What We’re Reading (Week Ending 04 July 2021) - 04 Jul 2021
Reading helps us learn about the world and it is a really important aspect of investing. The legendary Charlie Munger even goes so far as to say that “I don’t think you can get to be a really good investor over a broad range without doing a massive amount of reading.” We (the co-founders of Compounder Fund) read widely across a range of topics, including investing, business, technology, and the world in general. We want to regularly share the best articles we’ve come across recently. Here they are (for the week ending 04 July 2021):
1. RNAi is Setting a High Bar for Gene Editing – Maxx Chatsko
Although there are many therapeutic modalities in the field of genetic medicines, individual investors have been most excited about gene editing tools such as CRISPR-Cas9. The valuations of publicly-traded CRISPR companies suggest investors have very high expectations — perhaps unreasonably high considering the general lack of meaningful data.
In the second quarter of 2021, three different drug candidates or drug products based on RNA interference (RNAi) have demonstrated the ability to reverse three different diseases — all with convenient dosing…
…Intellia Therapeutics (NASDAQ: NTLA) is initially focused on developing in vivo drug candidates for gene silencing applications. The approach uses CRISPR-Cas9 to “knock out” a gene by disrupting the sequence responsible for its expression. At a high level, the clinical objective of reducing protein levels is identical to that of RNAi.
Not surprisingly, there’s plenty of overlap between the company’s knockout pipeline and those of RNAi drug developers. Intellia Therapeutics’ lead in vivo drug candidate is taking aim at hATTR, while the next most-advanced program is targeting hereditary angioedema (HAE), another Alnylam target. Discovery-stage programs that might use knockouts include PH1, A1AT liver disease, and others that promise to square off with RNAi drug candidates and drug products.
On the one hand, gene knockouts promise to drive deeper reductions in protein levels than current-generation RNAi. They would also represent a permanent, irreversible change to a patient’s genome. Although CRISPR gene editing compounds must be administered intravenously, a single dose is the most convenient dosing.
On the other hand, there are clinical and commercial challenges for investors to consider. First-generation CRISPR gene editing requires making a double-stranded break in the genome, which is repaired with mutagenic (“mutation-causing”) processes. Double-stranded breaks can result in random insertions and deletions of genetic material far from the cut site, while there’s evidence that sections of chromosomes can be rearranged. Each is a hallmark of cancer cells. (It’s important to note that these are on-target effects inherent to CRISPR gene editing, separate from the more familiar off-target effects discussed in the media, which are largely exaggerated.)
The long-tail safety risks from on-target effects might not become evident until years after clinical development is completed. Therefore, drug candidates utilizing CRISPR-mediated knockouts might generate promising safety and efficacy data in clinical trials, but regulators might balk at speedy approvals for knockouts or require long-term patient monitoring. This is especially true considering many indications being targeted by knockouts will likely have safe, effective, and convenient treatments provided by RNAi. In other words, although these indications are called “rare diseases,” regulators probably won’t be under pressure to approve knockout drug candidates for the sake of patients.
Even if knockouts earn regulatory approval, it could be difficult to dislodge RNAi treatments. For example, by the time Intellia’s lead drug candidate reaches the market (assuming it does), most global hATTR patients will be taking treatments from Alnylam. An estimated 3% of global patients are taking the relatively inconvenient Onpattro, although many more could be eligible for treatment with vutrisiran should it earn regulatory approval in 2021 or 2022. That could result in Intellia boasting an approved knockout drug product and frustratingly little opportunity to capture market share. It’s more likely that the market experiences cutthroat pricing competition between RNAi treatments and gene knockouts, which would be great for patients, but perhaps not so great for the drug developers arriving a little too late to the market.
2. The Devastating Decline of a Brilliant Young Coder – Sandra Upson
In Cloudflare’s early years, Lee Holloway had been the resident genius, the guy who could focus for hours, code pouring from his fingertips while death metal blasted in his headphones. He was the master architect whose vision had guided what began as a literal sketch on a napkin into a tech giant with some 1,200 employees and 83,000 paying customers. He laid the groundwork for a system that now handles more than 10 percent of all internet requests and blocks billions of cyberthreats per day. Much of the architecture he dreamed up is still in place.
But some years before the IPO, his behavior began to change. He lost interest in his projects and coworkers. He stopped paying attention in meetings. His colleagues noticed he was growing increasingly rigid and belligerent, resisting others’ ideas, and ignoring their feedback.
Lee’s rudeness perplexed his old friends. He had built his life around Cloudflare, once vowing to not cut his hair until the startup’s web traffic surpassed that of Yahoo. (It took a few short months, or about 4 inches of hair.) He had always been easygoing, happy to mentor his colleagues or hang out over lunch. At a birthday party for Zatlyn, he enchanted some children, regaling them with stories about the joys of coding. The idea of Lee picking fights simply didn’t compute.
He was becoming erratic in other ways too. Some of his colleagues were surprised when Lee separated from his first wife and soon after paired up with a coworker. They figured his enormous success and wealth must have gone to his head. “All of us were just thinking he made a bunch of money, married his new girl,” Prince says. “He kind of reassessed his life and had just become a jerk.”…
…WHAT MAKES YOU you? The question cuts to the core of who we are, the things that make us special in this universe. The converse of the question raises another kind of philosophical dilemma: If a person isn’t himself, who is he?
Countless philosophers have taken a swing at this elusive piñata. In the 17th century, John Locke pinned selfhood on memory, using recollections as the thread connecting a person’s past with their present. That holds some intuitive appeal: Memory, after all, is how most of us register our continued existence. But memory is unreliable. Writing in the 1970s, renowned philosopher Derek Parfit recast Locke’s idea to argue that personhood emerges from a more complex view of psychological connectedness across time. He suggested that a host of mental phenomena—memories, intentions, beliefs, and so on—forge chains that bind us to our past selves. A person today has many of the same psychological states as that person a day ago. Yesterday’s human enjoys similar overlap with an individual of two days prior. Each memory or belief is a chain that stretches back through time, holding a person together in the face of inevitable flux.
The gist, then, is that someone is “himself” because countless mental artifacts stay firm from one day to the next, anchoring that person’s character over time. It’s a less crisp definition than the old idea of a soul, offering no firm threshold where selfhood breaks down. It doesn’t pinpoint, for example, how many psychological chains you can lose before you stop being yourself. Neuroscience also offers only a partial answer to the question of what makes you you.
Neural networks encode our mental artifacts, which together form the foundation of behavior. A stimulus enters the brain, and electrochemical signals swoosh through your neurons, culminating in an action: Hug a friend. Sit and brood. Tilt your head up at the sun and smile. Losing some brain cells here or there is no big deal; the networks are resilient enough to keep a person’s behaviors and sense of self consistent.
But not always. Mess with the biological Jell-O in just the right ways and the structure of the self reveals its fragility.
Lee’s personality had been consistent for decades—until it wasn’t…
…In mid-March of 2017, Kristin and Lee went to a neurologist to get the results of an MRI. To Kristin, it seemed that the neurologist had initially been skeptical of her concerns. Lee was young, healthy, and communicative.
The MRI told a different story: There was atrophy in the brain inconsistent with the age of the patient, the neurologist reported to them. When Kristin asked her what that meant, she said Lee had a neurodegenerative disease of some kind, but they’d need to do more tests to get a specific diagnosis. One of their doctors suggested they go to the Memory and Aging Center at UC San Francisco…
…The neurologists delivered their verdict: He appeared to have a textbook case of frontotemporal dementia—known by the shorthand FTD—specifically, the behavioral variant of that disease. It targets a network of brain regions sometimes described as underpinning one’s sense of self. As the pathological process advanced, it was carving a different person out of Lee’s raw substance.
The term frontotemporal dementia refers to a cluster of neurodegenerative diseases that affect a person’s behavior or speech while leaving memory largely intact, at least early on. Unlike Alzheimer’s disease, FTD isn’t well known. It is a rare disease, affecting roughly one in 5,000 people, though many of the neurologists who study it believe it is underdiagnosed. What is known is that for people under the age of 60, it is the most common form of dementia. Still, as a man in his thirties, Lee was unusually young to be afflicted. For some patients, one of several genetic mutations turns out to be the likely cause, and a subset of patients have a family history of neurodegenerative diseases. But nothing in the neurologists’ investigations turned up even a hint as to why Lee had been struck down.
Regardless of cause, the prognosis is grim. There’s no treatment. Lee’s doctors warned that his symptoms would grow worse, and that over time he would likely stop talking, become immobile, and struggle to swallow, until eventually an infection or injury would likely turn fatal. The best the doctors could recommend was eating a balanced diet and getting exercise.
The family sat stunned at the neurologist’s words. The brain scans were undeniable. On a wall-mounted screen the doctors showed a cross-section of the four lobes of Lee’s brain. In a healthy brain, the familiar, loopy folds of tissue appear white or gray and push up against the edges of the cranium, filling every available space. Lee’s brain looked nothing like that.
Black voids pocked his frontal lobe, areas where brain tissue had gone dead. Seeing it, Kristin gasped. “There were huge dark spots in his brain,” Alaric says. “That’s what … that made it concrete.”
3. Interview: Marc Andreessen, VC and tech pioneer – Noah Smith and Marc Andreessen
N.S.: One of the themes of my blog so far has been techno-optimism. I have to say that some of that attitude comes from talking to you over the years! Are you still optimistic about the near future of tech? And if so, which tech should we be most excited about?
M.A.: I am very optimistic about the future of tech, at least in the domains where software-driven innovation is allowed. It’s been a decade since I wrote my essay Software Eats The World, and the case I made in that essay is even more true today. Software continues to eat the world, and will for decades to come, and that’s a wonderful thing. Let me explain why.
First, a common criticism of software is that it’s not something that takes physical form in the real world. For example, software is not a house, or a school, or a hospital. This is of course true on the surface, but it misses a key point.
Software is a lever on the real world.
Someone writes code, and all of a sudden riders and drivers coordinate a completely new kind of real-world transportation system, and we call it Lyft. Someone writes code, and all of a sudden homeowners and guests coordinate a completely new kind of real-world real estate system, and we call it AirBNB. Someone writes code, etc., and we have cars that drive themselves, and planes that fly themselves, and wristwatches that tell us if we’re healthy or ill.
Software is our modern alchemy. Isaac Newton spent much of his life trying and failing to transmute a base element — lead — into a valuable material — gold. Software is alchemy that turns bytes into actions by and on atoms. It’s the closest thing we have to magic.
So instead of feeling like we are failing if we’re not building in atoms, we should lean as hard into software as we possibly can. Everywhere software touches the real world, the real world gets better, and less expensive, and more efficient, and more adaptable, and better for people. And this is especially true for the real world domains that have been least touched by software until now — such as housing, education, and health care…
…N.S.: Your most famous quote is probably “Software is eating the world”. How is that likely to manifest over the next decade or so? Will A.I. automate whole business models out of existence? Will old-line companies that try to patch software into their existing operations and business models get outcompeted by companies that start out as software companies and then branch into traditional markets, as my friend Roy Bahat believes? Or something else?
M.A.: My “software eats the world” thesis plays out in business in three stages:
1. A product is transformed from non-software to (entirely or mainly) software. Music compact discs become MP3’s and then streams. An alarm clock goes from a physical device on your bedside table to an app on your phone. A car goes from bent metal and glass, to software wrapped in bent metal and glass.
2. The producers of these products are transformed from manufacturing or media or financial services companies to (entirely or mainly) software companies. Their core capability becomes creating and running software. This is, of course, a very different discipline and culture from what they used to do.
3. As software redefines the product, and assuming a competitive market not protected by a monopoly position or regulatory capture, the nature of competition in the industry changes until the best software wins, which means the best software company wins. The best software company may be an incumbent or a startup, whoever makes the best software.
My partner Alex Rampell says that competition between an incumbent and a software-driven startup is “a race, where the startup is trying to get distribution before the incumbent gets innovation”. The incumbent starts with a giant advantage, which is the existing customer base, the existing brand. But the software startup also starts with a giant advantage, which is a culture built to create software from the start, with no need to adapt an older culture designed to bend metal, shuffle paper, or answer phones.
As time passes, I am increasingly skeptical that most incumbents can adapt. The culture shift is just too hard. Great software people tend to not want to work at an incumbent where the culture is not optimized to them, where they are not in charge. It is proving easier in many cases to just start a new company than try to retrofit an incumbent. I used to think time would ameliorate this, as the world adapts to software, but the pattern seems to be intensifying. A good test for how seriously an incumbent is taking software is the percent of the top 100 executives and managers with computer science degrees. For a typical tech startup, the answer might be 50-70%. For a typical incumbent, the answer may be more like 5-7%. This is a huge gap in software knowledge and skill, and you see it play out every day across many industries.
As for Artificial Intelligence, as an engineer myself, it’s hard to be quite as romantic as a lot of observers tend to be. AI — or, to use the more prosaic term, Machine Learning — is an incredibly powerful technology, and the last decade has seen explosive AI/ML innovation that’s increasingly showing up in the real world. But it’s still just software, math, numbers; the machines aren’t becoming self aware, Skynet is not here, computers still do exactly what we tell them. So AI/ML continues to be a tool used by people, more than a replacement for people.
A famous story from the birth of computer science has Alan Turing, father of the computer, lunching with Claude Shannon, father of information theory, in the AT&T executive dining room near Bell Labs in the early 1940s. Turing and Shannon engage in an increasingly heated discussion about the future of thinking machines when Turing stands up, pushes his chair back, and says loudly, “No, I’m not interested in developing a powerful brain! All I’m after is a mediocre brain, something like the President of AT&T.'”
I think about AI like that — although, for the record, the President of AT&T is a friend of mine, and he’s actually quite bright. I suspect “Artificial Intelligence” is the wrong framing for the technology; Doug Engelbart was probably more correct with what he called “Augmentation”, so think “Augmented Intelligence”. Augmented Intelligence makes machines better thought partners for people. This concept is clearer for considering both the technological and economic consequences. What we should see in a world of rapidly proliferating Augmented Intelligence is the opposite of a jobless dystopia — productivity growth, economic growth, new job growth, and wage growth.
And I think this is exactly what we are seeing. It’s worth remembering that before COVID, only 18 months ago, we were experiencing the best economy in 70 years — rising wages, low and falling unemployment, and essentially zero inflation. The economy was even improving more for lower skill and lower income people than it was for people like us, despite computers everywhere. Unemployment among the most disadvantaged in our society — people without even high school degrees — was as low as it’s ever been. This is far from an automation-driven dystopia; in fact, it’s the payoff from three centuries of increasing mechanization and computerization. As the economy recovers from COVID, I expect these positive trends to continue.
4. Individuals or Teams: Who’s the Better Customer for SaaS Products? – David Sacks
There are three main reasons for the superior economics of Team products:
1. Deal Sizes
Team products have larger initial contract values as a result of the ability to sell multiple seats. By contrast, the small deal sizes of Individual products may be insufficient to justify the cost of a sales team. Unless the Individual product is highly viral, it will be easier to build a distribution strategy for a Team product.
2. Retention
Team products are stickier than Individual products. To use a gaming analogy, multiplayer mode is more engaging than single-player mode. Users can do more interesting things when coworkers are part of the experience; value creation is higher.
Once a team is collaborating in a product, no single user can easily make the decision to leave. The decision to migrate to another tool requires coordination (aka a “rip and replace”). By contrast, a solo user can leave an individual product at any time.
Finally, collaboration provides constant opportunities for reactivation. A subscriber who stops using an Individual product is likely churned whereas an inactive user on a Team product is just one notification away from being reengaged. As long as the team maintains some minimum threshold of engaged users, it will avoid churn at the account level.
For all of these reasons, account-level churn rates for Individual plans are commonly around 5% per month, but only 1-2% per month for Team plans. This translates into much higher revenue retention for Team plans.
3. Seat Expansion
Team plans have the ability to add new seats as the product spreads within a company, creating revenue expansion. As a result, successful Team products have “net negative churn,” meaning that expansion from retained accounts exceeds revenue lost from churned accounts.
5. It’s Official: US Government Says Electric Vehicles Cost 40% Less To Maintain – Steve Hanley
In its latest study, the Office of Energy Efficiency and Renewable Energy says,
“The estimated scheduled maintenance cost for a light-duty battery-electric vehicle (BEV) totals 6.1 cents per mile, while a conventional internal combustion engine vehicle (ICEV) totals 10.1 cents per mile. A BEV lacks an ICEV’s engine oil, timing belt, oxygen sensor, spark plugs and more, and the maintenance costs associated with them.”…
…Big deal, you say? Who cares about a difference of a measly 4 cents? Consider this. The light duty vehicles — sedans, SUVs, passenger vans and the like — owned and operated by the federal government traveled nearly 2 billion miles in 2019, according to the General Services Administration. That difference of 4 little cents translates into savings of about $78 million a year, according to Motor Trend.
The one thing that the EERE study doesn’t show is the reduction in fuel costs for those government owned vehicles, which allows us to do a little speculating. Let’s assume the average fuel economy for all of them is 20 miles per gallon. That means it would take about 100 million gallons of gasoline to drive 2 billion miles. Now lets assume that gas costs an average of $3.00 a gallon (I am math challenged so I like to use round numbers). 100 million gallons at 3 bucks a gallon equals $300 million, does it not?
Now let’s assume further that the cost of electricity is roughly half the cost of gasoline. The end result is that a fleet of electric vehicles would save Uncle Sam about $150 million in fuel costs every year. Add in the $78 million in lower maintenance costs and the total annual savings from switching the entire US government fleet to electric vehicles could be $228 million every year from here to eternity or $2.28 billion over the next decade.
6. Casualties of Perfection – Morgan Housel
So many people strive for efficient lives, where no hour is wasted. But an overlooked skill that doesn’t get enough attention is the idea that wasting time can be a great thing.
Psychologist Amos Tversky once said “the secret to doing good research is always to be a little underemployed. You waste years by not being able to waste hours.”
A successful person purposely leaving gaps of free time on their schedule to do nothing in particular can feel inefficient. And it is, so not many people do it.
But Tversky’s point is that if your job is to be creative and think through a tough problem, then time spent wandering around a park or aimlessly lounging on a couch might be your most valuable hours. A little inefficiency is wonderful.
The New York Times once wrote of former Secretary of State George Shultz:
His hour of solitude was the only way he could find time to think about the strategic aspects of his job. Otherwise, he would be constantly pulled into moment-to-moment tactical issues, never able to focus on larger questions of the national interest.
Albert Einstein put it this way:
I take time to go for long walks on the beach so that I can listen to what is going on inside my head. If my work isn’t going well, I lie down in the middle of a workday and gaze at the ceiling while I listen and visualize what goes on in my imagination.
Mozart felt the same way:
When I am traveling in a carriage or walking after a good meal or during the night when I cannot sleep–it is on such occasions that my ideas flow best and most abundantly.
Someone once asked Charlie Munger what Warren Buffett’s secret was. “I would say half of all the time he spends is sitting on his ass and reading. He has a lot of time to think.”
This is the opposite of “hustle porn,” where people want to look busy at all times because they think it’s noble.
7. Tobi Lütke – Sriram Krishnan and Tobi Lütke
This is a very natural segue to my next question. One of the theories behind this whole set of interviews is diving into the atomic bits of how we spend our time in meetings. This time compounds over the long term and has a massive effect. What does a good meeting with Tobi look like? Alternatively, what does a bad meeting with you look like?
So actually, agendas are not terribly successful with me. I admire how other CEOs I’ve spoken with always have a strict agenda where everyone has a speaking slot. I find that absolutely fascinating. Even if I really set myself to an agenda and say, “Okay, great, this is going to happen,” I can’t get through half of a meeting like this. Partly because a good meeting is, for me personally, when I learn something.
I started a company because I love learning. I went into programming because I found it fascinating. During meetings, I just love to hear the things that teams have discovered. When you’re discussing an idea or a decision, I want to know what has been considered. To be honest, I find myself more interested in the inputs of an idea than the actual decision. I say this because when I have my own ideas, the first thing I tend to do is just try to falsify them, to figure out why what I’m thinking about is probably incorrect. This is actually something that I have to explain to people that I work with. If I like someone’s idea, I tend to do the same thing: I try to poke holes in it.
I usually say, “Well, the implication of this choice means you’ve made the following assumptions. What inputs did you use to make these foundational assumptions?” Effectively, I’m trying to figure out if an idea is built on solid fundamentals. I find that shaky fundamentals tend to be where things often go wrong. The decision being discussed could be the perfect decision according to the various assumptions that everyone came into the room with. But if those assumptions are faulty, the seemingly perfect decision is faulty too. Interestingly, assumptions are rarely mentioned in the briefing docs or in the slide deck. Usually, I’m trying to make sure those are rock solid. Through this process, I invariably end up learning something completely new about a field. That gives me great confidence and comfort both in the decision and the direction…
…You try and design how your company spends time and attention. One particular incident came up recently which I found really fascinating. You wrote a script to delete every recurring meeting at Shopify. Talk about why you did that, and what you ended up learning from it.
[Laughs] So, going back a little bit further there—you know what, I should talk about books. One thing that is interesting is how people have accused Shopify of being a book club thinly veiled as a public company.
We tend to read a lot and talk about a lot of books. We read Nassim Taleb’s books and one person on my team began talking about Antifragile and gave an outline. He said, “I think Nassim is putting a word to the thing that you keep talking about…”
Now, I come from an engineering perspective. One of my biggest beefs with engineers, in general, is that they love determinism. I think there’s very little determinism in engineering left that’s of value. An individual computer is deterministic; once you introduce even just a network connection into the mix, everything becomes unpredictable and you have to write code that’s resilient to the unknown. Most interesting things come from non-deterministic behaviors. People have a love for the predictable, but there is value in being able to build systems that can absorb whatever is being thrown at them and still have good outcomes.
So, I love Antifragile, and I make everyone read it. It finally put a name to an important concept that we practiced. Before this, I would just log in and shut down various servers to teach the team what’s now called chaos engineering.
But we’ve done this for a long, long time. We’ve designed Shopify very well because resilience and uptime are so important for building trust. These lessons were there in the building of our architecture. And then I had to take over as CEO.
When that happened, I made two decisions: one, I’m going to try to learn as much about business as possible. But, if business is very different from software architecture, I’m going to be no good no matter what I do. And so, I ran an experiment to treat engineering principles, software architecture, complex system design, and company building as the same thing. Effectively, we looked for the business equivalent of just turning off servers to see if the system has resiliency. For instance, we used to ask people to use their mouse on their non-dominant hand for a day. We introduced these little nudges to ensure that people didn’t become complacent.
There are a bunch of really fun stories around this. I had a conversation with one of my co-founders, and we were discussing our unique problem: namely, Shopify was a company initially for American customers, built by German founders, in Canada.
[Both laugh]
It’s a very complex thing.
For instance, we talk a lot about how different cultures interact because we couldn’t have built Shopify without the Canadian optimism. These things were not necessarily things that we would recognize, at least when it comes to optimism, coming from Germany. That said, there are also challenges between cultures. For instance, Canadians are unbelievably nice. Like, no one wants to ever say anything to upset anyone. This is why we need to emphasize the importance of feedback. In this way, the uphill battle is more real for us than it would be for a Silicon Valley company.
There’s so much on the theme of how Shopify is not a Silicon Valley company. I think you pointed out one of these themes right here.
Exactly.
For instance, if we had built the company in Israel, this would not have been a challenge. It’s really important to understand that culture is real and multi-layered. The “host” city’s effect on the employees in that local office is very real. To do something world class, you have to show up with a lot of world class skills, and not a lot of downsides.
In this way, pushing people to give feedback is something very important for us.
That was a tangent, but to get back to the question you asked, we found that standing meetings were a real issue. They were extremely easy to create, and no one wanted to cancel them because someone was responsible for its creation. The person requesting to cancel would rather stick it out than have a very tough conversation saying, “Hey, this thing that you started is no longer valuable.” It’s just really difficult. So, we ran some analysis and we found out that half of all standing meetings were viewed as not valuable. It was an enormous amount of time being wasted. So we asked, “Why don’t we just delete all meetings?” And so we did. It was pretty rough, but we now operate on a schedule.
Disclaimer: None of the information or analysis presented is intended to form the basis for any offer or recommendation. Of all the companies mentioned, we currently have a vested interest in Shopify. Holdings are subject to change at any time.
