we've learned enough from like, like 26 million deaths didn't and like 10 trillion or whatever.
00:21:10:12 - 00:21:15:22
It was like global GDP impact did not seem to be like, like enough
00:21:15:22 - 00:21:17:10
Yeah. It's crazy. It.
00:21:17:10 - 00:21:43:23
It seems like any sort of top down globalist, kind of organizational, response to these things just is not going to work like we need to think about. Like, how do we have better, kind of bottoms up. Decentralized responses to these things, right? How do we have labs? Just like doing stuff. How do we have, you know, speed up companies abilities to make vaccines really quickly and, like, deploy them quickly?
00:21:44:02 - 00:21:46:22
Like, it just seems like the steps down things is just not going to work.
00:21:46:22 - 00:22:10:07
yeah, I don't really trust the WHO to figure this out, honestly. Like, it's, it just feel like, you know, you know, death by bureaucracy. I think this is one area where philanthropy needs to step in and, like, fund this stuff. I don't I don't see any business model for it. So maybe there is a business model, like, there's, like, some output of this kind of, you know, a global cooperation, but bottoms up.
00:22:10:07 - 00:22:27:21
Or as you say, that has like some kind of business, attribute as well. But I don't think we can trust governments to figure this out. We have to like, figure out a different path. That is, that is more like, you know, bottoms up driven. Because I've seen the same with immigration. Governments don't do anything, even though it's in our interest to bring in high skilled immigrants.
00:22:27:23 - 00:22:42:22
The government doesn't care. And like, it just, you know, this is like a big, you know, it's a big failure. We can all agree, like, hey, we're not bringing in high school immigrants, but it's not like the government is fixing itself. So this is another big failure by the government, and we can't just wait for it to be fixed.
00:22:44:10 - 00:23:10:22
I I think that there is there's probably some kind of private model of biosecurity, that that, we can we can envision, it does have some regulatory bottlenecks, but I, I, I could, I could imagine, like, like some sort of, biosecurity as a service model where, where individuals and companies actually pay for, like, immune subscription to, to be, to be updated, against, against sort of both emerging and and circulating circulating pathogens
00:23:10:22 - 00:23:12:22
an insurance policy or something. Maybe.
00:23:13:05 - 00:23:34:22
Yeah. I mean, I mean, there's lots of pieces to it, right? Like you can imagine, like labs over the world, just like sequencing all these viruses and submitting them to, like, an open source kind of place. Right. Like, like if you take all the different pieces that you might want the W.H.O. or someone else to do, you could probably think of like, kind of bottoms up, idea that like, gets there as well.
00:23:35:04 - 00:23:37:10
But who's going to pay for it? That's the question. You know.
00:23:37:19 - 00:24:05:04
I mean, there's lots of labs already paid for. I mean, the one of the benefits is sequencing is so cheap now, right? It's not an expensive thing to go sequencing these things en masse. Yeah, I think it's doable. But, yeah, it does require, like, thoughtful organization. I really want to get to this, this AI topic that you that you kind of briefly mentioned in the intro Hannu, you said, reasoning models kind of fit very well with mRNA.
00:24:05:06 - 00:24:09:10
Yeah. I guess talk a little bit more about how that kind of works together.
00:24:09:18 - 00:24:39:13
Yeah. So, so basically like like the if, if we kind of, remind ourselves about, mRNA, so, so, so information flows in the cell from, from the DNA which stores all the recipes for proteins through through mRNA, which then sort of carries that information over to the protein factories of the cell that then make the proteins and that the proteins go off and do everything in our bodies and, and, our construction materials and, constructors and, and everything else.
00:24:39:15 - 00:24:58:06
And, the kind of power of mRNA as a vaccine or a therapeutic modality has been that you can also put in mRNA from the outside of the cell and the protein with certain constraints. But like the protein protein factory, still the ribosomes still read through it and make basically whatever genetic code you want to to, to include there.
00:24:58:08 - 00:25:18:01
So that means you can really program the cell to make make anything you want. Or even more generally, like you can even affect the state of the cell, like it's kind of to, to get any kind of biological state state you want. Like you can find an RNA combination that will drive like some more rapid wound healing or, or actually restore the cell to a more useful state.
00:25:18:01 - 00:25:40:21
And these are the kinds of things people are. People are developing, so so you have this like enormous generality, like, like you have this like 3D biological 3D printer essentially, that you can program. So, so that that means that, if you can now, take sort of any kind of biological problem you want to solve, like I want this immune cell to, to, to be very active against these kinds of targets.
00:25:40:23 - 00:26:07:09
You can now start working backwards from that desired state and try to figure out what combination of mRNAs would. You would get you there. And that's actually a reasoning problem. Now, you know, you're actually reasoning about like like causality across all these different pathways in the cell. So at a kind of a higher level of abstraction than just saying, okay, like I have this protein target and I need to hit that with a small molecule, because with mRNA you already have all these starting points, natural proteins that you can immediately deploy.
00:26:07:11 - 00:26:26:11
So so that's sort of a second order problem. But the first order problem is, what is the combination of things you need, to get from like state A to state B? So, so it becomes much more like a logic puzzle, of of the kinds of the kinds that the reasoning models are actually actually pretty good at and, and are obviously getting increasingly better.
00:26:26:11 - 00:26:44:10
And then the really cool thing is that you can you can also start giving it experimental data. And you can also start doing reinforcement learning to, to, to teach it, to, to teach it specific reasoning modes that work well for biology, and these other kinds of things. We have now been exploring. And
00:26:44:10 - 00:26:58:10
It sounds a little abstract. Can you give me, like, I guess give me an example of, like. Oh, what is the. Yeah. What is the thing you'd want to sell to do or something to do that you might, like, go prompt, prompt the reasoning model to try to figure out.
00:26:58:15 - 00:27:26:12
Yeah. I'll give you, I'll give a specific example. So, so one of the things we, we found for, at various points, to Oksana, was that, quite often you have some interesting protein, like, in our case, actually, like, literally some some of the mutants of, of, of Covid, like SARS-CoV-2. Vaccine targets, get efficiently made by the cell when you put the mRNA in, but then they get stuck inside the cell.
00:27:26:13 - 00:27:49:23
You want the cell to secrete them so that the immune system to see them, but they actually get stuck inside the cell. So how do we get them on stuck? It's like one, one very obvious example. And, and like, with sort of human brain power, we, we we figured out some partial, partial answers. Where, where, you know, we look at sort of how do all the secretion pathways in the cell work?
00:27:50:01 - 00:28:10:22
Where are the bottlenecks where, where things might be getting, getting stuck? And then we add the to the mRNA mixture that we put in, things that that sort of, modulate some of those or might try to are sort of geared towards removing those bottlenecks. And that can actually get very complicated. Like you might include like five different different things.
00:28:11:00 - 00:28:30:02
Besides domain meant main target. So now you know now. So now you have to like hold this, this set of all the pathways that affect protein secretion for the cells in your head. And you hit sort of the limitations of human like context window pretty quickly. But again, like these, these are the kinds of things that reasoning models can start solving.
00:28:30:02 - 00:28:37:22
And we are basically seeing them generate very plausible, plausible answers, or, or actually working answers to, to to this problem.
00:28:38:03 - 00:28:47:22
Oh, that's really cool. And. And you're saying ideally, you then hook it up to an actual lab so you can, like, go say, like, try this out and then see what the result is and try to, like,
00:28:47:22 - 00:28:49:19
iterate.
00:28:49:19 - 00:28:52:22
done, what we've done actually is hook it up to lab robots.
00:28:54:10 - 00:28:59:22
can actually go directly from the model designed experiment to, to the robot executing the experiment.
00:29:00:00 - 00:29:07:20
Yeah. Well let's just hope the models never go evil because looking them up for RNA lab seems dangerous.
00:29:07:20 - 00:29:10:20
there's there are like there are definitely scenarios where this is a very bad idea.
00:29:10:20 - 00:29:38:11
But of course, like we of course these are, these are sort of constrained systems. Like this is not like a universal universal biology lab, but like a specific set of the robot robot sort of sort of carries it carries out a specific assay rather than like, like any conceivable conceivable one. But the but it does it does give you this like very quick iteration loop, which where, where now even if, even if you sort of like what, what if we're actually seeing this, and there's a, there's like a really, really cool result.
00:29:38:11 - 00:29:55:14
We saw that, that I, that I can't, can't talk about, but the, the but basically in a different problem than the one I described, but analogous, we saw like, the kind of performance with two experimental iterations that you would expect from like a smart grad student for a year. So that's kind of the level level now.
00:29:55:14 - 00:30:25:14
But then, of course, the system is tireless and you can just keep keep going and even if it's sort of not superhuman, it actually actually will be functionally superhuman just by, by the virtue of infinite patience and, and timelessness. But the, the the interesting question now is that when are we going to get that sort of move 37 moment like, like like with with AlphaGo, where, where it starts generating designs, where, where, that a human would not have been able to and we're not there yet, I would say.
00:30:25:14 - 00:30:36:22
But but like, given the advances both on the sort of, baseline model side and then the ability to do RL with the kind of experimental data that we're generating, I don't think it's far.
00:30:37:01 - 00:30:38:22
But what's. What's move 37.
00:30:38:22 - 00:30:49:04
Oh. So, so this was in the, in the match between Lee Sedol and AlphaGo, the famous, famous go game between Lee Sedol and AlphaGo.
00:30:49:06 - 00:30:58:10
AlphaGo made its 37th move. And and all the go enthusiasts were like, what? And and there's this,
00:30:58:10 - 00:31:00:10
And everyone thought it was a bad move, right?
00:31:00:10 - 00:31:10:13
everyone thought it was a bad move, but it was it was an absolutely brilliant move. And there was a, there was, another, another go champion who was watching the game, commentator and I can't remember his name.
00:31:10:13 - 00:31:18:22
Name, but he basically said, that is not a human move. I have never seen a human make a move like that. And then and then they got sort of entranced by it and said, like, it's so beautiful. It's
00:31:19:06 - 00:31:22:22
Yeah. Yeah. That's really interesting.
00:31:22:22 - 00:31:29:02
that that moves. And that move actually spawned a whole sub literature of, of goal like, like people then sort of.
00:31:29:04 - 00:31:38:10
So the human goal game with go level sort of went up just by the result of, of there being like a demonstration of superhuman got ability. So.
00:31:38:10 - 00:31:55:10
Yeah. Have you been in the last, like, 12 months? Have you been surprised by the progress of AI or. Or kind of like, you know, where if you if one side is like, oh, my God, is going much faster than I thought, and the other side is like, oh, it seems like it's plateauing. Where are you on the spectrum?
00:31:55:20 - 00:32:20:10
I was surprised by the biology results that I talked about. I mean, that that that's, that's was definitely closer. Like, like that happened much faster than I thought it would. Also the, like, just generally the reasoning, like, chain of thought process, kind of has been, has been really rapid. Does this, arc AGI, benchmark, like for solving these visual puzzles,
00:32:20:16 - 00:32:21:22
Yeah I saw the. Yeah.
00:32:21:22 - 00:32:24:16
where, from from who's the gentleman who was behind it?
00:32:24:17 - 00:32:27:22
Behind, French la France? Who actually who
00:32:28:00 - 00:32:33:05
That's the one that O3 Pro is doing really well. That right. Like the open AI model.
00:32:33:05 - 00:33:03:01
so that that seemed initially like, oh, like the models are just, like, terrible at this. And, and, and humans are much better at this. But now it's actually like, like, getting, getting to sort of, top human performance, especially with scaling, with inference time. So, so, so yeah, this is all the reason and, and then like, I mean, I, I don't know if, like deep sequel is fundamentally surprising, but it's sort of sort of showing the really the power of RL also applied to, to reasoning, very, very, in a very impressive, impressive way.
00:33:03:02 - 00:33:14:10
So, yeah, that's this is for me. It's it's been like, like definitely updating priors on on progress happening at least in these fronts faster than I that I, that I expected,
00:33:14:23 - 00:33:39:23
Yeah. It's also cool to hear that there is these kind of. Yeah I guess like I see the generalized stuff but like it's cool to hear that bio has like these like specific application that are powerful. And I can imagine like every industry, that has like this kind of deep knowledge work is kind of exploring. It's, you know, probably take a few years for the US to have, like, actual dividend from it, but, it seems like it would be it will be huge.
00:33:39:23 - 00:33:43:10
As like we continue improving the models and rolling them out in other industries.
00:33:43:10 - 00:33:56:10
Yeah. I mean, the other biology specific area, of course, which is much more directly biological is protein model models. So protein language models applied to proteins. So I think we might have talked about that a bit last time. And
00:33:56:10 - 00:33:57:22
AlphaFold.
00:33:57:22 - 00:34:12:00
yeah. AlphaFold is yeah. I mean, in the same, same category. But the, the what is actually more surprising was that the, the, just applying transformer architectures and next token prediction also worked really well for protein structure structure prediction.
00:34:12:02 - 00:34:23:09
And some of the more recent models incorporate that. And they are also generative. So so you can use them to design and generate new proteins. And maybe the most surprising thing
00:34:23:09 - 00:34:47:02
There, was I guess that's almost almost 12 months ago now. Was a, paper from, a company called Profluent, who, published a paper where they used, one of these, these protein LLMs to basically generate new versions of Crispr that do not exist in nature but still work as genome editors.
00:34:47:04 - 00:35:07:00
So that's been kind of, I think that's something that's only, only kind of still also percolating through the industry that, now we can basically take a protein that has a function that we want and make like a completely synthetic mimic of it, that, that has a different sequence from, what is found in nature, but still does the job we want.
00:35:07:04 - 00:35:15:10
And, and the immediate corollary of this immediate consequence of this is that old biologic, biologic like drug IP is broken
00:35:16:10 - 00:35:29:20
because you can always make a mimic that, that, make it. Yeah. That that is not covered by the, the IP. So so that's, that's, that's kind of interesting. And that, that, that, that is going to have a lot of consequences, I think.
00:35:29:22 - 00:35:35:10
But but that's yeah, that's a, the reasoning model stuff was more, more recent update.
00:35:35:13 - 00:36:03:01
Hannu To switch topics a little bit. I'm curious too. What do you think of China's progress in the biotech industry? There's an article in The Economist just four days ago about talking about, you know, how China has made dramatic progress on, on biotechs and actually leads the US in many ways. I think I saw, you know, a Twitter post by, David Lee talking about how China is much faster on preclinical as well as clinical development.
00:36:03:03 - 00:36:17:13
And there's certain areas where they just are outright leaders now, on, on biotech, it seems like this is an area where the Chinese government is also supporting them, but it's leading to a place where I think a significant percentage of new, innovative drugs are now coming out of China. All right.
00:36:18:09 - 00:36:42:13
I mean, that that was my. That was the point I was making earlier that like, the pharma, pharma, pharma pipelines are now fueled by by drugs developed in China. I don't know if they're necessarily like, cutting, cutting, cutting edge, but they are executing, faster in the clinic for sure. And, and are able to basically, do the process of from a concept to, to a human to human proof of concept, much, much faster.
00:36:42:13 - 00:37:04:01
And, the so, so I think which, which does then obviously suggest that, in that in the, in the US, we should be focusing on really pushing the cutting edge even even farther, further. And because every, every kind of easy thing or very obvious thing, we make is going to be be executed faster in China. So,
00:37:04:01 - 00:37:05:13
What do you think is happening there? Like.
00:37:05:13 - 00:37:12:07
it's obviously it's obviously good that, drugs get developed faster and actually actually get deployed, deployed faster.
00:37:12:07 - 00:37:26:13
But, the, the from the standpoint of like us companies remaining competitive, I think like, it just means we need to do more moonshots rather than try to big because the obvious things will get executed better by, by by others.
00:37:26:23 - 00:37:35:00
Why do they move faster, though? Is it, is it. They're just better at execution? Or is it like, you know, lower, lower regulations when it comes to, you know, they're the
00:37:35:00 - 00:37:45:01
lower lower regulation and like like, yeah. Like like really a concerted push to, to support support like, that,
00:37:45:01 - 00:37:52:01
I mean, you don't really associate China with lower regulation. So, you know, it's, like they've targeted this industry to move fast on. It seems like.
00:37:52:11 - 00:38:00:19
Are they going to do the same thing they did with the EVs? With biotech? Right? Can they. Can they become the leaders if they just, like, subsidize it and focus on
00:38:00:19 - 00:38:16:00
I mean, I mean, China has also been basically the CRO of the world, like they, they're the that they've been like the biologics manufacturer and the, the sort of sort of also running clinical trial or already like running clinical trials for others or animal studies for others. So, so
00:38:16:13 - 00:38:18:07
What does CRO stand for?
00:38:18:07 - 00:38:20:02
contract research organization.
00:38:20:02 - 00:38:22:19
So, so, so people have been basically for many
00:38:22:19 - 00:38:30:19
So they've already been, like, the kind of factory of the world as I'm extending. And now they're, like, trying to grow up. That's interesting.
00:38:30:19 - 00:38:35:07
that now there's just using that infrastructure, to, to do the actual clinical programs.
00:38:35:07 - 00:38:38:09
Yeah. Great. So if there's ever a war with China like we're going to do, we're
00:38:38:09 - 00:38:40:19
Okay. Now are we going to have like bio.
00:38:40:19 - 00:39:04:07
Oh I mean that's already like like the, the, like, like, legislation and, and restrictions on, on like, using, using materials from China have already impacted the biotech industry. So, so there, there was like, like the Biden administration had a whole push on on how to be of, the US should become more independent in biomanufacturing, which, which I would imagine the current administration would actually actually agree with, at least theoretically.
00:39:04:07 - 00:39:06:19
But, I don't know where they're going to go
00:39:06:19 - 00:39:08:19
hard to tell these days, honestly. Yeah.
00:39:09:00 - 00:39:16:03
Well the vaccines that were produced for Covid 19 like the you know those were us companies that did them. Did they not produce them in the US.
00:39:16:07 - 00:39:21:07
Well, I guess, I guess, I guess Moderna, US company, BioNTech BioNTech German. German
00:39:21:07 - 00:39:23:07
So, against Western Company Limited,
00:39:23:07 - 00:39:27:20
but did they not do it in the US like was the production wasn't in the Western.
00:39:28:00 - 00:39:38:07
Moderna for sure. I think the, the, I think for the Pfizer BioNTech, there was at least some manufacturing in Germany. So, so I don't, but I mean, they obviously
00:39:38:07 - 00:39:43:22
It's not like they're being manufactured in China, right? It's not like. It's not like we don't have the capacity to manufacture the things.
00:39:43:22 - 00:39:51:07
some of the materials, though, some of the materials, like, like the, the reagents, and, and, kind of upstream.
00:39:51:09 - 00:40:00:07
I might it might have come from China. I am not, not actually, pretty pretty sure there was some supply chain, global supply chain dependency. And so in certain materials there was a massive shortage of. So,
00:40:01:19 - 00:40:09:07
yeah, it's it's the whole like, globalization, like like, other other side of the coin.
00:40:09:15 - 00:40:27:04
Well, I mean, I think there's huge benefits as well. Right? If China is, like, producing drugs cheaply, you know, there's, globalization has benefits, and we can get cheaper drugs. And, you know, if they're doing a bunch of work, I mean, to deregulate and bringing drugs to market faster. I mean, we need those drugs because, I mean, you know, us.
00:40:27:04 - 00:40:33:13
And it's not like U.S. has the best life expectancy in the world. You know? So we actually need, you know, we need those medicine.
00:40:33:13 - 00:40:52:13
And one of the benefits of drugs is like, you know, once you create that IP, it's out there right. Like we'll be able to use it and people will be able to learn from it. It's not like a closed industry, even if like, there's geopolitical, political kind of, conflict.
00:40:52:13 - 00:41:03:17
Yeah. No. And I think as, as we sort of came up like I think like IP, IP is actually hard to protect. No, no. Or it much harder to predict now with, with these AI advances.
00:41:04:08 - 00:41:25:17
Do you think that that, like, fundamentally changes the pharma industry if they can no no longer rely on like, oh, we discovered this thing. We spent $5 billion on that. Now that we have a patent and we can just like own that space for, for a few years. But if it's like super easy to go replicate it, then you know, what happens to, like, this kind of pharma pipeline and,
00:41:25:21 - 00:41:47:21
Yeah. It's a it's a good question. I mean, I think the moats just have to move somewhere else. And I think, like, to think of some historical examples. There there is this concept of biosimilars, which is like the biologic drug equivalent of generics and, and that an end for many of the big monoclonal antibody drugs, like, like, let's say like Humira or Keytruda.
00:41:47:23 - 00:42:19:09
There are, biosimilars like clones that others have made, but they haven't yet, but they haven't fully taken over. And the process of, you still need to make them, and manufacturing biologics is, is harder. Than, than than small molecule drugs. And then, you know, obviously you have to also have the, support of the, the clinicians and, the, the actual markets and the sort of sales force to be able to, to end relationships, to be able to, to get them to patients.
00:42:19:09 - 00:42:30:05
So, so I think the moats may move from the IP to, to that. That's like manufacturing and clinical and, and and actual execution. So but yeah, I think that's, that's not necessarily bad.
00:42:30:05 - 00:42:41:17
Immad. I had a question for you. I mean, you know, you obviously worked in a very different space than biotech. You know, throughout your career, do you do a lot of biotech investing? Is this, an area of passion for you that, you know?
00:42:41:17 - 00:42:59:05
honestly, biotech is like the one space that I think is exciting, but I don't invest in because the problem is, like, I just have no idea. Like, I have talked to many people and I'm like, oh, wow, this person's curing cancer. And then they talk for a while and I'm like, I have literally no idea what you're talking about.
00:42:59:07 - 00:43:18:04
You know, most of the spaces like, like my general interest, I feel like, like I love talking to high tech about it, but, yeah, if I talk to, like, a space tech company, you know, I can, like most of them actually are not doing science like they're doing mostly doing engineering and, business model and application of that, anomaly.
00:43:18:04 - 00:43:30:20
I can, like, get my head around it. Like, I'm not saying I'm a space expert, but I'm like, okay, fine. Like, you're taking pictures from space. Like, this is the cost of getting a satellite at space like, this is like the type of market you have, and this is how you can do it. Okay? You know, I can get that.
00:43:30:22 - 00:43:49:22
But when it comes to, like, you know, someone who sounds fairly smart and is doing like a cure for something that sounds important and that's a huge market, but it's going to take them six years to get to market. Like I just have no idea. And and the more I talk to people actually in that space, including investors and like, okay, it's a very specialized field where like, they really do get it.
00:43:50:00 - 00:44:05:17
And I just don't feel like I need to play that like I have, I have there's enough investments to be made in like hot tech and software and all these kind of things that are not in this, like, somewhat weird space that, like, I think it is like like everything is different about like the investors are different.
00:44:05:17 - 00:44:26:17
Like the path to exit has been like pretty strange, right? Like a lot of these companies go public before they have revenue and then they sell to pharma. It's a very different space. Like I think it's actually, I know I, I really respect the investors that started in software and do successfully play in biotech. And there's a few of them, but yeah, I just can't get my head right.
00:44:27:09 - 00:44:28:17
Hannu, do you have any tips for us?
00:44:28:17 - 00:44:47:00
Just a quick comment and that that in mind. So, so I'm reminded of this, heuristic that I heard from, Charlie Longhurst, who's, who's one of our, early angel investors and, and has done successful who, who's actually not not a, not a scientist, but, has has done quite a lot of successful biotech investments also.
00:44:47:02 - 00:45:09:12
And, and he basically said that the way he thinks about it is in all investing, you can you can essentially be successful in two ways. It's sort of like the IQ meme, right? Like the caveman and the Jedi. And then there's this, the sort of Midwich guy, first person in the middle. So. So either you take the caveman approach approach and you totally go with your gut and you're like, okay, like, like does does this person seem smart?
00:45:09:12 - 00:45:33:17
Like like, do I do I trust them like, like like, does this from first principles make sense? And then you make the make the decision based on that. Or like, you really like spend $1,000, like understanding every single, every single detailed detail of it. And then, then you, then you make a much more informed decision. But the middle but in fact, I can tell you that 99% of biotech investors live in the middle.
00:45:35:05 - 00:45:48:17
so, so, so so they sort of project this. They're very good at projecting the aura of, of specialized expertise. But actually actually they are I, I don't think they are better. I think the sort of caveman
00:45:50:05 - 00:45:54:17
actually better. And most biotech investors are not that. So, so, so as a, as a
00:45:54:17 - 00:46:04:05
I mean, I've, I've actually been through and all all investing. Right. Like 90% of VCs don't make money basically. Like I think that's like probably, probably similar.
00:46:04:05 - 00:46:10:05
I like the caveman approach. I think is, it's probably. I mean, you know, betting on good founders generally is
00:46:10:05 - 00:46:29:05
mean, I think right at the start, like basically all you have is a caveman approach, like how much analysis are you going to do on like this? These people with like a DAC and like they seem smart and all. The huge idea, right. It's hard. It's hard to actually be like amazingly analytical in that position. Do you do any investing in, in biotech Hannu?
00:46:29:16 - 00:46:35:17
I, I don't sometimes I get asked for it to, to do advice but I don't myself just don't have the, the capacity.
00:46:35:17 - 00:46:42:05
Assuming you had more time and money, would you do it? And then if you did, like, what would you look for in these kind of ideas?
00:46:42:17 - 00:46:44:08
Give us some tips here. Yeah.
00:46:44:08 - 00:47:04:10
if I, if I was able to I might just, just to basically try to pay forward the support that I've been given also and try to empower more people to try to do exciting things. I would certainly try to be very first principles about it. Like, like there, there would be that human element, like are you trying to do something, something, ambitious.
Share this post