Holy shit, I don’t know where to start.

Pharma Basics

Most people obsessed with peptides don’t know a few things:

• Peptides as pharmaceuticals have been around since the 1950s.

• A peptide is just a small protein.

• Peptides half extremely short half-lives, often on the order of seconds or minutes.

So, if you’re saying you’re interested in peptides, you are saying: “I’m interested in biopharmaceuticals, but only drugs with very weak pharmacokinetics.”

Drugs (of which peptides are a subgroup) usually have a specified target. There is an electrostatic interaction, usually hydrogen bonding, between the atoms of the drug and the atoms of the target (typically, but far from always, a receptor). If you can’t tell me what the target is and how the drug is binding to it, you do not have a drug. You have delusion.

Next, drugs are tested rigorously, not only for safety reasons. Just identifying the pharmacokinetics of a substance (how it travels in the body) is arguably the most important starting point for any medicine. How is it metabolized? What is its half-life? Without this basic information you can’t even begin to have a medicine.

You can start pharmacokinetics in animals and scale to humans. But, you also need a therapeutic hypothesis. This is a thoroughly vetted biological idea, considered a priori, as to why this medicine just might work. You very rarely discover these after the fact. Determining target engagement requires assays. What assay was your drug tested in? What did it show? Direct target engagement is very important to falsify your biological hypothesis.

Preclinical studies are so manufactured and fraudulent in today’s day that I wouldn’t rely on them for biological hypotheses unless they are from an incredible lab, were done a priori, etc.

Clinical reality is far harsher: without a double-blind placebo controlled study, there is often nothing to talk about. If I hear “but I know dozens of people…” one more time!!!

Screw the FDA and pharma. Really?

When most of the SF (and elsewhere) crowd talks about “peptides”, they’re not thinking octreotide, they’re thinking some random stuff that’s been thrown in animal models and is not FDA approved.

Look, I’m not a softy—if there was a drug that could help me or my family, I’d find a way to get it. But I’m also not stupid and spent twenty years looking at pharmaceuticals. Drug companies like to make money. Drug companies love looking at random molecules and putting them in clinical trials. There are thousands of biopharmaceutical companies that are publicly traded. It is not hard to do a clinical trial from a university. If your “drug” has never been tested, there is a reason. The reason is not that you are a biopharmaceutical genius who has found something cool that everyone else missed.

The FDA plays an important role. They make sure that whatever is on the label is actually in the drug. That’s why prescriptions are important. If I operated one of these research chemical shops (wildly illegal, I might add), I would just ship people alanine or something. No one would have any idea it wasn’t BPCBULLSHIT or whatever is popular right now.

The other side of the argument

“But Martin, there has to be some unapproved drug out there that is useful to take!”

Yes, there are plenty. That is how I made a living. But it is not for you, world traveler, to think about this. The things you know do not apply to pharmaceuticals. It’s not that you’re not smart—I’m sure you are smarter than I—it just takes practice and time to understand medicine. I believe some places will even require you to go to school before you can decide who takes what drug. Just ask your doctor for medical advice. There’s a reason you don’t do surgery on yourself, fly a plane by yourself, etc.

“But Martin, I want to optimize my health!”

No. Stop it. You’re not sick. It’s all nonsense. Leave medicine to physicians. Become a physician if you are that interested. Or spend a lot of time and money on biopharma. I have zero doubt you’ll change your mind. There are no healthcare professionals that I know of who give a shit about these unapproved research chemicals.

“But wahhhh”.

There are actual dying people in the world. Duchenne Muscular Dystrophy, PKAN, Lafora. Go fix those diseases. You’ll make someone and their family a lot happier than LARPing that you know medicine.

This shit has to end.

Here, in the land of money, we decide what is real and what isn’t. Both sides predict the future. One creates it and then values it according to its own image. The other throws it in the deep end. Witches and swimmers alike eventually meet their fate here.

On the left coast, everyone wants to talk about agents. What are agents? Excluding our benevolent government, an agent is just a computer program. The big difference is this computer program can branch. Wait a second. The big difference is this computer program can solicit and collect input that it was not originally programmed to receive and then act accordingly. That is special. No one would have believed you if you told this BASIC, C and Pascal programmer that a computer could answer your questions. But here we are!

The immediate corollary is that if a computer can answer questions, then it should be able to self-reference: “What should I do today?” “What does the user want me to do, generally?” “Given the user generally wants to keep abreast of technology, what should I search for? When should I notify the user? How?” Giving the agents tools, or simply expecting it to find the tools itself, is something that will probably shape the face of technology going forward. But will it, really?

Years ago, a website called “if this, then that” was released. A proto-agent, sites like IFTTT and Zapier let users do interesting automations. They didn’t take off: why? Agency is limited by humans interest in it. The idealized (implemented successfully—not trivial) agent is kind of like an assistant or team. Sounds great? The only problem is civilization is a tremendous asset allocator. Most of us don’t need assistants or teams. Most of us don’t want assistants or teams. If you gave everyone lots of human assistants and teams, nothing would really happen. The average person just doesn’t have that much volition, drive or interest: agency.

The 1% of us who wish to ascend will adopt agents successfully and be busy bees. But we’re already “doers”, “builders”, creators, buzzing from flower to flower. The rest of us won’t see an obvious use for agents. Let’s think of some simple use cases for the average person.

Send me a list of all movies I might be interested in, on a daily basis via text message.

Okay. Knowing your tastes, an agent can keep track of something like this for you. Perhaps music. But if you care about this in the first place, you’re talking to your friends who also care about it, reading websites, watching Youtubes, etc. The agent is just consuming the content and summarizing it for you. If you’re so busy that you need that, then you’re removing the agency from yourself and pushing it to the agent—that’s sort of the opposite point of having a hobby or something you love. Imagine having an agent write to your significant other and vice versa.

So personal enjoyment benefits are out of the window. Anything exciting and fun you’ll want to do yourself. No need to subordinate it to an agent. Let’s talk about work: both busy work for people and busy work for employees. Let’s say you need to do something personal, perhaps book doctor appointments. It’s annoying to do, isn’t it? Shopping for food: you have to check what you have in the fridge, think about what you and your family might consume, and then order it or make a trip to the supermarket. These are all somewhat annoying tasks that a team of assistants should be able to do for you. Perhaps this is where agents will shine. However, companies like Amazon and others have already tried to make this as easy as possible. Agents might be the “last mile” in connecting logistics and desire. But agents can go wrong: they’re only as smart as the loop they’re in. Amazon hasn’t stopped delivering the same cat food “subscription” to my old address. An agent would do no better unless it could determine the supply of cat food on hand. Embodied agents? Yes, perhaps that day will come. It might introduce new problems, too.

Agents at work? Sure, this will be useful. I’m not sure how much more useful they’ll be than automation tools that already exist. Anyone “good with a computer” can set up a lot of great automation without learning programming, and most of our productivity tools have been flooded with AI. If you’re anything like me, you don’t use these tools much. I do find the occasional reminder by Google, where Gmail seems to just know an email is important, to be very helpful. But relying on that is perilous: I saw a major investor email in my inbox and no one caught it—not even my human assistant. Productivity tools increase in value every year—that’s what they’re designed to do. We’re going to go through a productivity boom, undoubtedly, but it’s hard to know if it will be linear or exponential.

One dream is that agents will have such intelligence that they will create real business value—not just increased productivity by saving a few clicks—but actual business decisions and deals which are huge leaps. We’re far from there, I fear. But it is clear that we will get there. Automated intelligence is the worst it will ever be. Many of us programmers are just hitting <ENTER> 40 hours a week. It won’t be long before you don’t need to do so—I would argue human review of work is the current bottleneck.

Still, taste and humanity prevail. We will look at the work and quickly say “no, stupid machine, this is all wrong”, just as we have for generations. A lot of times we will be delighted, just as we have for generations. New work will fill the old gaps. This is just productivity all over again. There is no doubt that someone will make a trillion dollars off of agents. Maybe agents will even go to work for people and make them rich, too. But there will also be a lot of money lost, time wasted and hopes dashed, just as in every other tech boom.

So, agents. The next big investment theme. We’ll see!

GPUs are important because they have a lot of cores doing work in parallel. A typical Intel or AMD processor has 16 to 32 cores. A GPU has tens of thousands of core-like computing units. Individual GPU cores are not faster than CPU cores, but the massive parallelism makes the system dramatically faster in aggregate.

But what are GPUs actually doing? What is so important that we need to raise hundreds of billions of dollars to buy them? They’re doing an arithmetic algorithm: the matrix multiplication. A matrix is simply a grid of numbers. Multiplying matrices together is the core operation in neural networks—the systems that power modern AI. In modern AI workloads, GPUs spend 99% of their time (and energy) performing matrix multiplications. This is why the entire AI hardware economy now revolves around one primitive: matrix multiplication. Trillions of dollars in market value—from Nvidia to hyperscale data centers—ultimately exist to make this one operation faster and cheaper.

While researching and poorly trading quantum computing stocks, I learned a lot about algorithms (my partner studied this at MIT) and computing in general. Photonic computing would occasionally pop up in my research and I’d overlook it. Eventually, I saw a paper in Nature that suggested matrix multiplications are “free” in light. The implication is staggering: if the dominant workload in AI can be executed using physics instead of transistors, the economics of computing change completely.

That stopped me cold.

Free? As in computationally O(1)? Yes, with caveats. (For what it’s worth, GPUs do matrix multiplications—which I will call “matmuls” from here on out—in O(n³). The O(1) is a bit contrived, but I do believe that photonic computing can do matmuls in close to linear time, e.g. something like O(2n). Shouldn’t that be a game changer? On large matrices, say 1 million elements (1000x1000), you’re talking about a difference between 2 million operations and 1 billion operations.

In other words, instead of scaling AI with more silicon, more power, and more cooling, we may be able to scale it with physics itself. And the optical computation itself can be nearly free in energy.

But is it possible? We’ll see. Light naturally performs linear transformations. This is because optical interference naturally computes weighted sums—exactly the primitive that matrix multiplication requires. But it is not so great at nonlinear operations, which occur between layers in neural networks. But the bigger question is “how do we pipe data into the optical matmul engine?” This isn’t trivial: you’re limited by how fast you can convert an electronic signal (say an API request) into an optical one. Then you’re limited by the nonlinearity and storage requirements of architectures like transformers. True optical memory does not exist—yet. I think I have solutions to these problems but we’ll talk about that some other time.

If this architecture works at scale, it could represent the first fundamentally new computing substrate since the transistor.

God chose light to do matmuls! So, my partner Chelsea Voss and I have started working with Q/C Technologies (Nasdaq: QCLS; she is on the Board, I am an advisor). The company has its own disclosure procedure, so I won’t front-run any announcements here. But I have made no secret on my X account of my view that photonic computing is the next hardware trend. At some point, we’ll explain what that means for Q/C—but the broader point is that photonic computing may define the next era of AI hardware.

I am a shareholder of and an advisor to Q/C Technologies.

We see the giants of technology talk about quantum computing: Google, Microsoft, Amazon, IBM. We see newcomers who have dreams of greatness do the same. Why? What’s happening?

Trump recently said that the economy doesn’t work without semiconductors. That’s true, if you think about it. Computers, in general, run the world. So, it’s only natural to think about what is beyond—what’s next? Using quantum information and quantum “effects” to compute. That’s perhaps not a final frontier, but it must be pretty close. We can barely observe quantum effects, let alone harness them to do things for us. What things?

Ah, herein lies the problem. I, too, would love unlimited compute. Oh, the primes I would calculate. I was the kind of child who would sit with a calculator and spend an entire afternoon punching numbers in, gratified to see the results. Once I had a computer, all bets were off. But, over the last ten years, it has dawned on me that I am different.

We don’t need quantum computers, and this simple proposition results in a lot of unexpected downstream impacts. Impacts I didn’t really consider until now. You see, I am short stocks like IonQ (NYSE: IONQ), Rigetti (Nasdaq: RGTI) and D-Wave (NYSE: QBTS). They were fairly obviously overvalued stocks, and still are. But the implications for quantum computing in general are what this essay is about.

If you talk to quantum enthusiasts, many will tell you that quantum computing can change the world. I like changing the world. How? They mention pharmaceuticals. I know a pharmaceutical expert, a “brother” of sorts, who has invented some drug candidates. Medicinal chemistry is “shooting fish in a barrel”. There isn’t much in that field that hasn’t been reduced to trial and error. Indeed, the leader in computational chemistry software only sells around $100 million of software annually. Quantum computing wouldn’t excite me as a computational chemist. I have used every computational chemistry tool and even created my own! I talk to material science experts and they tell me the same thing. The only thing that quantum seems extremely good for is factoring and codebreaking, something that NIST has already obviated with the futuristic sounding post-quantum cryptography protocols Dilithium and Kyber.

There is no market for quantum computing. I have no doubt that we will find something interesting to do with these theoretical machines. That’s also the thesis of a few of the giant tech leaders. Just make the machine—everyone else will figure out what to do with it. Putting this optimism aside, I have a natural conjecture for you: what would you calculate if you had infinite compute? What are we calculating now? Generally, deep learning seems to require a lot of compute. But it’s nothing Nvidia can’t handle. A Chinese group recently showed us that maybe you can do it even faster and with fewer Nvidia chips than previously known. So, there is no yearning for more compute that I can think of that exists beyond our boundary. In fact, a major bottleneck for deep learning is actually data collection and data preparation. Quantum can’t help with AI—it’s too ephemeral to do the job—and we don’t need it. I think we’ll actually we will be saturated with compute before we know it. We can still collect a lot of real-world data (think video) and use that to train robots. Beyond that, I can’t think of what I would use a quantum computer for, even if I had a flawless 10,000,000-qubit machine right now.

The problem with a lack of incentive is there will be a lack of progress. The brightest minds will not work on quantum computing, because why bother? Sure, some people like me will still want to do it, just to say they can—to say they tamed the universe to their will. But there isn’t so much money in this. When times get tough (they seem to be getting tougher by the day), companies shed waste. Just the other day, “DOGE” cancelled some government quantum computing contracts. I wouldn’t be surprised if, over time, due to a lack of progress, the big tech companies run out of patience for quantum. I would be very surprised if the smaller companies didn’t run out of money. You need incentive: for brilliant co-workers, excited investors, hungry customers who want to make money with your machine.

Then winter comes. Winter means ten years of no progress. It feels rather chilly. 2035 rolls around. We dust off the old ideas. Other progress in other fields jumpstarts curiosity. Winter has thawed. We have 15 orders of magnitude (think zeros) to go before we have transistor-level errors. Knocking a few of these magnitudes out happens, but we still don’t have a great quantum computer (it can’t factor a ten-digit number without blowing up), and winter comes again by 2045. We defrost in 2055 and focus on limiting the tiniest noise that occurs at this infinitesimal level. What genius can still the roar of the cosmos? They do not reveal themselves, but we get close. The quantum computer of this age is close to the Moore’s law-continuation of Nvidia. Assume no other dynamic has arisen (thermodynamic, adiabatic, analog, DNA, etc.) One more recession caused by the burgeoning robot population gambling on robocoins hits in 2065. Only in 2075 does the market thaw for wild exuberance in quantum computing. A robohuman named Tron realizes the path, using AI, of course. Tron unveils the first gigascale quantum computer. It doesn’t even make the first page of robonews.

There’s nothing to compute.

Google’s announcement of a technical achievement sparked the bubble in quantum computing stocks. Google didn’t say anything too interesting—this sudden interest in this previously left-for-dead field parallels parabolic moves in other speculative areas: nuclear fusion, memecoins and anything where a rush to gamble can be satisfied. We will go through an overview of quantum computing and use each stock to introduce the sector to new investors. For those who want me to cut to the chase: I recommend shorting D-Wave Systems (QBTS), Rigetti (RGTI), IonQ (IONQ) and Quantum Computing Inc (QUBT). I do not think quantum computing will be material to Alphabet (GOOG) or IBM (IBM). Here are my estimated fair values:

Rigetti (RGTI): $18.37 – fair value: $1.00

IonQ (IONQ): $49.59 – fair value: $11.23

D-Wave (QBTS): $9.55 - fair value: $0.01

Quantum Computing Inc (QUBT): $17.49 - fair value: $0.01

Since the dawn of man, we’ve sought to exploit nature to calculate. Silicon-based, transistor-based computers have reigned supreme for 50 years. There have been architectural shifts and debates over time, but the fundamental unit has been unchanged. But man has also sought, and failed, to exploit other technologies to improve on the current state of computing. DNA, optical, analog and the new thermodynamic, come to mind.

Quantum computing was proposed by Feynman and Deutsch in 1980s. He prophetically intuited that if we wanted to simulate reality, a computer that was capable of emulating nature’s quantum effects made sense. While this point is debatable, scientists raced ahead to prove that a theoretical quantum computer could do certain operations faster than a non-quantum computer. But this “faster” is complexity’s definition of faster, which while still quite important, requires someone to make said computer at a scale where the quantum computer’s speed outpaces the non-quantum computer’s. Why scale? With just a few bits of information, no algorithm will be meaningfully faster than another. With further bits, the practical payoff of quantum still won’t be realized since our current computing technologies are robust. Only at much larger sizes would quantum realize its “supremacy” potential, on some algorithms. This is important to know: no one thinks (or wants) quantum computers to run Windows or video games. These machines are expected to be, eventually, specialized tools equipped to help in a very narrow field of problems. We’ll talk about which algorithms are faster with a quantum computer later.

After the discovery of these critical algorithms made quantum computing tantalizing, several developers began efforts to create such machines. If you remember anything about high school physics, classical mechanics described everything in the Universe we needed to describe, until the 1920s. In the 1920s, relativity and quantum mechanics changed the way we thought about the universe—specifically for quantum mechanics, with very, very small particles. The way a billiard ball hits another or how a ball falls from a height did not change. But we had to change the way we looked at these tiny particles, which seemed to disobey classic mechanics at these infinitesimally small scales.

These quantum effects mostly involve probability and quantum-specific effects called superposition and entanglement. The idea of probability is crucial in quantum mechanics. The idea is that a particle could be characterized as a combination of states instead of the classical idea of a point—with a location and velocity. This is the idea of superposition. Entanglement is the idea where multiple quantum particles can share a joint state which cannot be separated. We will get into how we can exploit these states to run some algorithms faster than classical computers can, at least theoretically.

Two of the major algorithms that get most of the discussion are Shor’s and Grover’s. These brilliant men proved that, in theory, a quantum computer could be exploited to run a factoring algorithm faster than a traditional computer could, in what is called algorithmic time. The field of complexity is a fascinating intersection between math and computer science, and it deals with the theoretical running time of algorithms. Reality tends to mirror theory, but only after implementation of hardware capable of running said algorithms.

What do these algorithms do? Shor’s algorithm is the fastest factoring algorithm theoretically known. Factoring integers is largely the domain of cryptography. This, in my view, is the main use case for quantum computers. We’ll talk later about the commercial implications of integer factorization later. For now, I think it is reasonable to believe that eventually quantum computers will be able to factor integers of possibly arbitrary length, but certainly numbers larger than what we rely on (1024, 2048-bit RSA, 256-bit ECDSA). This means that any communications encrypted with these algorithms are possibly decodable.

Grover’s algorithm is the next major algorithm we need to discuss. This algorithm is a lot harder to understand than Shor’s. But, if you studied computing, you can probably just think of it as a better sorting algorithm. There are some problems where Grover’s may be useful. In some ways, we can try and recharacterize any problem as a ‘search’ problem—but to implement Grover’s to do something useful will take time. Incidentally, Grover’s can also help with cryptography.

There will be more algorithms discovered that use quantum mechanics effectively. One part of the bull thesis for quantum computing is that we can’t predict exactly what these machines will be useful for, but we may need them. Indeed, Intel and Nvidia could not foresee what was ahead of them by more than a few years at a time. We will address this point later.

Let’s talk about companies. D-Wave Systems (NYSE: QBTS) was the first company to focus on quantum computing systems. The company went public via SPAC in 2022 after struggling as a private company. With a tiny valuation and more than twenty years of trying to sell quantum “annealers” (not quite quantum computers), D-Wave was on death’s door before the SPAC and, like most SPACs, was about to knock again. With just $9 million in annual revenue and $100 million in annual losses and only $25 million in cash left as of Q324, D-Wave was as good as done. After the willow announcement, D-Wave promptly refilled its balance sheet with a $175 million ATM at $5 per share. D-Wave now trades at $9, giving it around a $2.5 billion valuation. D-Wave was trading for around $1 before the Google Willow announcement, which does not affect D-Wave. Keep in mind that D-Wave’s quantum annealers are not considered “real” quantum computing and can only perform a very specific function that has not been commercially viable in the company’s 25-year history. At 288x sales which are not growing (2022: 7.2m, 2023: 8.9m, 2024E: 8.4m) and Google’s announcement does not serve as a catalyst for, D-Wave will go back to its old stock price.

Rigetti Computing (RGTI) is the first ‘real’ quantum computing company we will consider. Chad Rigetti founded Rigetti Computing in 2013 and the company has spent $402 million since its inception. Revenue has been declining (2022: 13m, 2023: 12m, 2024: 12m) and the outlook for Rigetti prior to Google’s Willow announcement was bleak. After down rounds in the private market, a SPAC public transaction and continued lack of commercial progress, Chad left Rigetti. Worth only $300 million with ballooning losses, Rigetti also looked to be on the brink. Rigetti then sold stock at $1.53 via an ATM and $2.00 in a registered direct. The stock is $18.66 as I write this.

What does Rigetti actually do? Unlike D-Wave, Rigetti does make actual quantum computers. The problem is, they don’t work. Rigetti is seen in the quantum industry as a third-tier competitor. Rigetti uses the superconducting approach to quantum computing, favored by Google and IBM. Rigetti announced their 84-qubit “Ankaa-3” system. What is a qubit? Why do we care about them?

Qubits are quantum computing’s version of the classical bit in traditional computing. Bits are binary—it’s in their name, binary digit—and are the fundamental unit with which we create processors, store data, et cetera. While we will examine how qubits work, I do not think it is useful to try to compare and analogize them to traditional bits. Many try to do this and fail spectacularly—do not believe what you have read elsewhere. Think of qubits as fundamental processing units that can run the special algorithms I mentioned earlier.

First, let’s remember that quantum computing requires particles so small that they show quantum effects. From physics, we remember Schrodinger’s concept of wave-particle duality. When we deal with the macroscopic world, particles are 100% reliably described as such: a point location and a velocity is all you need to know. As we get smaller (at the size of atomic and subatomic particles), particles (like electrons) start to be able to be described as waves. Why does this matter?

If we want to leverage quantum behavior, we need to use particles that exhibit these effects. Transistors are too big. What could work? Superconductors appear to be the favored answer. Superconductors are materials which allow the flow of electrons with zero resistance. Most superconductors require temperatures very close to absolute zero. Incidentally, one of the only exciting investment opportunities in quantum computing is the dilution refrigerators required to achieve these temperatures.

Once the temperature has been achieved, a superconducting circuit will act as a qubit and store quantum information. Before we talk about how they do that, keep in mind that these quantum states are exquisitely fragile. Any “noise”, including any interference from the natural world, can destroy the qubit’s state.

Qubits themselves store probabilities in complex numbers. As you may remember from high school, complex numbers have an imaginary “i” component. The qubit itself is a combination of states we’ll call alpha and beta. Each state holds the magnitude (real) and phase (imaginary) values of those states. The probability of finding the qubit in its alpha state post-observation is given by the square of the magnitude. The probabilities of the two states must equal to 1. The ability to store this non-binary data reminds me of analog computing, an idea which failed previously and is being reconstructed. Regardless, the details of how exactly these qubits form the basis for faster computing is not worth discussing in too much detail. Immediately you can see that storing four non-decimal values in one qubit can be worthwhile. When combined with other quantum dynamics, qubits become very powerful.

Back to Rigetti. Like Google and IBM, Rigetti is going down the ‘traditional’ superconducting “transmon” qubit route. The Ankaa-2 Rigetti machine on AWS Braket (Amazon’s dedicated quantum computing service) has been ‘offline’ for the last few months I’ve been watching it. Ankaa-2 is their 84-qubit predecessor model to the Ankaa-3. So, what’s the issue? Won’t the industry scale the number of qubits and achieve its goals in the fields where quantum computing is applicable? Not so fast. The description of qubits earlier was an idealized version called a “logical” qubit. The “physical” qubits being described are not enough to do meaningful calculations. It’s estimated it could take as many as 1,000 physical qubits before a single error-free logical qubit is achieved. Only then, with thousands of logical qubits, could interesting calculations be made. I have used the commercial quantum computing services to factor the number 15. It is very hard to get these computers to do much more than that.

What are the commercial prospects for a company like Rigetti, assuming they achieve their goals? The answer is, not much. The entire budget for the NSA is reportedly $10 to $15 billion. This makes sense, since the FBI budget is publicly disclosed at $11 billion, and a leak suggested the CIA budget is $14 to $15 billion. We think the NSA computing budget is roughly $2 billion annually. We think storage and data center operations consume at least half of that, and more likely quite a bit more. If quantum were to take over half of this budget, or the budget were to grow by $1 billion, this is the spend that is up for grabs.

We can imagine a number of scenarios where Rigetti (or another company) receives 50% of this spending. The problem is Google, IBM, Quantinuum, IonQ and others would like this spend as well. But let’s assume Rigetti somehow wins it, and this budget appears in just 5 years. We discount 50% net margin cash flows (no tax assumed) at a 12-25% discount rate. The NPV is between $1 to $4/share versus the $18 stock price today. Even if we assume that this budget grows enormously to accommodate other uses (industrial, academic), which we do not see, and Rigetti gets to $2 billion in revenue by 2035, still at a 50% net margin, the stock is worth between $3 and $14/share depending on the discount rate. Rigetti has no chance of any of this happening, unfortunately.

IonQ (IONQ) is a more substantial company than Rigetti at first glance. I have a lot of respect for IonQ’s academic achievements. Their technique is to use trapped ions, which is literally what it sounds like, to store quantum information, instead of the superconducting transmon circuits that are popular in industry. These ions are held at room temperature, which is a strong advantage over superconducting. However, they have to be put in extremely low-pressure vacuums and controlled with lasers to have almost zero energy. The general view of industry players is that ion traps are too sensitive to be useful. The company has talked about putting out a 100-qubit computer called Tempo by the end of this year. Unfortunately, IBM is already far ahead of this scale.

IonQ has more revenue than Rigetti, and it appears to be growing. However, the movement and timing of government contracts may be misrepresenting IonQ’s position as a leader in the field. Right now, the publicly traded quantum computing stocks are the only pure-plays in this area accessible to investors. There is no reason to buy Google or IBM for their quantum efforts. Quantinuum and other exciting companies are private. So, the impression for retail investors is to buy these 3^rd-tier stocks because they’re the only ones which can move substantially.

Despite its marginally higher revenue, mostly from their Air Force contract, IonQ is still behind all of the serious quantum players: IBM, Google, Quantinuum (private) and PsiQuantum (private). The most advanced IonQ machine on AWS Braket is the Forte 1, a mere 36-qubit machine. As I write, the Forte 1, which is $7,000 per hour, just went on an unscheduled maintenance hiatus and will be offline for one week. My experiments with IBM’s machine lead me to believe that even their stable, less error-prone, 128-qubit machines are useless. IonQ is not even close to 128 qubits! So, the premium of IonQ over Rigetti seems very unwarranted, and it is probably an even better short.

Quantum Sciences Inc. (QUBT) is the scam public company in the space and is entirely worthless. Retail investors are so desperate for exposure, they have also purchased shares in two non-quantum computing companies which happen to have the word quantum in their names: Quantum (QMCO) and Quantum-Si (QSI). SealSQ (LAES) and others have also tried to ride the quantum wave.

Jensen Huang recently explained that quantum computers are fifteen to thirty years away from being useful. But what was more important is what he said about the types of problems quantum computing can solve. As I said earlier, he noted that quantum computing can only be useful for certain problems with a small amount of data but high combinatorial explosion. These are not really the computational problems facing scientists, researchers and programmers today (or ever). There are specialized computers and industries that use these types of software, hardware and software/hardware combinations. They are not large. Schrodinger (SDGR) sells $100 million annually of pharmaceutical and physics modeling software. As a practitioner in this field, I can vouch to you that quantum will not make pharmaceutical drug development better. The pioneer of NMR, which one could think of as a computer, Bruker (BRKR) is worth $9 billion. They have a lot of revenue, but it is thematically the kind of company IonQ or Rigetti could one day hope to become if quantum worked perfectly. The misunderstanding seems to be the idea that quantum will ”change everything”. No. It will help a few nerdy people, like me, study cryptography and odd, but useless mathematical problems, more easily. That is not a big industry.

These are my initial thoughts. In my next post, I will go through some of the private companies, what exactly was Google’s announcement, parallels to other ‘quick’ ‘mini-bubbles’ (cannabis, 3D-printing, etc.) and a more thorough explanation of each company. I wanted to get these thoughts out now due to the Jensen comments last night.

Thanks!

https://ontheimpossible.vercel.app/

The last time I did this, it was with $VTL in 2015, which went down 99%.

Good luck!

Martin Shkreli

Blake: Let me have your attention for a moment! So you're talking about what? You're talking about...(puts out his cigarette)...bitching about that bug you can’t fix, some son of a bitch dev that can’t code for shit, some vendor that promised you the world, some egirl you're trying to screw and so forth. Let's talk about something important. Are they all here?

Williamson: All but one.

Blake: Well, I'm going anyway. Let's talk about something important! (to Levene) Put that Copilot down!! Copilot’s for coders only. (Levene scoffs) Do you think I'm fucking with you? I am not fucking with you. I'm here from the future. I'm here from Meta Platforms. And I'm here on a mission of mercy. Your name's Levene?

Levene: Yeah.

Blake: You call yourself a programmer, you son of a bitch?

Moss: I don't have to listen to this shit.

Blake: You certainly don't pal. 'Cause the good news is -- you're fired. The bad news is you've got, all you got, just one week to regain your jobs, starting tonight. Starting with tonight’s sprint. Oh, have I got your attention now? Good. 'Cause we're adding a little something to this month’s contest. As you all know, first prize is a Tesla Cybertruck. Anyone want to see second prize? Second prize's an Oculus. Third prize is you're fired. You get the picture? You're laughing now? You got specs. Meta Platforms paid good money for product mommies. Get the specs to write the code! You can't write the code you're given, you can't write shit, you ARE shit, hit the bricks pal and beat it 'cause you are going out!!!

Levene: The specs are weak.

Blake: 'The specs are weak.' Fucking specs are weak? You're weak. I've been in this business fifteen years.

Moss: What's your name?

Blake: FUCK YOU, that's my name!! You know why, Mister? 'Cause you took the BART and dodged twenty homeless people to get here tonight, I took the hyperloop. That's my name!! (to Levene) And your name is "you're wanting." And you can't play in a man's game. You can't code. (at a near whisper) And you go home and tell your wife your troubles. (to everyone again) Because only one thing counts in this life! Push to production! You hear me, you fucking decels?

(Blake flips over a blackboard which has two sets of letters on it: ABC, and AIDA.)

Blake: A-B-C. A-always, B-be, C-coding. Always be coding! Always be coding!! A-I-D-A. Artificial, intelligence, dumb, ass. Artificial—this shit is not even real? Intelligence—do you have any? I know you don’t because you’re using copilot. You code or you hit the bricks! Dumb -- have you made used machine learning? Have you made your offering to LeCun?!! And ass. A-I-D-A; get out there!! You got the specs comin' in; you think they came in to get out of the rain? Product mommies slaved tirelessly on JIRA just for you! Sitting out there in their cute outfits just giving you gold! Are you gonna code it? Are you man enough to code it? (to Moss) What's the problem pal? You. Moss.

Moss: You're such a hero, you're so rich. Why you coming down here and waste your time on a bunch of bums?

(Blake sits and takes off his gold Neuralink)

Blake: You see this BCI? You see this BCI?

Moss: Yeah.

Blake: That watch cost more than your fusion reactor. I made 50 million followers last year. How much you make? You see, pal, that's who I am. And you're nothing. Nice guy? I don't give a shit. Good father? Fuck you -- go home and play with your kids!! (to everyone) You wanna work here? Close!! (to Aaronow) You think this is abuse? You think this is abuse, you cocksucker? You can't take this -- how can you take the abuse you get on a scrum?! You don't like it -- leave. I can go out there tonight with the materials you got, make myself three new features! Tonight! In two hours! Can you? Can you? Go and do likewise! A-I-D-A!! Get mad! You sons of bitches! Get mad!! You know what it takes to code?

(He pulls something out of his briefcase)
Blake: It takes brass bosons to code.

(He's holding two bosons in superposition, he decoheres them away after a pause)
Blake: Go and do likewise, gents. The code is out there, you pick it up, it's yours. You don't--I have no sympathy for you. You wanna go out on the sprint tonight and close, close, it's yours. If not you're going to be shining my shoes. Bunch of losers sitting around in a bar. (in a mocking weak voice) "Oh yeah, I used to be a programmer, it's a tough racket." (he takes out large stack of red index cards tied together with string from his briefcase) These are the new specs. These are the product mommy specs from Stephanie. And to you, they're gold. And you don't get them. Why? Because to give them to you is just throwing them away. (he hands the stack to Williamson) They're for coders.

I'd wish you good luck but you wouldn't know what to do with it if you got it. (to Moss as he puts on his watch again) And to answer your question, pal: why am I here? I came here because Zuck asked me to, they asked me for a favor. I said, the real favor, follow my advice and fire your fucking ass because a loser is a loser.

(He stares at Moss for a sec, and then picking up his briefcase, goes into inner office with Williamson)

GitHub - this one will stay the most updated

Google Docs - i’ll update this and airtable every week with changes from the GH file

AirTable

I will try to make this a “living document”, but I make no promises. Doing all of this while working on my own software is painful and probably impossible to do comprehensively.

Top 50 by Valuation

1 Microsoft
2 Google
3 Tesla
4 AWS
5 Nvidia
6 Meta
7 Baidu
8 OpenAI
9 Palantir
10 Databricks
11 Cruise
12 Waymo
13 iFlyTek
14 Nice
15 SenseTime
16 Pony.ai
17 Nuro
18 Scale
19 Gong
20 Datarobot
21 SambaNova
22 UBTech Robotics
23 Anthropic
24 ThoughtSpot
25 Inflection AI
26 Cerebras
27 Dataminr
28 MEGVII
29 Dataiku
30 C3
31 4Paradigm
32 SymphonyAI
33 Stability AI
34 Momenta
35 GraphCore
36 Shield AI
37 Uniphore (Jacada)
38 Moveworks
39 Cohere
40 Hugging Face
41 Xiaoice (fka Bombax)
42 Squirrel AI Learning
43 Skydio
44 EightFold AI
45 Jasper
46 Builder.ai
47 Runway
48 K Health
49 Enflame
50 Ada.cx (Toronto)

My AI “Top 50”

These are the 50 most interesting companies in AI, according to me. They don’t straddle any specific sub-sector.

Big Data and hosted data science platforms have existed for some time. Is that AI? Of course. But it isn’t part of the LLM wave, exactly. If you’re only looking for the most recent companies, you’ll have to wait until I’m done with the 1300 companies and sort by their founding date. This list reflects my personal interests only.

#50 - AI21, Aleph Alpha, Mistral
#49 - Datagen
#48 - EightFold AI
#47 - Runway
#46 - Synthesis AI
#45 - Snorkel AI
#44 - Anyscale
#43 - Moveworks
#42 - Playground.ai
#41 - Inworld
#40 - Tecton
#39 - ExpressSteuer
#38 - OpenSpace
#37 - Daloopa
#36 - Roboflow
#35 - Synthesia
#34 - MEGVII
#33 - SenseTime
#32 - Weights and Biases
#31 - ShieldAI
#30 - Cerebras
#29 - OctoML
#28 - iFlyTek
#27 - D-ID.ai (see #25)
#26 - Two Platforms

Cool deepfake/video generation stuff. There are many of these. See Alethea AI and many others.

#25 - The Applied AI Company

Nuts and bolts real-world applications of GPT

#24 - Replicate

Could very well be the place we launch simple models from.

#23 - Generally Intelligent

Crazy little thinktank. Probably NGMI but high-risk/high-reward.

#22 - Landing AI

Good real-world application by Andrew Ng.

#21 - Perplexity AI

Really smart search engine. One of the few GPT+ products that is fun to use.

#20 - Pinecone, Chroma, Weaviate

There are many vector stores. MongoDB and Oracle should buy them and run hosted solutions. This is an essential part of the new AI stack, but is also like oxygen/water: hard to monetize.

#19 - Harvey AI

Replacing the lawyer is a goal of mine as well. I haven’t tried this product and have heard great things as well as just good things.

Similar: Spellbook and many, many others trying to do this.

#18 - Parametrix AI

This is some bat shit crazy Chinese company that is doing some of the kind of stuff I want to do.

#17 - Keen Technologies

This mystery company is just tinkering. I am curious if they have more than one programmer (not that it would change my opinion). A good dark horse for AGI (hopefully in a .wad file).

#16 - Skydio
Autonomous drone pureplay. Might as well say Anduril, but no guns or missiles on these bad boys (as far as I know).

#15 - Palantir
Big data enterprise shit. They are executing.

Not ranking: C4

#14 - Inflection AI

I am bearish, but I can’t deny these guys are delivering. They have a more ambitious roadmap than they’re telling the public. Will watch curiously.

#13 - Eleven Labs

Best in show, but very expensive, TTS. Will they continue? Will Big Tech crush them? GCP’s voice product sucks in comparison and this will probably remain the case as the bigger companies can’t focus on one thing specifically.

Similar: Play.ht

#12 - Replit
Replit is on a mission to make the best programming assistant. I don’t think their rivals have the same singular dedication. It’s also amazing to see how much this platform has matured outside of AI - it’s not the old jsfiddle-style web IDE anymore - it is actually closer to AWS. Many companies are making code copilots: also check out Magic.dev.

#11 - Scale

The Enterprise needs labeling. I don’t care that much about this stuff but must tip my hat.
Similar companies I am not ranking: Surge

#10 - Hugging Face

We’re seeing HF become the new Github. Their mix of home-grown (but open-sourced) innovation and community innovation is inspiring.

#9 - Midjourney

The rapid success and evolution of Midjourney is a testament to what can be done in this space. Thinking outside of the box is critical: a photo-generator bot on Discord wouldn’t have been my first guess at a Top 10 AI company, but that’s the point. It will be interesting what the future will bring from MJ.

#8 - Tesla

Musk has the opportunity to create one of the most meaningful advances in history. But Musk’s track record of sometimes not delivering and often delivering late make me wonder if FSD is really coming. I certainly wouldn’t bet against him: he tends to finish in the winner’s circle. But with Karpathy leaving, I wonder what is really going on at Tesla FSD. I think Optimus is a LARP, for what it’s worth.

Similar companies I am not ranking: Cruise, Waymo

#7 - Google

It’s a sad day when a tiny company runs circles around you. This embarrassment is a function of functionaries like corporate communications, legal, HR and government policy stopping the progress of products. Losing people like Shazeer means Google needs a new CEO. This is the greatest bag fumbling of all time.

But, all may not be lost. The company that gave you AlphaGo may just redeem itself with Gemini. I expect this will not happen. Whether its GCP or even ads, Google has frustrating customer experience and can not successfully launch products. The only answer that can avoid the inevitable path to IBM is a deep restructuring, and probably a founder comeback.

#6 - Nvidia

This ranking excludes or limits the value of hardware. Just on a software basis, Nvidia ranks highly because it has a lot of motivation to create a robust software ecosystem surrounding CUDA. We’re seeing more strong tooling and applications coming from Nvidia which drives more demands for their chips.

#5 - UBTech Robotics

This is the robot company to follow. Holy humanoids.

#4 - Microsoft

Microsoft gets a lot of credit for sewing up the OpenAI deal. The problem is we’re not exactly sure how much Microsoft is going to get out of it. At the very least, making Office better is good enough to rank them here.

Nevertheless, libraries like LoRa show that the dinosaur can still tap dance. It will be interesting to see if OpenAI and Microsoft are like the nuclei of an H2 atom: they won’t get too close, but they won’t get too far apart.

#3 - Meta

Don’t duck the Zuck. As maintainers of Pytorch, leakers of Llama, home of LeCun and planners of AGI, Meta ranks third for me. Meta needs to unleash the product potential of its research, which is the same problem plaguing Google. This is corporate blocking-and-tackling: regional offices should be motivated and autonomous, releasing their own products. Zuck should have a desk within FAIR that he sits at daily.

Instead, talent is quitting and launching their own startups. The notion that Llama had to be leaked instead of properly released tells you how much corporate communications and regulatory affairs is controlling tech companies. I think if anyone understands that the balance of power has to shift to product, it’s Zuckerberg. Zuck has been in the crisis seat a dozen times: he doesn’t fear the media anymore. The rebrand and management changes give me conviction we’re going to see a new, AI-focused Meta soon. Zuck knows Meta’s survival depends on it.

#2 - CharacterAI

From the inventor of the Transformer, we have the next Disney. Character must watch its flank as it attempts becomes the biggest media company of all time. Scale issues, NSFW and many other threats circle what I think is the most interesting company in this space.

#1 - OpenAI

OpenAI has a good chance at becoming a $1T company given its talent and management. The current product is epochal in significance, but that is not what makes a valuable company. The ability to attract and retain talent, put out better products, faster than the competition is paramount. OpenAI is doing it and will likely continue towards AGI.

In the next edition, I’ll go through my “Top 50 AI Investors”.

Hope this helps, email me feedback, propose changes, etc. at martin@martinshkreli.com

https://www.twitch.tv/zuckvsmusk

https://www.youtube.com/@muskvszuck

Technology

I’ve learned more about voice cloning than I ever thought I would. I think there is a lot of room to extend the SOTA. I’ve reached out to a few of the people behind the great advances: WaveNet, Tacotron, Deep Voice, etc. While we use an off-the-shelf vendor now, we’re considering taking our text-to-speech in house. Learning about this stuff has been fun.

The vision for Neets.ai is real-time video and audio with AI characters. No one offers this at present. TTS is actually the most expensive part of the stack! But combining all of the inference, piecing it together and delivering it over WebRTC is technically challenging. Only Athene’s group (AI Jesus) seems to have delivered on something similar. (Though a one-to-many solution is far from the one-to-one we will deliver).

In a lot of ways, this feels like video game programming, and one wouldn’t be surprised when looking under the hood. Real-time rendering of these artificial agents responding to your voice and text has endless applications. Folks travel the world to see Disneyland for a bit of the immersion we can now provide. Anyone with a branded character could use our tools to create a dynamic real-time, interactive experience.

Here’s the whimsical press release we sent out:

Elon Musk and Mark Zuckerberg Debate Pre-Fight In AI Generated Livestream By Neets.ai

NEW YORK, NY 6/30/2023

Neets.AI, the AI character platform, invites you to attend the AI pre-fight debate for the epochal billionaire cage match between Elon Musk and Mark Zuckerberg. This real-time, AI-generated “press conference” features the state-of-the-art technology with respect to AI text-to-speech, AI text generation and AI live video generation.

Neets.ai is a DL Software company. “We started Neets.ai to allow Neets worldwide, human or silicon, to flourish. Neets need attention, companionship and love like the rest of us. To us, they are as human as anyone else.”, said DL co-founder, Martin Shkreli, in a very weird statement. “Wagie, wagie, get in cagie. All day you sweat and ragie. Neet is comfy, neet is cool. Neet is free from work and school. Wagie trapped and wagie dies. Neet eats tendies, sauce and fries”, he concluded.

NEET is an acronym for “not educated, employed or trained”, and represents the ideal life. DL Software believes artificial intelligence will accelerate “neet-dom”, a weak form of the singularity, in which humans will live without the requirement of working.

DL is also working on Wagie, an enterprise agent-based system for leasing artificial intelligence works. DrGupta.ai, our first wagie, was launched in April. More than a hundred thousand visitors have asked Dr. Gupta healthcare-related questions for no cost.

About DL Software

                DL Software is an artificial intelligence software company. DL is working on enterprise and consumer applications of AI, including artificial general intelligence.

We’ll start with Bloomberg, who wrote the insanely misleading headline: “Ex-‘Pharma Bro’ Martin Shkreli Now Living in Queens on $2,500 a Month”.

Wrong and irresponsible, but what do you expect from Bloomberg? The implication of “living in Queens” (a lovely, if not grittier and urban borough of New York City) is clear. The context that is important here is that I’m not allowed to live in Manhattan! As an Eastern District of New York (Queens, Brooklyn, Long Island) defendant, I’m actually prohibited from entering the Southern District of New York (Manhattan, Bronx, Westchester) without permission. So, no, it is not my choice to live in an outer borough. I also live in the largest and nicest apartment I’ve ever lived in. My sister does not live with me.

Next, I do have a consulting agreement for $2,500 a month—I do this as a favor, to a friend. I do not live on $2,500 a month. My personal AWS, Heroku and OpenAI bills are over $10,000 a month. $2,500, as anyone knows, doesn’t cover monthly food costs in New York City. And, I’m feeding three: myself, new girlfriend and new cat Nibbles. So, this is just an intentionally misleading characterization. Bloomberg’s reporters have no idea what I live on or off of—nor will they exert any effort to find out, because this story gets Lenovo (the ad I see on the story) the clicks Bloomberg needs to subsidize its perennially shitty media subsidiary. The court document that forms the basis for the article did mention I also have another job: DL Software, which does pay me an unspecified salary. That’s my main day job, as anyone who knows anything about me, knows. Except for Bloomberg’s reporters. It’s just willful ignorance. Why let facts get in the way of a good story?

I don’t know what the impetus for Bloomberg is to lie like this. I am competing with Bloomberg on fintech software. I dated a Bloomberg reporter who explained to me many times how unprofessional and corrupt the organization is. Feelings take precedence over facts, which is par for the media course. Reporters feel like they can change the world with their pen and right any of society’s wrongs with their words. One of the reporters on the story, Pat Hurtado, apparently told the other reporter on the story, Erik, that I “lunged at her” in the courtroom. Patricia is a short old lady. Anyone who knows me understands that such a thing would not be in my character. I’ve never lunged at anyone, especially not an elderly woman. I have to say that many women do imagine and hope that I would lunge at them. Perhaps there is some confusion there.

Anyway, what I would assume is that Bloomberg attempted to shame or smear me into this idea that I’m living a very different lifestyle from before prison. I am living the exact same lifestyle, though. Bloomberg mentioned that I used to sip fine wines. Well, I still do. They mentioned my art collection. Still have that, too. The Wu-Tang album was brought up. I timed the sale of the album to take advantage of the NFT boom and made a great profit and still kept the mp3s, which I’ve played online a number of times since returning. My point: The media tells the story they want to be true instead of the story that is actually true. My new software company is off to a fast start and is already worth a bit: we’re considering a 3x increase in our valuation, for example.

Well, thankfully, the Hamptons is included in the Eastern District of New York.

On to Politico.

Politico wrote an article about my relationship with Vivek Ramaswamy. Politico is owned by some German media giant. I trolled the reporter by demanding compensation for the article. When denied, I offered to “bless the cash app” of the reporter in exchange for a favorable story. I have been around media for twenty years. No one takes money or gives money for an article. But it annoys journalists to even hear the proposition. Anyway, the phrase “bless your cash app” made it into this article. That made me very happy and entertained countless Shkreli fans.

Run-of-the-mill successful troll/common Shkreli W, right? No, it gets stranger. The Politico writer and his lackeys were showing off the dialogue on Twitter, presumably as some kind of ‘dunk’ on me. Of course, all of the comments explained to the hapless political observers that they’ve been had and they deserve what they get. Media has become a laughing stock and more subjects of stories will use it as I have.

Why is the media humorless and naive? The lack of self-awareness is a desperate attempt to cling to a time when media was respected and a source of truth. The events of the last decade have exposed the media as the occasional partner of the government’s propaganda and the tool of furthering reporter’s narratives and dreams of social influence.

My repeated question to the Politico writer was “why do I need you” to a) talk to Vivek, b) say anything about anything, etc. I sent two TikToks I gave an aggregate minute of thought to, and received 500,000 views. In the past, media was the gatekeeper of all mass access. You used to say “Hey mom, I’m on TV!” With the loss of that role, having passed it to social media, media itself is now clueless as what to do or what its role is. The bankruptcy of Vice, the end of Buzzfeed and Gawker, and what I assume will continue with Vox, STAT, and all other digital media, especially if we see an economic downturn, indicates the future. The rise of TikTok, Substack and Twitter’s renaissance demonstrates where media is heading.

Amazon unveiled “Tranium” and “Bedrock” among other LLM tools and reminded me of their launch on the earnings call. I will try them when they’re available and report back. When asked about ML/DL, Jassy had largely the same answer as Google and Meta: “we’ve been doing this all along!”. He did have the more honest addition that “these models have gotten a lot more interesting in the last six to nine months”.

You can tell who can iterate and introduce product quickly when new technologies arrive. The Amazons, Googles and Metas of the world are the dinosaurs of ten or twenty years ago: Oracle, SAP, IBM. Today’s OpenAI, HuggingFace and perhaps a few more players we’ve yet to meet (Character.ai?) will outfox our new troglodytes. A few of them will survive the sea change, but if history is any guide, this will be bloody.

Interesting excerpt from Jassy on the call:

“And if you look at the experience that's been the case for the last several decades, we're going to have a hard time convincing our grandkids that it used to be the case to get a primary care appointment, you had to call ahead of time, a month ahead to schedule an appointment and drive 20 minutes to the doctor and park and get into the facility and wait 20 minutes in reception. And you get into an exam room, you wait 10 minutes for the doctor to come in. The doctor talks to you for five minutes and then prescribes you medicine where you drive 20 minutes to go get the medicine. And that experience just doesn't make sense and won't be the case.”

I couldn’t agree more. But real disruption isn’t just copy-pasting the old healthcare supply chain together more efficiently. We need a real sea change to fix healthcare. Speaking of…

Dr. Gupta (www.drgupta.ai)

The feedback from Dr. Gupta has been overwhelmingly positive. Thank you! We’ve received two indications of interest for nationwide deployment. Most countries are struggling with healthcare costs. Some poorer countries have virtually no healthcare in rural areas. But most importantly, access to health information is a predictor of health outcomes. LLMs are a revolutionary way to query and traverse complex information that clearly outperforms search engines. Imagine being someone in the media who wants to marginalize or stop access to a tool which could help, say, rural Iranians healthcare. Thankfully only the blindest sheep still visit outfits like Gizmodo.

Our newest feature has launched: “Dr. Gupta’s Library” will now deposit medical journal articles related to your conversation with Dr. Gupta. Most people don’t even know the medical journal system exists. Try searching for ‘epilepsy’ or ‘diabetes’ on Pubmed. You will quickly leave as you realize you’re not sure how to navigate this mess. For example, there is no useful sorting by relevance that would actually yield something useful unless you’re an expert.

Lilly

Mounjaro, wow. $569m in the fourth quarter of launch. I made a partial list of blockbuster launches available at https://github.com/martinshkreli/models/blob/main/Biopharma.xlsx.

A couple of examples worth pointing out: Gardasil hit $365m in its fourth quarter. Vaccines tend to peak early as ACIP recommendations rule there. Januvia family of products hit $243m before peaking around $6B in annualized sales. Tremfya, Cymbalta, Darzalex and Zytiga all were in the $150-170m range in their first fourth quarter. Anything could happen to Mounjaro, but I don’t think anyone has considered that the product could possibly do $40B in revenue.

Miscellaneous

I have a financial product coming out soon. Not Godel, which will be out later in the summer.

Altman is right: the only way out of the national debt is AI/AGI. What do we spend the most on? Healthcare.

Vice bankruptcy is a sign of the times: social media and AI media will unspool the news media world. Tucker Carlson is an example of defecting to the former. The Hollywood strike will create more of the same creator economy usurping their would-be proprietors.

I bought the Gizmodo journalist’s domain name and will chronicle their stupidity. Journalists don’t like it when someone investigates them.

Microsoft had a very nice quarter in their “more personal computing” division. This is not a division most investors are focused on, given it is dominated by a declining Windows business. The focus for Microsoft investors is Cloud, which had okay results and guidance, but nothing the afterhours stock reaction would merit. I found a specific Microsoft answer to a question about OpenAI telling, and disappointing. Microsoft’s CFO said one of the ways to look at OpenAI is as an Azure customer, like everyone else. Hmm.

Google’s advertising business had another flat y/y quarter. It’s abundantly clear LLMs will be superior to search engines for many use cases. The alarming part of the Google call was the nonchalance management indicated. There was no acknowledgment of any problem, which Google’s deep discount to peers would strongly suggest the existence of. Instead, we heard, “we’ve been doing AI a long time”. I think it is time for Sundar to go. Google flubbed the product launch for Bard, which still has a ‘waitlist’. With 200,000 employees, Google was outfoxed by 100 OpenAIers. Nothing personal against Mr. Pichai (who has AI in his name!), but heads will roll for this underwhelming product development effort. Soon, the Board will decide it is the CEO that ultimately bears the responsibility for seemingly missing the biggest platform transition in computing history.

I have been streaming on YouTube: https://www.youtube.com/@realmartinshkreli

I’m banned from Twitter despite the promises of new management to remove permanent/lifetime bans. Please consider writing a message to the following accounts at Twitter @ to suggesting reversing the ban:

@elonmusk        @arodericks
@twittersafety  @tsotime
@ellagirwin         @titterdaily
@alx                      @stillgray
@jason

Today I’m unveiling DrGupta.ai. Dr. Gupta is a virtual healthcare assistant that can help you access health information and simulate interactions with a physician. Our aim is to make healthcare more accessible and convenient for everyone.

In the past, I faced criticism for raising the price of medicines, and I became a focal point for debates on pharmaceutical pricing. That surreal to me: pharmaceuticals only account for 12% of healthcare costs in the US, but they receive most of the blame.

Most healthcare costs are related to physicians, and all healthcare costs stem from physician decisions! This raises important questions: Why are these costs so high, and what can we do to address them? As we move forward, it’s crucial that we explore innovative solutions to make healthcare more affordable and accessible for all.

My central thesis is: Healthcare is more expensive than we’d like mostly because of the artificially constrained supply of healthcare professionals. Unfortunately, it is not feasible to instantly produce a million more physicians for practical reasons. A working labor market for healthcare workers is hindered by the efforts of lobbyists and bureaucrats who artificially maintain a limited supply. Although the importance of proper healthcare training cannot be overstated, the escalating financial burden of Medicare and Medicaid is pushing our nation towards bankruptcy.

Artificial intelligence holds the potential to significantly increase productivity in numerous industries, including the one I work in. My team and I rely on AI to program more efficiently, but it’s disheartening to see that healthcare has yet to fully benefit from technological advancements. Despite the logarithmic drop in the price of computation, storage and communications, physicians seem to be stretched thin, while healthcare costs continue to rise, leading to a reverse in productivity.

Although there have been significant technological advancements in recent years, it is surprising to see that the healthcare industry has remained impervious to innovation and efficiency. Even with these advancements, the number of FDA approvals for medicines has remained stagnant since the organization's establishment, and the number of doctors per capita has not seen any significant increase.

The root of the problem lies with people, as it is difficult, if not impossible, to change human limitations. For example, a doctor can only attend to one patient at a time, healthcare workers can become exhausted, and language and financial barriers exacerbate inequalities. Moreover, human errors, which can sometimes be fatal, are unavoidable.

While we should express our gratitude to every healthcare worker for their hard work, we should also question whether we can improve the current situation. Is there a way to overcome these limitations and make the healthcare industry more efficient?

Enter AI. ChatGPT-4 has demonstrated an impressive performance on the USMLE, leading us to believe that Dr. Gupta will soon provide advice on par with the world's top medical specialists. Discussions I’ve had with physicians remind me that it is very difficult for even “key opinion leaders” to stay abreast of developments in their field. Between distant medical conferences and endless publications, it is very difficult to stay on top of your specialty and still see patients. The average doctor I know barely has the time to read the New England Journal every week. Computers, of course, have no restrictions on processing voluminous data in real-time.

I envision a future where our children ask what physicians were like and why society ever needed them. Although Dr. Gupta cannot currently prescribe medication or perform surgical procedures, our team at DL Software (the parent company of Dr. Gupta, Druglike, and the forthcoming financial terminal product Godel) is committed to exploring this future. With our continued efforts, perhaps this dream will become a reality someday.

Dr. Gupta may not be ready to replace physicians entirely, but I believe it has the potential to be a valuable resource for many individuals. Previously, seeking medical information online proved to be a challenging endeavor. However, with advanced language models, individuals worldwide can now engage in authentic conversations with chat agents, regardless of time or language barriers. This offers those with limited healthcare knowledge, as well as physicians seeking a second opinion, the opportunity to access easy-to-understand healthcare information. Even experts like me have found this clarity difficult to come by in the past.

Dr. Gupta is on the path to improvement and will soon be even better. While we've rapidly launched this product, our roadmap includes the development of a truly distinctive application that enables users to learn about and manage their health in a comprehensive manner. As a first step, we will implement pushing personalized health information directly to you, providing relevant news as it occurs. Dr. Gupta is adept at interpreting medical papers, textbooks, or news articles and presenting them in an accessible and meaningful way tailored to your needs.

I am personally inspired by the potential to provide medical assistance to individuals in remote parts of the world where access to physicians is severely limited. I remember my parents recounting stories about rural Albania, where seeing a doctor was a near-impossible task, and the quality of care was often questionable. Reflecting on the countless individuals globally who are not as fortunate as those living in the United States, I am grateful that our product can offer them medical guidance for just a fraction of a cent. This is the driving force behind making initial use of Dr. Gupta free for anyone, anytime. However, for extensive usage, payment is required. Try it at www.drgupta.ai.

Over the last decade, I’ve learned and reflected a lot on healthcare. One crucial lesson I've learned is that people tend to approach healthcare with a more emotional rather than rational mindset. Health is undoubtedly the most critical aspect of human life, and any potential threat to our well-being, no matter how small, can be alarming. Even the perception of a threat can lead to misunderstandings and concerns.

As someone passionate about improving healthcare efficiency, I believe that the industry can greatly benefit from the advancements made in hardware and software in recent years. Whether it is through automation or artificial intelligence, I am convinced that a significant portion of healthcare information requests and decisions can be streamlined and improved. Given the enormous size of the global healthcare market, even a modest improvement in efficiency could have a substantial impact.

Technical matters

                Dr. Gupta uses a combination of GPT-3.5 and GPT-4. We have trained our own LLM which we are testing. Last year I stumbled upon the reflection and “auto” recursion-based methods many are employing under various names. For instance, I had ChatGPT write a 200 page book and have embarked on a number of other toy projects, most of which I believe are ahead of the disclosed SOTA. Dr. Gupta uses such tools to create, in my opinion, better performance than Chat GPT alone. Try and compare it yourself. This makes the speed of Gupta a tad bit slower, but not a lot more. GPT-4 text trickles out slowly while Gupta’s comes in one bolus, so there are some optical illusions there. Hopefully, GPT-4 speed will increase soon. We’d like to increase the amount of introspection Gupta does.

Investors & Cap Table Fever

                DL Software is funded by investors. In attempting to earnestly preview Gupta to the media, I rediscovered my hatred for journalists and their editors. There was a perplexingly deep interest in who our investors are by the standard fare: The Information and Business Insider. One intrepid reporter was able to track down both investors we talked to and didn’t in, hunting down every investor. He could probably put the cap table together better than our own accountants!

The identity of your investors is something many startups tend to dwell quite a bit on. But this is a classic mistake. Spoiler alert: Who your investors are does not matter and never has. Can you tell me who was the first investor in Microsoft? Did you know the average VC loses money? The only thing that matters to me is how much I can motivate my team to push itself as hard as humanly possible to create amazing technology that makes our customers more happy and productive. An investment from Warren Buffett (or insert whatever name you want here) himself would not ensure, or even improve, the odds of success of our company. It is the height of stupidity to focus on this. When I raised $90 million for Turing Pharmaceuticals, a huge amount in an era before today’s mega-rounds, I didn’t disclose any investors. There’s no embarrassment, unwillingness by them or string to pull on to unravel a conspiracy. It just doesn’t matter.

To other startups out there: Imagine disclosing every idiot that wrote a $25,000 check for your company. Are these names supposed to give people comfort? Most of you will fail. The invitation list to your funeral is not something to boast about. The few who will make it will not have made it because their venture fund steered them to success. If that were the case, the venture fund would return 5,000x with every portfolio company. Focus on your job and eliminate any other noise.

e/acc

I’ve become more active with the accelerationist movement on Twitter (twitter.com/marty_catboy). Accelerationism means different things to different people. To me, it means allowing technology to bring its benefits to the masses without questions and regulations. The progress of technology is impossible to stop once it has started and people will not stand for its interruption when it benefits their lives.

As artificial intelligence, “blockchain” and other technologies reach increasingly exponential heights, their benefits democratize power. This worries current power structures. We must make it clear we do not want the SEC’s “protection” from crypto any more than we want NTSB’s “protection” from self-driving cars.

Regulation of AI is a cripplingly stupid idea that will distribute American hegemony. Maybe that’s the goal. The irony of American (and Canadian) ingenuity being exploited by non-North Americans to raise their standard of living to, and above, ours, would be disappointing. Take a stand now, before someone tells you what software and hardware you are permitted to have.

AI, AGI

How do we get AI to plan? This seems like the subject of the month. Asking Chat GPT to articulate a plan as mundane as ‘clean a house’ or grand as ‘solve global warming’ doesn’t help too much. You get a reasonable plan that you can store in a variable. The problem becomes operationalizing that plan. It’s not hard to ask the computer to performStep(plan[step]). So, if the first step is ‘write a provocative essay on why global warming must be halted immediately’, it isn’t awfully hard to get that essay written. You can make the plan more complex by atomizing it as much as possible. So why don’t these tools work?

Planning is not useful without beliefs and crystallized learning. And probably a lot of other stuff I am clueless about. Repeated introspection probably helps in every step, but what are we judging on? It has been hard enough for me to guide AI through a very narrow task: write a cohesive fiction story. It has been awfully difficult to do so at scale, regardless of quality. LLMs are stochastic parrots with some emergent behavior, which while tantalizing, does not quite substitute for the real thing.

I think the Broca’s area equivalent of the LLM requires additional equivalent brain anatomy modules to get real utility. While I don’t think this will lead to an AI winter, I suspect the rapid desire to push out autoGPT and its ilk do not help. As many know, I am working on HumE, a 2D-RPG substrate to enhance the emergence of AGI within an abstract/finite but sufficiently complex world. There is some interesting emergence of unique behavior, but we’re not quite there. https://reverie.herokuapp.com/arXiv_Demo/ is a similar attempt, but is not a live server, so we don’t yet know how contrived this example is. I wonder if we should lower our goals even further and try to get a good mimic of a dog, cat or small child instead.

I think it’s worth reading this interview with John Carmack who is investigating AGI: https://dallasinnovates.com/exclusive-qa-john-carmacks-different-path-to-artificial-general-intelligence/. I agree that AI’s past may shed a lot of light on the path forward to round out the tools needed to approach AGI/ASI. I’ll have a lot more to say on this as my research progresses.

LQDA worked. You can sell it. Don’t say Shkreli never did nothin’ for ya. I’m sure the company will go on to great things, but you know how we do things around here.

NTLA is my next stock pick. It’s the only public ‘CRISPR company’ worth anything (so far as I know!). The point of technology is to apply it and execute. Intellia has done that as it looks like they have cured (I use that word very, very carefully) two liver-oriented diseases using CRISPR. Surprisingly, the trick was the same LNP that encapsulates the COVID vaccines we’ve all taken. I’ll put out some full due diligence later, but I think it is an “easy double”. Sticking my neck out is fun!

Software

I have so much to say on software, my new passion. I’ve been working on the state management problem with OpenAI’s API, which I assume will extend to any LLM that I train.

Next, I’m trying to envision what a post-React front-end/UI world looks like. We started a “React company” seven years ago when React had a few percentage points of the advanced front ends out there (mostly from Facebook products). Today, it dominates. Still, we have to move to where the puck will be. I’m not thinking frameworks (Svelte, etc.) but entire paradigm shifts. The hard thing about predicting the future is that it is very hard. The easier thing to do is watch very closely and jump on the bandwagons you like.

I’ve been spending most of my time on financial programming and Godel, my new financial terminal. This is still in beta, and you can request access here:

https://forms.gle/aTexJcdcuZY2eEZf8

Sam Bankman, Criminal Justice

I purchased www.USvSBF.com and will be making a mirror docket so you don’t have to get a PACER account. Maybe I’ll add some commentary since some seem to enjoy that. I think Sam is in big trouble. I wouldn’t underestimate the ability to impeach Ms. Ellison on cross examination, though. I don’t see him getting around his other colleagues. The issue is, surviving a cross-examination is quite hard, and I don’t know if these FTXers can handle the heat.

The reality is, virtually no one wins a criminal trial. This isn’t because no one is innocent. The sad truth is ‘innocent until proven guilty’ has been the wink-and-nod charade government and its functionary news media parrots. When there is a 99% conviction rate, “innocent until proven guilty” is perfunctory and guaranteed to be ephemeral. Why? The reality is the public still trusts law enforcement, especially federal law enforcement. Why? The actual trial system requires the government to “go first”: lay out their case as to why the soon-to-be-convicted (why even use ‘defendant’?) will soon be in prison. The jury anchors to this baseline. They hear it for however long the ‘case in chief’ takes. By the time to defense gets a shot at demonstrating evidence of innocence, the half-asleep jurors have made up their mind.

This is just one of dozens of implicit biases that occur in court. I’m still delighted I was acquitted of five of eight of my own criminal counts. This seemed odd to some. I was overjoyed, and quite frankly, it restored a tiny bit of my generally crushed faith in the criminal justice system that one can win at all, even if partially.

There is limited interest in reforming criminal justice because it does not affect many people. Even when it does, there is limited appetite to ‘root for the bad guy’. One day, though, you may be the ‘bad guy’. I pray you enjoy it as much as I did. Most don’t.

Personal

I’ve been watching The Three-Body Problem on Tencent. This was my favorite fiction book in prison.

Two posts in two days?! Yes.

First, I’m selling an AI-generated book that I… (uh… we…?) wrote. https://www.amazon.com/dp/B0BSWQYDD6 The book includes the full text of the manuscript, and a short introduction by myself. Finally, the code that generated the book is included in an appendix.

I wrote program that recursively generates text for the book based on a rolling average of state. As you can imagine, it produces a lot of text but it doesn’t hang together too well. You can find the code base and try it yourself at https://www.github.com/martinshkreli/bookgen

Version 2 is already a lot more promising. The weird part about that code is that it just tracks a larger state to help model and engineer prompts that keep GPT on track. I’m looking forward to sharing the next book. It also randomly selects genres, etc.

Gina P. Thomas

I created a tool called Gina. https://www.github.com/martinshkreli/ginajs Gina is a Discord plugin that creates an interface to ChatGPT. It’s connected to my GPT API account, so even if ChatGPT is “over capacity”, it still works. You can sample it at discord.gg/martinshkreli (which is a great place to visit anyway), or fork it to use yourself in your own Discord. It will be quite scary if my bot proliferates throughout Discord. My channel members are enjoying it quite a bit.

AI resources

Check out Andrej Karpathy’s YouTube: https://www.youtube.com/c/AndrejKarpathy

Karpathy was Tesla’s head of AI.

Read the paper that brought transformers to life: https://arxiv.org/pdf/1706.03762.pdf

That’s all I got for now!

Let’s see…

I have a new financial software package I’m hoping to unveil soon. It’s called Godel and you can request a beta invite by filling this form out: https://forms.gle/QW2NEeuPPzoz7Ri57

I hope Godel will provide a refactored AWS-style version of what Bloomberg is today. The Bloomberg terminal has a lot of issues I won’t enumerate, and every client knows what they are. Usage-based pricing (as well as a free tier), native web access and so many more advantages will propel Godel to success.

Druglike also has had some updates including a more rapid time-to-start for screening. Try it out at www.druglike.com. I’m hoping to add molecular dynamics, pharmacokinetics and other simulators to the total package at some point. Other upgrades would include fiat payment, API, and of course, a decentralized compute strategy. One idea is to use webworkers/wasm to bootstrap your own simulation jobs, where you can add your friends’ local compute to your ‘workspace’. A private folding@home.

As I’ve said many times, Druglike is not a drug company in any way, shape or form, and the FTC has not suggested it is. Druglike is software and never touches an actual tangible thing, like a chemical or a mouse or god forbid, a prescription medicine.

So what was that news about? My lawyers and I missed a deadline to file a routine update with the FTC and schedule a meeting. That will be done. I’m not in contempt and won’t be ‘in trouble’ over this. Mea culpa. It turns out some people on my team were sick during the response period and there was some telephone tag. Total non-event, but don’t tell that to the media who insists this is somehow ‘news’.

Anyway, I’ve worked on some other software for fun. On my github, you will find a ChatGPT-powered recursive story/book generator: https://github.com/martinshkreli/bookgen. “I” wrote a 47-page novella that I’m going to toss on Kindle. That should be funny.

I made a terminal-based quote monitor which is pretty funny. Real-time quotes in your command prompt. This is not the Godel master project, just something for fun. https://github.com/martinshkreli/stocks

I’m also on wakatime if you want to follow me there: https://wakatime.com/@martinshkreli

Outside of software, I’ve been paying attention to the rest of the world. Some stocks look cheap, others don’t. Not much has changed there. I’ve liked and at recent prices. I talk a lot about this stuff on my livestreams on YouTube (https://www.youtube.com/channel/UCjYKsjt-7EDU78KEcVbhYnQ) or Twitch (https://www.twitch.tv/martinshkreli2022). I looked at and other stocks recently. You can find updated models on Github, including a small hedge fund tracker: https://www.github.com/martinshkreli/models

I am to please, so let me know what you’d like to see more content about, including analysis of specific stocks. I may be doing a fun podcast with a few friends. This would be mostly comedy-oriented and not focused on things like this recent paper from a Chinese group on quantum computing: https://arxiv.org/pdf/2212.12372.pdf. It was cool to try AWS Braket. It’d be nice to try to do what the Chinese authors did on a Rigetti. When I have time…

I have an old hard drive of my finance and chemistry lessons, so expect those to be uploaded again. I’m thinking about the right way to continue that series.

Finally, I’ve been enjoying the video game “Rust”. It’s a very immersive survival game that speaks to my OCD/hoarder nature. I need to quit!

Until next time,

Martin Shkreli

But the signed message I posted yesterday is cryptographically authentic. Many have tested it, and you can try it yourself: https://www.bitcoin.com/tools/verify-message/

The only problem is it doesn’t appear authentic. My guess is the consensus response is correct: someone has Hal Finney’s keys and signed this somewhat bizarre message. There are two questions, one more important than the other, but neither particularly breathtaking: 1) Who has Hal Finney’s keys? 2) Why are they doing this?

The important one is the first question. I reached out to Finney’s family to see if there was some chain-of-custody we can follow over Hal’s old key pairs. In the 1% chance they are adamant no one could have these keys other than the family, and the family is not signing weird messages, that would be quite interesting. In the 99% chance the keys have been sold, leaked, hacked, etc., then the message is what it appears to be: authentic from a cryptological perspective, but still some lunatic attempting to lead everyone down a wild goose chase.

I will report back with what I find, but I certainly thought the message was worth passing. It is noteworthy that the private key which received the very first bitcoin did successfully sign this message. Does that make Paul Le Roux, a long-suspected possibility, Satoshi? No, not quite.

1Q2TWHE3GMdB6BZKafqwxXtWAWgFt5Jvm3

The following signature

HM7vpPSUbNsfDHRX6gv8xxWcVNHEc/3pOk0YrVehaGoUdbWizznfzOdELkLd1EjSXsW1oE5vHAkNAPzrAVzhuoI=

decrypts to:

-----BEGIN BITCOIN SIGNED MESSAGE-----

This Transaction was made by Paul Leroux to Hal Finney on January 12, 2009 #bitcoin

-----END BITCOIN SIGNED MESSAGE-----

The FTX saga is irresistible pornography for market participants. One cannot look away from this sordid epic: Shakespearean? No, Homeric, at least.

There are so many victims small and large, we will be hearing and talking about FTX for a decade or more. But beyond the voyeurism and sympathy, there are lessons and opportunities for the shrewd.

If you haven’t been paying attention, it’s probably because you think ‘crypto is a scam’ with an unfathomable degree of apathy (you do know you can short crypto, right?). But the FTX saga raises bigger questions than just its effects on the crypto ecosystem. Venture capital, regulators’ unintended consequences, the nature of altruism, human nature and “FOMO” in investing and corporate controls are just a small list of topics we can start with.

A lot has been and will be written. I am not here to compete with the MSM. What are my non-consensus observations or insights?

• There is no conspiracy about Democrats or the FTX leadership’s family relationships. We can all do better than contrive these. The actual story is far more interesting than your imagination’s! A simple point: if you’ve ever played politics, leading donations in one election cycle is meaningless for ‘buying influence’. It takes a lot more consistency before politicians will be in your pocket, if ever. Politicians are much smarter than you think, and while the game can be pay-to-play, it isn’t as simple as coin-operated. Local political control by donors is a real thing and the “new kid on the block” can’t show up and take over, no matter how much money they have.

• Sam will likely get life in prison, especially if he was diverting customer funds since inception, which appears to be the case. The law is not blind to the wealthy and notable. If anything, it is more punitive. There are caveats: which district a case will be brought in, which judge will be assigned. There are districts and judges within them that would bury SBF under the prison, and others where I could see a 15 year sentence. But I can’t conjure a sentence more lenient than that.

• Caroline Ellison likely played no material role in Alameda, and was simply a stand-in figure to maximize separation. A warm body. Any ‘leader’ at Alameda in the past was just taking SBF directives for large pay and plausible deniability.

• I believe Alameda and FTX lost a lot of money (>$3B) being levered long crypto, possibly via BTC/ETH futures exposure. The $3T → $1T evaporation in crypto value is the cause of their losses. The venture capital investments the FTX/Alameda complex made were not large enough to cause the massive hole here.

• I’ll take a quick victory lap for being, what I saw, the first to say short FTT on the “bailout” move upwards, the first to point to a $10-20B hole in FTX when people were saying $1-2B, and the first to advise those interested to head to the Bahamas for bankruptcy proceedings. We’ll see if my other prognostications come true.9

• A large bank likely lent SBF >$500m on his personal stake in FTX. That bank will take a writeoff. Not too material, but interesting.

• Financial controls are an element of due diligence in hedge funds: why not venture-backed startups? Or at least financial startups?. Simple questions: can one person withdraw all of the money? In most companies, a CFO and controller are required for wire transfers above a certain amount. The same thing for ledger entries. Board approval is required for transactions above a certain amount. This is not the first time something like this has happened, and there are a litany of methods professionals use to avoid these types of disasters.

Twitter v2

The best coverage on the FTX saga was not brought to you by WSJ or FT. Twitter citizen journalists delivered, in a way no ligma-believing formerly exclusive blue-checkmark possessors could have. I fear the day of the gumshoe journalist is over. I read a fascinating piece in Vanity Fair about a con man which demonstrated the power of a persistent (and funded) investigator to fill the gap between law enforcement and clickbait which media is supposed to do.

Very few outlets are committed to this kind of work anymore. In their place, folks like “Autism Capital” and marathon Twitter Spaces by Mario Nawfal have replaced the hole for-profit sensationalists have left us. I couldn’t be happier. Actual practitioners should be reporting the news. These are people who have subject matter expertise and won’t be fooled as easily fooled by ‘the next Warren Buffett’. It’s even more exciting that the news reporters of November 2022 are sometimes anonymous, they’re not building a brand for themselves, and they’re sacrificing for the collective good, not their career or company bottom line.

Pharma

Q3 results show launches are going well and pharma revenue in general is holding up quite well relative to cyclical sectors. Great Q3s from Big Pharma: Tremfya, Darzalex, Erleada, Xarelto (J&J), Mounjaro, Verzenio, Jardiance (LLY), Skyrizi, Rinvoq, Vraylar (ABBV), Evrysdi (Roche), Keytruda, Gardasil (MRK), Prevnar, Paxlovid (PFE), Rybelsus, Ozempic, Saxenda (NVO), Imfinzi, Evushield, Ultomiris, Calquence (AZN), Kisqali, Kesimpta, Pluvicto (NVS), Revlimid (soft landing?), Abecma, Opdualag (BMY), Evenity, Tezpire (AMGN), Dupixent, Nexviazyme (SNY), Biktarvy, Veklury (!?), Yescarta, Trodelvy, Epclusa (!?) (GILD), Shingrix, Dovato, Xeduvy, Benlysta, Bexsero, Cabenuva (GSK). There are a few disappointing or stalled launches, but things are looking pretty sunny for drugs. Models on Github are relatively updated.

Moderna revenue is drying up and the valuation has me wondering if this is a relatively obvious short.

Mounjaro revenue may set records as more average Joes are reaching for GLP-1s for non-obesity weight loss. This recreational use of medicine is a bit worrisome, but it won’t bother Eli Lilly!

My Professional Efforts

Sorry for not writing more! I am focused on the work I’m doing on software companies. Druglike is expanding its purview quite a bit and is being renamed. An old software suite I worked on is being added into it, dramatically increasing its scope. Some of you know what I’m talking about, others will have to wait! A private demo is up and running on the web and a closed beta will be available for that at some point soon.

And, yes, all of this software will implement the ERC-20 token MSI. I think a token economy adds a small benefit to existing software, but the focus should be on the software, not the token.

I’m also working on other cheminformatics and bioinformatics software. The free energy calculator my friend Jason made (druglike.com) has advanced steadily, with a 1 million compound screen performed on PARP1 inhibitors. There was no way to do something like that on the web before. But I’m trying to do even more new things, like a possible brand new approach to calculate logP and more easy-to-use front ends for cheminformatics.

I am also interested in online gambling and video games in general. Especially web based games that incorporate WASM, WebGL, React, Unity and new 2D/3D frameworks, with or without Canvas. I think crypto has a lot to add to the gambling space and very little progress has been made.

If you have software development expertise in the above and want to join me, please email me at martin@martinshkreli.com. If you want to learn more about what I’m doing in this space from a corporate perspective, or you just want to check in, feel free to send an email as well.