YT YouTube2Text

YouTube2Text → Transcripts → I bought palantir at $7. these 8 stocks are the biology version

I bought palantir at $7. these 8 stocks are the biology version — Transcript

by Antonio Linares · 6,133 words · 985 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hello everyone. Today we're going to
0:01look at eight stocks, eight names
0:02companies which I believe are going to
0:04change the world in the next 5 to 10
0:06years time. The broader thesis is that
0:08in the next few decades biology is going
0:09to deliver most of the returns in the
0:11stock market. Welcome back to the
0:12world's best investing podcast. I went
0:14long Palantir at $7 per share. The
0:16evidence of that you can find it in my
0:18original deep dive written back in 2022
0:21explaining why Palantir was the most
0:23important company in the West and then
0:24me actively defending the thesis all the
0:26way up with the various 50% plus
0:28declines and everyone trying to short
0:30the stock on the way up. Today as I
0:32said, we're going to look at eight
0:33tickers in depth. The very least this
0:35presentation is going to do for you is
0:36make you smarter. If you relate to the
0:38AI trade, this presentation is going to
0:40enable you to at least understand the
0:42next big huge trade early which is the
0:44intersection between AI and biology.
0:46Nothing I say today is financial advice,
0:48nor should it be interpreted as such.
0:50This is just for educational and
0:52entertainment purposes only. Let's get
0:54started with today's presentation.
1:00Right. So, life extension per token.
1:02Many of you guys have heard me talking
1:03about this. This is a new AI scaling law
1:05which I've coined myself. Essentially,
1:07the idea that every time an AI model
1:09runs that's trained on biological data,
1:12human lifespan is going to be extended.
1:15Longer lives, better lives as an
1:17emergent feature of this very simple
1:19data set which is your ability to read
1:21into someone's biology at present, write
1:23back into it, and then be able to tell
1:25what happens afterwards. What's the end
1:27state in someone's biology. This is the
1:29very simple key value pairs we'll cover
1:31now that will train superhuman AI
1:33doctors. In the future, it's actually
1:35happening now and this is going to give
1:37way to the most valuable subscription
1:39service ever in the history of humanity.
1:42And that is the bottom line of the
1:43thesis. Today we have eight tickers. I
1:45know many of you guys like me to cover
1:47the tickers early. So, we have Hims 10x
1:50AI, Nautilus Biotechnology, Butterfly
1:53Network which is the next deep dive.
1:54Today we're only going to cover it
1:55superficially. Recursion
1:57Pharmaceuticals, AbCellera, ImmunityBio,
1:59and MiNK Therapeutics, Hims, and Tempus
2:02AI at the distribution level. It's very
2:04important for the read and write
2:06function in biology. I'll explain the
2:08importance of that now. So, actually
2:09read patients so we can see what's
2:11happening with patients before and after
2:13a treatment. That is the essence for
2:15training an AI. And then there's
2:17actually deeper components of this value
2:19chain. We have Nautilus Biotechnology,
2:21which is very important, increasingly
2:23so, for the read function at the
2:25proteomic level. Butterfly Network seems
2:27to have extraordinary technology that
2:29can tell the state of specific organs at
2:32present, and this is actually quite
2:33unique in the market via ultrasound,
2:35essentially chips that run ultrasound in
2:38situ inside the chip. Then we have
2:39Recursion Pharmaceuticals, which is the
2:41drug manufacturing machine which builds
2:44itself. It's recursive in nature,
2:45apparently. AbCellera, which is a
2:48company that I've covered a lot.
2:49Congratulations to you guys that bought
2:51in early. I didn't. It's essentially a
2:54factory that just prints antibodies and
2:56can therefore target any receptor in the
2:58human body. By the way, if you don't
3:00know anything about biology, you can go
3:02and check out my past videos on that. In
3:04the course, I have a module in which I
3:05break down the first principles
3:07framework so that you guys get up to
3:09speed quickly without spending years
3:11studying biology like I did. And then we
3:13have ImmunityBio, which is basically
3:15something that jacks up your immune
3:16system by binding onto the interleukin
3:1815 receptor. I do believe that will be
3:20fundamental
3:22in the next 5 to 10 years, specifically
3:24so in the next 5 years. So, that's very
3:26important to take into account. And then
3:28we have MiNK Therapeutics, which is the
3:30only company in the world, at least
3:31publicly traded, that I know of, that's
3:34capitalizing on a third branch of the
3:36immune system, which is not the innate,
3:38not the adaptive branch, it's something
3:39in between that targets the bad guys via
3:42what is known as lipid. So, it doesn't
3:43use the peptide complex, therefore does
3:45not leverage the major
3:47histocompatibility complex, and
3:49therefore is extraordinary. You can go
3:51and check out
3:52my deep dive on that company.
3:54>> [snorts]
3:54>> Now, let's get started with the
3:56worldview that I was promising. Biology
3:57will, in my view, deliver most of the
3:59stock market returns of the next decade,
4:01primarily by extending human health
4:03span. This is going to be the most
4:05valuable, desirable subscription service
4:07on Earth because ultimately, as a human,
4:09if you don't care about sticking around
4:11and in a good state, what do you really
4:12care about? If you don't stick around,
4:14you can't possibly care about anything.
4:16Therefore, on a first principles basis,
4:17if this does come to life, as I believe
4:20it will do, and actually, as I will
4:21prove in today's presentation, is coming
4:23to life, this is going to be extremely
4:25valuable. Semiconductors have been have
4:27proven to be extremely valuable, and
4:29they're only getting started. As many of
4:31you know, I'm an early AMD shareholder
4:33since $4.2 per share. At the very least,
4:35you have billions of videos of me
4:37defending AMD last year when it was when
4:39it was trading at $75 per share, saying
4:42it's the next AI
4:44trillion-dollar AI giant, and so it has
4:46been. It's now trading just over 1
4:48trillion. All right, so
4:50we have an emerging data value chain
4:52that's already extending lifespan per
4:54token. This is something that the market
4:56is missing. The financials are starting
4:58to click, and several key assets were
5:00actually severely underpriced. And to
5:02me, that's certainly an arbitrage, which
5:04which I am exploiting. There are a
5:06series of newly started long positions.
5:08Newly means in the past year, or
5:10recently in the past few months, which I
5:12will disclose during the presentation.
5:14The building block of this new scaling
5:16law, which is more lifespan per token
5:18every time an AI model runs that's
5:20obviously trained on the adequate data,
5:23is a single key value pair, as is the
5:24case with the rest of AI. The key is
5:27someone's biological state before a
5:29treatment, then plus the treatment, and
5:31the value is the state after. So, if you
5:33have a snapshot of someone's biology
5:35before a treatment, then you add the
5:36exact treatment, and then you see what
5:38happens to someone's biology after.
5:40Enough pairs of that, and AI does become
5:43biologically predictive. Now,
5:45historically,
5:47every major advance in has come from a
5:49better read or a better write function
5:51in biology. So, as we as we understood
5:54biology better and we've gained an
5:56ability to then write back into biology,
5:58we've cured things. Now, both run on
6:00data and compute uh compute that's
6:02recursive in nature. So, you essentially
6:04have AI models that are starting to
6:06improve themselves. So, each better read
6:08actually trains a better write. And we
6:11are right now going through revolution
6:13in this space, which, you know, is
6:14increasingly acknowledged by the wider
6:17uh public, but it's still very niche,
6:19right? So, if you tell anyone about this
6:21that's not on, you know, in in our
6:23sphere of investing and technology,
6:25they're going to think you're insane,
6:26right? So, you have the key value pair
6:28behind predictive biology. As I was
6:29saying, it's very important we nail
6:31this. You have someone's biological
6:32state before captured by the read
6:34function, then you have the treatments,
6:36which is delivered by the write
6:38function. So, we have, you know,
6:39something in the write function that's
6:41booming right now is peptides, for
6:42example. You're just writing back into
6:44biology.
6:45And then you have the state after.
6:47Millions of these pairs teach the model
6:49the link between key and value. Then,
6:51you know, when you have millions of
6:53patients on board, for example, and this
6:55key value pair is increasingly training
6:57an AI model,
6:58when a new guy joins the platform, for
7:00that guy, the AI is going to be
7:01predictive. And that's what's going to
7:03change not just medicine. It's not only
7:06going to produce the most valuable
7:07subscription service on earth, in my
7:09opinion, it's also going to change the
7:11human condition forever.
7:14Now, here's where we get into a spotting
7:16the asymmetries, you know, the tickets
7:18that we were discussing, how good the
7:19prediction is. So, how much someone's
7:21health span is extended per token, per
7:24every time a model runs, depends on how
7:26deep the read goes, from how you feel
7:29through each organ down to the proteome,
7:31epigenome, and genome. By the way, if
7:33you've never heard of the proteome,
7:35epigenome, and genome, I have lots of
7:36videos of that. As I said, we have a
7:38module on the course with the first
7:40principles framework.
7:41Let me just illustrate what I mean by a
7:43deeper read. How the person feels, how
7:45they're sleeping, how much they work
7:47out, how their heart is performing. We
7:49have that in the wearables today. That
7:51is being democratized at scale with
7:54Whoop, which I've been a customer and
7:55now actually taking a break from it
7:57because it was a little bit painstaking.
7:59But, um you know, Aura stuff like that,
8:01this is happening across the board. You
8:02have the intelligent mattresses that
8:04people are using. That's the very
8:06high-level surface biomarkers. Then we
8:08have organ performance. This This seems
8:11to be some kind of innovation happening
8:13inside Butterfly, ticker BFLY. I'll be
8:16discussing that in depth in the upcoming
8:17deep dive. The proteome is very well
8:20done by Nautilus Biotechnology, frankly,
8:23in a way which no one else seems to be
8:25able to do so at the moment. The thing
8:27everyone is using in the industry is
8:29mass spectrometry, which we'll be
8:30covering now, but it doesn't work to
8:32truly capture nuance in the proteome.
8:34So, this ticker to me is fundamentally
8:37interesting and it's to me it seems
8:39incredibly undervalued. And then,
8:42at the epigenome, RNA, genome level, we
8:45have two companies which I think are
8:46very interesting, which is Tempus AI,
8:49ticker TEM, and then Recursion
8:51Pharmaceuticals, ticker RXRX. So, just
8:54understand that and I'm not saying these
8:56tickers will succeed 100%, but just
8:58understand that the deeper the read into
9:01biology goes, the better, more
9:03predictive, more efficient AI will be at
9:06extending human health spans across the
9:08board. Now, to test whether this chain
9:11is actually viable, it's real and it's
9:12not something I'm making up, you have to
9:15look at the distribution layer. So, the
9:17platforms in direct contact with the
9:18customer where read and write actually
9:21reach a body. Is this happening across
9:23the board or not? Yes, two companies,
9:25Hims and Tempus. You guys know I'm a
9:28long-time Hims bull. I'm a gigabull as
9:31I've been with companies like AMD and
9:33Palantir and I continue to be to date.
9:35These two companies,
9:37as I've covered in the past, perform the
9:39same function but are coming at this
9:40through opposite doors. Both are
9:42actually growing, as I will show now,
9:44for one simple reason. And that's that
9:46data improves outcomes. Whether you
9:48believe in the emerging data value chain
9:51that I'm going to cover or not, you
9:53can't deny the fact that actually the
9:55top line of these two companies are
9:56growing really fast.
9:58Whether AI is truly driving patient
10:00outcomes or not, we can debate back and
10:02forth. But the fact that these are two
10:04data-driven
10:06and the top line is growing
10:08exponentially for now, just because the
10:11intel that the data is generating
10:13enables these companies to deliver more
10:15value per dollar spent is actually
10:17increasingly harder to debate. Right?
10:19So, this is a hypothesis. There is no
10:21guarantee of success in investing. But
10:23you have Hims doing this revenue in the
10:25last 12 months, 11.3x revenue growth
10:28with respect to the start of this
10:29timeline. Tempus 1.4 billion in the same
10:33period, 5x revenue growth. You will see
10:36there is a tentative arbitrage in the
10:37market cap. So, Hims 6.7 billion dollars
10:41market cap, Tempus AI 14.9 billion
10:44dollars. Although there is a disparity
10:46in the revenue. This is because the
10:48market at present interprets Hims's
10:51revenue as being lower quality of that
10:53of Tempus. However, as we know in
10:55technology, things that come to be very
10:57meaningful eventually actually look like
10:59toys at the beginning. Hims has been
11:01labeled as an erectile dysfunction
11:03company, GLP-1 company, whatever pill
11:06company. It's actually building an
11:07infrastructure that's D2C. It's just a
11:10different way of nailing this key value
11:13pair. Right? So, Hims Labs picks up data
11:15on people before treatment,
11:18biomarkers and increasing volume of
11:19them, then the treatment that they
11:21receive, and they do so via this
11:23closed-loop infrastructure, and then the
11:24state after. So, on a first principles
11:26basis,
11:27it's very much training in AI, just like
11:29Tempus is.
11:31Now,
11:32there's some interesting qualitative
11:34remarks from the co-founder and CEO of
11:36Tempus AI in the Q2 2026 earnings
11:40report. He said, "They, the customers,
11:42don't just want our data. They want
11:44access to Lens. They're uploading data.
11:47They're building models that remain in
11:48Lens, and the business just feels super
11:50healthy, super sticky." Algorithm attach
11:53rate 45%. That's up from 40% in the
11:56latest report. Data licensing revenue
11:59growth 36% year over year.
12:02Now, the company is actually for the
12:03first time ever
12:05positive GAAP net income. This is
12:07because Tempus is becoming a machine
12:10that not only delivers value just by
12:13selling raw data, it's actually a place
12:15where customers are beginning to
12:17generatively build their own AI models
12:20on top of Tempus's data. So, Tempus is
12:23entering this new chapter in which at a
12:25marginal cost, it prints AI models that
12:28then drive additional incremental value
12:30at a marginal cost to their customers
12:33inside their infrastructure. So, rather
12:35than just a data platform, it's becoming
12:38a place where new AI models become
12:41bootstrapped by customers via prompts,
12:44which is extraordinary.
12:46Now, that we've covered the top line and
12:48we we can sort of broadly agree that
12:50something is happening for these two
12:51companies to grow so fast, it's
12:53interesting to go deeper into the value
12:55chain. Because here is where I believe
12:58also the case at the top of the value
13:00chain is where we will uncover many
13:02multi-baggers in the decades to come.
13:05This thesis, life extension, health span
13:07extension per token, I believe is
13:09extremely fertile ground for
13:12extraordinary investments as has been
13:14the sort of
13:16vanilla standalone AI value chain with
13:18companies like AMD and Palantir over the
13:20past few decades.
13:22Recursion is a machine that essentially
13:24explains why a treatment should work.
13:26And Tempus owns the outcomes, right? So,
13:29they can read a patient's state before
13:31and after the treatment. So, they own a
13:34higher level of the same key value per
13:37abstraction. Now, they recently closed
13:39the deal
13:40in which Recursion pays Tempus $42
13:42million over 3 years
13:45for data regarding their patients. And
13:47Tempus pays $12 million to Recursion
13:50Pharmaceuticals over 2 years for data
13:53for their TX FM RNA model. So, for a
13:56deep comprehension of what's happening
13:58inside cells at the RNA level.
14:01In case you haven't seen my Recursion
14:03Pharmaceuticals deep dive, they own
14:05roughly 50 petabytes of lab data. They
14:07have a factory which processes cells and
14:10generates deep low-level biological data
14:13of what's happening inside cells. This
14:15is hard to do because you have to build
14:17out all the physical infrastructure to
14:20get the data of what's happening inside
14:22cells, and then you have to build the AI
14:23models on top of them. Of course, as AI
14:26scaling laws continue accelerating, the
14:29intelligence built per petabyte of data
14:32generated in Recursion's factories
14:34continues going up. So, every day that
14:36goes by, Recursion, in my view, is
14:38actually exponentially harder to
14:39replicate. Bottom line is
14:42Recursion has a deeper level
14:44understanding, lower level
14:45understanding, of what's happening
14:47inside cells and how the chemistry and
14:50drugs interacts with that biology.
14:52And this deal just shows you a
14:54delineation
14:55in the data value chain. These two
14:57companies are paying for each other's
14:58data because they can't replicate that
15:00data. So, this is a very strong signal
15:03from Q2 2026. Now, what really caught my
15:07eye in Q2 2026 for Recursion
15:09Pharmaceuticals is that the model seems
15:12to be discovering novel biology. Not a
15:15hypothesis that someone prompted the
15:17model and to gain clarity on. The model
15:20came up with something that no
15:22specialist actually even thought of
15:24apparently.
15:26This is the result of a novel atlas that
15:29Recursion has built by processing over 1
15:31trillion lab-grown neurons and
15:34microglial cells. Neurons, as you may
15:36know, are essentially the cells that
15:38power the central nervous system
15:40including the brain, and microglia are
15:43the resident immune cells inside the
15:45brain. This is mapped together with
15:4717,000 genes that Recursion has
15:50essentially looked at. So, this atlas
15:52has produced, apparently, a novel
15:54neurological target, so something in the
15:56brain that can be interacted with to
15:59produce to improve patient outcomes that
16:01Genentech is now taking to the clinic.
16:04What's interesting is that this atlas is
16:06obviously reusable. This was an emergent
16:08property of the atlas, and this is sort
16:11of increasingly generalized intelligence
16:13such that it would seem that Recursion's
16:15model, there is a combination of them,
16:17but I think we can speak of these models
16:19as one emergent model, came up with
16:21something which really no one thought
16:23about, according to management's
16:25qualitative remarks in the Q2 2026
16:28earnings call. Now, I'm going by what
16:31they say because this seems like an
16:32extremely capable management team, and I
16:34think I increasingly trust them. What's
16:36interesting here is the relationship.
16:39So, we talked about how
16:41the growing top line of the top part of
16:44the value chain, which is the
16:46distribution layer, seems to be powering
16:48demand for lower levels of the emerging
16:51data value chain. What's interesting,
16:53however, is the recursive nature of the
16:56relationships there. Meaning, if this
16:58thing is coming up with novel biology,
17:00does that supercharge the top line
17:02itself? Does that open a field? Does
17:05that open a field of possibilities such
17:07that Tempus and Hims can radically
17:10improve patient outcomes in a way which
17:12was previously considered impossible.
17:14This is absolutely the case, and this is
17:16actually happening now. Again, we just
17:18go back to the idea of more treated
17:20patients, more key value pairs,
17:22and then that sort of feeds back into a
17:24richer Atlas. So, this emerging value
17:27chain is extremely real. And as I will
17:29explain now, across these series of
17:31places, it's actually being tremendously
17:33undervalued.
17:35>> [snorts]
17:35>> Recursion by itself is an extremely
17:38interesting company to me because the
17:40engine seems to run at a level of
17:42efficiency
17:43which the broader pharmaceutical
17:45industry cannot match. And as I said, it
17:47seems to be recursive, meaning that the
17:49infrastructure improves itself. They do
17:51have a closed loop where they have a
17:53physical factory that generates the
17:54data, then the models, and then the
17:57models sort of improve the factory
17:59itself, and that's recursive in nature.
18:01But, if you look, for example, at REC
18:034881,
18:04which is essentially a molecule, like I
18:06covered this in depth in the deep dive,
18:08but it's a molecule
18:10that reduces polyps in the
18:12gastrointestinal tract. It did so by
18:15roughly 43% actually, concretely 43% in
18:20phase two.
18:21And what's interesting is that this is
18:23done at a level of efficiency which the
18:25market, the broader market, cannot
18:27match. Right? So, the industry norm is
18:30roughly thousands of compounds
18:32synthesized in order to produce one drug
18:34that does something useful in the
18:35clinic.
18:36So, we can see, however, that REC 7735
18:40only required 242 compounds. And just
18:43the the recursion average is 330. So,
18:47it's a lot lower than the thousands
18:48required by the industry
18:51across the board. What's
18:53incredibly interesting, even more, is
18:55that Recursion Pharmaceuticals is doing
18:57this while the guidance in terms of
19:00OPEX, operating expenditures for 2026,
19:03is 40% below the OPEX in 2024. So, the
19:07machine is not only outperforming the
19:09broader industry, it's not only
19:11allegedly discovering novel biology,
19:13it's actually doing so in an
19:14increasingly agile,
19:17lean manner. To me, this is
19:18extraordinary and it is very much worth
19:20keeping an eye on. However,
19:23this is part of this map, obviously, is
19:25the whole multi-omics thing. So, here we
19:27we we said they processed 17,000 genes.
19:30That's at the genomic level. They did
19:32say in the Q2 2026 earnings call, they
19:35do that at the other levels of
19:37abstraction that we covered, like
19:38epigenetics, proteomics, and so forth.
19:40However,
19:43Recursion Pharmaceuticals is doing this
19:44with what is known as mass spectrometry.
19:46And this, until Nautilus Biotechnology
19:49came along and actually recently did
19:51something great with their new
19:52technology, this has been, until
19:54recently, state of the art.
19:56The problem with mass spectrometry is it
19:57can't tell the difference between
19:59proteoforms. So, a protein is
20:01essentially just a combination of amino
20:03acids per electromagnetic attractions.
20:06The amino acids take on a specific
20:08shape. However,
20:10mass spectrometry cannot tell between,
20:12for example, an OH group,
20:13oxygen-hydrogen group attached to a
20:16specific location of a protein or
20:18another location. And that actually
20:20changes the function of that protein
20:22variation a lot. Mass spectrometry
20:24cannot do that. Nautilus, with iterative
20:27mapping, and I've actually made a quite
20:29a few videos about this company already,
20:31can do it.
20:32So, here what we have is, again, an
20:35emerging data value chain and we can go
20:36deeper. Nautilus can read intact
20:39proteins one molecule at a time. It
20:41understands perfectly the difference
20:43between one proteoform and another. And
20:45therefore, it gains a level of nuance
20:48that traditionally has been impossible.
20:50It's unblocking, unlocking a data layer
20:53which is foundational to everything in
20:56this emerging value chain that we are
20:57talking about. If you can get an
20:59exhaustive read at the proteomic level,
21:02there's nothing that you can't do in the
21:03body to mediate any physiological
21:05process.
21:07Here's what's very interesting about the
21:08company.
21:09The first assay, the first vertical
21:12read, the first read of a small part of
21:14the proteome took them 5 years. This was
21:16the tau protein. Now, the last one,
21:19oncology, which they're moving into now,
21:21took 1 year. And now they're looking
21:23into months and projecting 20 assays by
21:26mid-2028.
21:28This data layer, I think, is going to be
21:31among the most valuable parts
21:33of the value chain. Because if you can't
21:35read the proteome, you can't really do
21:36all that much, and AI actually doesn't
21:39fully extend. I mean, you can't play the
21:40LEGO puzzle. And therefore, I don't
21:42believe you can unblock the whole health
21:44span thing. Anyways,
21:46there's there's essentially a pathway
21:48called AK-1,
21:50and it's the same pathway that Recursion
21:527735
21:54is um working with. 7735,
21:57remember, is essentially this one that
21:59just received the FDA green light. And
22:02it was designed in 10 months and with
22:04just 242 compounds. However, Recursion
22:07doesn't really have an in-depth read of
22:10of what's happening in the proteome
22:11there. So, how much better could the
22:13molecule get if they do gain a read on
22:15that on the AK-1 pathway, which, by the
22:17way, is the one that I mentioned
22:19previously on oncology? The answer is
22:21Recursion's machine could get a lot
22:23smarter, a lot more predictive, and
22:25actually just turn this condition, AK-1,
22:29it's essentially
22:30a pathway involved in a good number of
22:33tumors. It could just turn this thing
22:34into background noise. If you as a human
22:37have these biomarkers being generated in
22:38the background and AI models working for
22:40you, and then the right function is
22:42seamless and delivered via the
22:44distribution layer, be it Hims, be it
22:46Tempus, be it Oura, be it be it Whoop,
22:49essentially this becomes a nothing. You
22:51know, it just disappears into the
22:52background, and AI truly becomes
22:54predictive for you, and thus extends
22:56your health span. So, therefore,
22:58Recursion is doing very well. It's
23:00already outperforming the broader
23:02market, as we're talking about in this
23:03slide, but it's doing so with a very
23:05thin read of proteomics. So, my bet is,
23:08you know, these guys are exchanging
23:10money now for data. My prediction is
23:13Nautilus essentially and Recursion form
23:16another link in the chain very soon.
23:17Now,
23:18we can trace the emerging data value
23:21chain along separate things that these
23:23companies are doing,
23:24which and you know, maybe that data
23:26value chain hasn't been formalized yet,
23:28but if you look at the biological
23:29advances, then it can be. There's a
23:31thing called PI3K, which essentially
23:35changes stuff in the cell membrane, and
23:37that change
23:39when it goes, you know, when the when
23:40the switch goes uncontrolled, it
23:42basically makes the cell proliferate and
23:45divide and grow uncontrollably. It does
23:47so via the AKT pathway that we talked
23:50about. It switches on the mechanism of
23:52grow and survive, and this absolutely
23:55seems to make cells grow uncontrollably.
23:57So, we have Nautilus
23:59working on the AKT1 proteoform assay,
24:03getting an exhaustive read of what's
24:04happening there with the proteome. Then
24:06we have Recursion Pharmaceuticals with
24:09REC 7735,
24:12which seems to switch off
24:14the jam switch. So, the thing that goes
24:15wrong that makes PI3K trigger this
24:19uncontrolled growth in the cell. And by
24:21the way, it does so with 100x plus
24:24more selective efficiency than does
24:26anything in the market, which is the
24:27reason the molecule is working so well,
24:30and it seems to be far less toxic than
24:32alternatives in the market. This is all
24:34covered in depth. Then,
24:36we have
24:37Tempus sequencing the H1047R
24:41tumors, which seem to be an ultimate
24:43expression of this pathway being
24:45uncontrolled. And you know, they they do
24:47acknowledge that it does emerge from the
24:50PIK3CA pathway, which I believe is more
24:53specific to solid tumors, but it's
24:55essentially a specific version of this
24:58pathway that I'm explaining here. Bottom
24:59line is
25:00these companies are doing things along
25:02the same pathway and they're just not
25:04connected yet. They're naturally going
25:06to be connected because they have modes
25:07that to me are clearly defined. Nautilus
25:10is very hard to replicate at this stage
25:12even though this is pre-commercial and
25:14obviously the financial risk in these
25:15companies I think is meaningful. Less so
25:18as you move up to the distribution layer
25:20which is where I think we're seeing
25:21meaningful traction at the moment. But
25:23this is clearly just an emerging data
25:25value chain and it'll goes back to the
25:28key value pair. For now
25:30the company that I'm most interested in
25:31out of these and it doesn't exclude me
25:33going long any of these companies that
25:35I'm mentioning is Nautilus.
25:37The iterative mapping technology that
25:39they use that I cover in depth in a deep
25:40dive no one's close to.
25:43Recursion efficiency is unmatched and
25:45then Tempus has a strong hold on US
25:47hospitals meaning
25:48I have clear visibility into who is
25:50going to be using
25:52the iterative mapping technology that
25:53Nautilus is bringing to the world. And
25:56they are now starting to commercialize
25:59the platform. I covered this in the Q2
26:002026 earnings reports digest. I believe
26:04I didn't make a video. Essentially
26:05what's happening with Nautilus
26:06biotechnology is they've pivoted away
26:09from releasing into the market as
26:11generic platform that allows you to read
26:14any parts of the proteome you want and
26:16they're now focusing on verticals in
26:17which they have seen they allegedly
26:20claim to have seen real demand from the
26:21marketplace and the science is at a
26:23place where it could actually unlock
26:25meaningful
26:27resolution at the proteomic level. So
26:29this thing is working and my view is
26:32that every condition at some point runs
26:34through proteins. I believe that on a
26:35first principles basis
26:37this is a reality. Of course proteins
26:40emerge from the epigenome which tell
26:42cells what parts of the genome to read.
26:44So you could actually very much change
26:46the proteome just by getting deep into
26:48the genomic layer. But certainly the
26:50proteomic layer is how you read where
26:53this function is manifesting and
26:55perpetuating. And so, this is incredibly
26:57important. I believe that every
26:59condition runs through proteins. And so,
27:01you have Nautilus Biotechnology working
27:03on these various conditions that emerge
27:06from dysfunction in these proteins. So,
27:08you have tau, alpha-synuclein,
27:11AKT1, P53, and EGFR, which are the ones
27:14in the oncology
27:16um data value chain, which are very
27:17important. Insulin for diabetes, cardiac
27:20proteins for heart disease, cytokines
27:22for for inflammation. This is where we
27:25have the thinnest read at the moment. I
27:27don't believe that reading into the
27:29genome is a problem. I think RNA we're
27:30doing pretty well. Protein amounts with
27:33mass spectrometry is actually quite well
27:35solved. What we don't have is proteoform
27:37resolution. So, we have Nautilus, which
27:40at the time of me recording this video,
27:42I'm actually I started doing this before
27:44market open. Yesterday, it was trading
27:47at $170 million. That's the market cap.
27:50They have a thin balance sheet, but I I
27:52have seen management
27:54exhibiting a clear ability to push the
27:56science forward. Commer- um They They've
27:58started to commercialize. Um the first
28:01revenue actually came in in Q2 from a
28:03foundation that they're working in. I
28:05believe it's the tau proteoforms that
28:07they're leveraging.
28:08And they're running lean, too.
28:10So, OPEX is down year over year. There
28:12is meaningful financial risk here. So,
28:14this company is burning $14 million a
28:16quarter and actually expects to raise
28:18capital before mid-2027.
28:21So, there is clear dilution risk. This
28:24is very early stage, but to me, this is
28:27a very appealing company. Right? So, if
28:30management continues executing in a
28:32disciplined manner and just continues
28:34doing what they've been doing, this
28:35valuation for a key components of the
28:38value chain to me seems ridiculous.
28:40If this works, what are we looking at
28:42here? A company valued at least in the
28:44tens of billions of dollars, because
28:45everyone is going to be pulling from
28:47this data layer that no one else really
28:49seems to be able to unlock for now.
28:51>> [snorts]
28:52>> However, you know, I'm long Nautilus,
28:54but my allocation there is micro. I do
28:56continue to plan to build on it, but my
28:59allocation is Hims I have a giga
29:01position, Immunity Bio I have a smaller
29:03one.
29:04It's reasonably sized, but I would say
29:06very much pursuing a level of asymmetry
29:08in which, you know, the whole biology
29:10thing could go up 100,000, 200,000 X,
29:13whatever over the next 5 to 10 decades,
29:15but it's something that I'd be prepared
29:17to have it go to zero. So, it's an
29:19additional layer of asymmetry in my
29:21portfolio. As I said, there are
29:23companies that are also building
29:24appealing components of this emerging
29:26value chain. We have AbCellera, which
29:29has essentially built the discovery
29:31platform once and can really just bind
29:34onto any target onto the body. I covered
29:36that in the video, I believe it was last
29:38week. Immunity Bio, I'm long this
29:39company as I said, has the ability to
29:42reboot the immune system and program it
29:44in any specific direction. And then Mik
29:46Therapeutics, which I'm not long, but
29:48I'm quite fond of that third branch of
29:50the immune system, which has also been
29:52widely abandoned
29:54by the rest of the industry.
29:56So, this is going to be a long journey.
29:58I'm going to continue studying
29:59companies, I'm going to continue
30:00building positions, and I'll let you
30:02guys know.
30:03But the bottom line is that every ticker
30:04adds key value pairs or makes them
30:07sharper. We are gaining a read on
30:09biology, which is going to be probably
30:11at the atom level resolution,
30:13and we're just going to have something
30:15which is going to be far more valuable
30:17than anything we've seen in the world
30:18today. So, you have companies like
30:20Amazon and and Netflix and stuff like
30:22that being pretty valuable just by
30:24letting people buy whatever or letting
30:26people watch whatever. This is the
30:28equivalent, but on the biology side of
30:30things. The risk actually goes up as you
30:33go down the value chain. So, in my
30:35opinion, the distribution layer makes a
30:37lot more sense. If you look at the unit
30:39economics of Hims or Tempus AI, they are
30:41far better. Hims is actually has been
30:43producing producing cash from operations
30:45for a long time. Even free cash flow per
30:48share went up exponentially before they
30:50entered this recent period of
30:52investment. So, the risk goes up as you
30:55look at smaller components of the value
30:57chain, but also does the upside. So,
30:59just to recap today's video.
31:03This superhuman AI that we're building
31:05that's going to extend health spans is a
31:06function of reading into someone's
31:08biology, understanding the treatments at
31:10the atom level increasingly and then
31:12reading the state after. We have all
31:15these tickers that we covered. Butterfly
31:16for organs, Nautilus for proteoforms,
31:19Tempus for tumor DNA and RNA, Hims is at
31:23the distribution level,
31:24but also very much getting a read into
31:26an increasing level volume of biomarkers
31:29via Hims Labs, for example. Grail's
31:31Galleri cancer test, which uh recently
31:34received some interesting encouraging
31:36data,
31:37and uh the stock went up a lot. Then we
31:40have the treatments, you know, we have
31:41Recursion Pharmaceuticals discovering
31:43novel biology and understanding how to
31:46tune drugs to address that biology. XmAb
31:49antibodies and T-cell engagers uh as
31:51verified during the Q2 2026 earnings
31:54report.
31:55Immunity Bio for the IL-15 uh super
31:59agonist to bind onto that receptor and
32:01reboot natural killer T-cells.
32:04And then we have MiNK Therapeutics for
32:06off-the-shelf invariant natural killer
32:08T-cells, which again I did cover in
32:11depth and the science is fascinating.
32:13For the state after again, it's very
32:14much the same. We have Butterfly,
32:15Nautilus, Tempus AI and Hims.
32:18Every every one of these tickers adds
32:20pairs or makes them sharper. So, the
32:22resolution is going up. But again,
32:25the lower you go into the value chain,
32:28I think risk goes up. Eventually,
32:31if I had to bet, the most extreme
32:33rewards will come from the lowest level
32:35participants in the value chain that
32:38also produce unit economics that no one
32:40else can replicate. A candidate to that
32:42tentatively in my opinion is Recursion
32:44Pharmaceuticals. Very much and this is
32:47also the case with Nautilus, although as
32:48I said, risk goes up. But everything is
32:52going to emerge from the proteomic
32:53layer. So if Nautilus is the single
32:55dominant player in that, the rewards can
32:58be extreme, although again, it can go to
33:00zero. So I am certainly allocating
33:03accordingly. But if you look at
33:04Recursion and its ability to uncover
33:07novel biology, that's extremely
33:08interesting as well. So I wouldn't
33:10discard a company because it's lower
33:12down in the value chain, but certainly I
33:14would
33:15bear in mind that the risk does go up.
33:18So
33:19that's it for today's video, quite a
33:21dense one. Uh lots of information to put
33:23together. You have the written form
33:24available in the blog. You can see the
33:27link there on the screen. My positions
33:29are Hims, ImmunityBio, and Nautilus
33:31Biotechnology to date. And as I said,
33:34I'm long these companies in essentially
33:37logarithmically declining allocation.
33:39I'm very comfortable making a massive
33:41bet on Hims. I've been building that for
33:43a long time. And then I'm comfortable
33:45making bets on ImmunityBio and Nautilus,
33:48but in decreasing levels of allocation
33:50to the point that Nautilus is
33:52essentially a micro allocation in my
33:54portfolio that nonetheless, I do plan to
33:57make bigger.
34:06All right, guys, giga video today, lots
34:08of information. I hope that at the very
34:10least this video has transformed your
34:12worldview and opened your eyes to an
34:14emerging value chain which I as I said,
34:16I think it's going to give birth to the
34:18most valuable subscription service on
34:20earth. A big candidate to that by the
34:22way in my opinion is Hims. So thank you
34:24very much for joining me today. As
34:25always, if you enjoyed the deep dive,
34:27could you please like and subscribe and
34:28share this with someone else whom you
34:30think will enjoy it. These deep dives
34:31updates videos are for free. And so, the
34:34only way this grows is with your help.
34:35So, thank you very much in advance. Take
34:37care, and until next time.

About this transcript

This page contains the full transcript of I bought palantir at $7. these 8 stocks are the biology version by Antonio Linares, generated from the public captions YouTube serves with the video. The transcript has 6,133 words across 985 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.