Compounding Crazy

Created

Jan 19, 2023 8:01 PM

Die Progress Units and the Law of Accelerating Returns

In 2015, Tim Urban wrote a two-part series on artificial intelligence (AI) in his deservingly wildly popular blog, Wait But Why. Because the piece covered some pretty wild ideas, Urban kicked it off with a 2,000 word preamble (a man after my own heart) to explain to his readers why they shouldn’t dismiss what he was about to write as too sci-fi or futuristic.

Urban’s big idea here is that human progress is exponential. In the present moment, though, it’s hard to recognize or comprehend that the graph is about to go vertical.

Source: WaitButWhy

As Urban explains it, “You have to remember something about what it’s like to stand on a time graph: you can’t see what’s to your right.” It’s possible to look backwards with relative certainty; it’s impossible to look forward with any certainty, particularly when progress itself is increasing at an increasing pace.

To illustrate the point, Urban wrote up a thought experiment:

Imagine that you got in a time machine back to 1750, grabbed someone alive then, and then brought them back to the present to show him around and watch his reactions. Everything we take for granted -- cars, tall buildings, iPhones, live sports broadcasts, recorded music, Google Maps -- would blow his mind in such a major way that he wouldn’t just be shocked, he might actually die.

If that guy wanted to pay it forward and kill someone with shock by bringing them to 1750, he couldn’t just go back ~250 years and grab someone from 1500. 1750 and 1500 are too similar. There would be some new things, but nothing that would kill the 1500s guy. To get the same effect, Urban guesses that he would have to pick someone up who lived around ~12,000 BC, before the First Agricultural Revolution. That guy, in turn, couldn’t just go back 12,000 years to 24,000 BC, “he’d have to go back over 100,000 years and get someone he could show fire and language to for the first time.”

He called the amount of time you’d have to go ahead to be shocked enough at the world to die a Die Progress Unit (DPU). The DPU is a colorful illustration of an idea from futurist Ray Kurzweil, the author of The Singularity is Near: the Law of Accelerating Returns.

The Law of Accelerating Returns states that the rate of change of progress accelerates because humans can use the technology at their disposal to progress faster than previous generations could without the technology.

This is demonstrably true, using DPUs or actual numbers, but according to Urban, there are three reasons we can’t grok the impact of exponential progress:

When it comes to history, we think in straight lines. It’s really hard to grasp how much crazier the future is going to get, because most people extrapolate based on the past growth rate, and even the more optimistic and adventurous ones will extrapolate from the present growth rate.

Source: WaitButWhy

If you accept that the rate of progress is accelerating, then it follows that using past or even present rates undershoots.

The trajectory of very recent history often tells a distorted story. Making this even harder, in the short-term, assuming an accelerating growth rate might make you look silly or lose money. The progression isn’t smooth. A period of rapid growth might give way to a period of slower growth. Progress happens in S-Curves -- similar to the Gartner Hype Cycle that we so often talk about here -- and it might even look like we’re regressing in certain areas at points. But over time, if you zoom out, those are just squiggles on an exponential curve arcing up and to the right.

Source: WaitButWhy

Our own experience makes us stubborn old men about the future. We have lived through past growth rates. We’ve never lived through future growth. Therefore, it’s a lot easier for us to intuit what the future will be like by extrapolating linearly based on what we’ve actually experienced. Yes, things will be different, we all know that, but they’ll be different in predictable ways. We expect that anything that deviates too far from our past experience, or historical averages, will revert to the mean.

So we’re making progress, and progress keeps getting faster, but how does that happen?

Two years later, in a 2017 post called Neuralink and the Brain’s Magical Future, Urban introduced another useful analogy: the Human Colossus. We evolved from blobs of cells that did nothing to humans who learned to communicate with language to people who could mass produce books and share them across the world, at which point humanity became the Human Colossus, essentially one large organism with shared knowledge. Computers and then the internet turbocharged that knowledge (and tool) sharing to the point that, today, you can go online, connect with anyone, download practically any information, and even fork full codebases or download designs for physical things.

Source: WaitButWhy

By connecting to the internet, any individual human can access all human knowledge. Things that took thousands of people millennia, centuries, or years to develop can be used and remixed by one person in one second. We have more building blocks than ever before, and we can use those building blocks to make new building blocks more quickly.

The takeaway is that human knowledge compounds, we build new things on the shoulders of giants at a faster and faster pace, and our brains are not good at understanding all of that in the moment.

If you pause, though, you can feel it happening all around us, right now.

The Venture Funding Rorschach and $320T Global Markets

There’s a lot of new and hard-to-understand things happening all at once right now. To simplify, let’s start somewhere familiar and quantifiable: venture funding.

This chart captured the attention of the venture and startup community over the past couple of weeks. It shows the total US venture capital activity over the past decade.

Source: Pitchbook-NVCA Venture Monitor

US venture capitalists invested more money in 2020 than in any previous year: $164 billion. In 2021, they’ve already invested 91% of that amount, $150 billion, split evenly at $75 billion in Q1 and $75 billion in Q2. That’s for only half the year. If they keep it up, VCs will invest $300 billion into startups in 2021, 83% more than they have in any other year.

The funding chart is a Rohrschach Test.

If you think that we’re living through a Tiger-and-COVID-fueled anomaly, maybe your brain projects out something like this:

You’re not a luddite, for chrissakes. You get that venture funding will increase. But come on. It’s a little overheated out there. We were already growing at a 15.4% CAGR in funding from 2011 to 2020 (we’ll throw out 2021 for now, weird year), it’ll probably revert to the mean.

If you subscribe to the DPU / Law of Accelerating Return view of the world, though, you might see something very different:

I’m not a blind optimist! Sure there will be down years! But if the curve is exponential instead of linear (as history, Kurzweil, and Urban would suggest it will be), it will look something like this. In fact, even I’m being conservative here. If I drew the curve out based on the 20.8% CAGR from 2011 to 2021, we’d be at $1.9 trillion in 2031, twice as highas the last bar on the chart above, and that’s assuming no acceleration in growth. Imagining accelerating growth is really hard, even for someone like me whose pessimist gene was lost at birth.

It’s even hard for Tiger Global, the fastest-moving venture investor in the world. The FT reported last week that Tiger had deployed most of the new $6.7 billion fund it raised in March by June. $6.7 billion in three months! According to the FT:

Writing to investors in June, the firm said it had “consistently underestimated” the market for private tech companies. Six months earlier, data suggested a $3tn market opportunity. It was now closer to $5tn, the firm said.

While that’s obviously self-serving -- Tiger needs to explain why it spent its investors’ money so quickly and why they should give them $10 billion more -- it’s also instructive. If Tiger is consistently underestimating the market’s potential, the rest of us are probably dramatically undershooting what’s going on.

Tiger’s rapid deployment strategy may be right for all of the competition-based reasons that Everett Randle highlighted in Playing Different Gamesand because they might have picked up on a more permanent acceleration of the market before anyone else did. If this is still, somehow, early innings, you want to put as much cash into fast-growing companies as possible.

That said, it’s a big leap from $300 billion today to $2 trillion in venture funding per year a decade from now. Particularly because that money needs to generate returns.

Source: Correlation Ventures

Based on my very rough estimates taken from this venture returns distribution chart (take the midpoint of ranges, assume 30x for the 0.5% of funds that return >20x), we can come up with an average cash on cash return for a US venture fund: 2.13x. That means that to generate historically average returns on $2 trillion, there would need to be startups worth ~$4.25 trillion created every year. That seems… implausible, given that, as I wrote in my Not Boring Capital memo, all of the world’s unicorns are worth only $2.37 trillion today.

That information is actually stale now. Since I hit send on July 12th, 29 new unicorns have been minted, and all 779 are worth a combined $2.485 trillion, now slightly higher than Apple’s $2.41 trillion market cap.

Anyway, that means that every year, we’d need to produce nearly twice as much unicorn value as exists in the world today. And not just on paper, we’d need actual exits! Again, seems implausible when you look at the size of the total market today.

All of the public equities in the world are worth roughly $117 trillion by combined market cap. Of that, $36 trillion is in the S&P 500 and $20 trillion is in the Nasdaq 100 (there’s overlap between the two: Apple is the largest company in both). That would mean that over time, just five years’ worth of venture-backed companies would need to be worth more than all 100 companies that make up the Nasdaq index today. This keeps getting more implausible!

But markets are not static. Based on data from the World Bank and Sibil Research, the combined market caps of all of the world’s publicly listed companies has grown from $1.138 trillion in 1975 to $117 trillion today, a roughly 100x increase, or a 10.6% CAGR over the 46 year period.

Source: World Bank

Check out the fancy gif I made above, of global public equity market cap growth a decade at a time. At the end of each decade, looking back, the graph looks impossibly steep, like it couldn’t possibly continue to go up. Then it does. That’s important to keep in mind when you look at this graph, which projects global public equity market caps forward a decade at the same 10.6% CAGR.

Based on that projection, $203 trillion in public equity market cap will be created in the next decade, nearly double today’s combined market cap in new value alone.

You can do pretty much the same thing just looking at the tech-focused Nasdaq.

Source: MacroTrends + Not Boring Projections

Just the 100 largest US tech/tech-ish stocks today are worth $20 trillion. At a 10.5% CAGR, that grows to $54 trillion in a decade. Just to be clear, the Nasdaq doesn’t even include some of the largest global tech companies, including Chinese giants Tencent and Alibaba.

Both projections are smoother than the future will actually be. Things could totally fall apart. But if history is a guide, we’ll wake up a decade from now to a market full of much higher numbers than we can reasonably wrap our minds around today. Some of the new market cap will come from today’s public companies, some will come from today’s private unicorns, some will come from existing startups that aren’t yet unicorns, and some will come from companies that don’t even exist yet. It’s impossible to predict what the mix will be.

But what’s wild is that that’s what the current $300 billion in venture funding is investing into -- things funded today might not be public for another decade. The $2 trillion in hypothetical 2031 funding, which would need to return $4.25 trillion per year, would feed into the following decade. By 2041, at the same CAGR, the world’s public equities would be worth $877 trillion, of which $147 trillion would come from the Nasdaq.

That seems fucking insane. And, you’re probably screaming at me through your computer, all of that is just based on typing x*(1+10.5%)^10 in a spreadsheet a couple times! That’s easy. You can make a spreadsheet say anything. How could we possibly create all of that value in two decades?

For any of this to be at all believable, we need to get back to innovation itself, and for that, we need to go back to the Human Colossus.

The Compounding Colossus

Numbers have a clear scoreboard. It’s easy to tell when they compound. It’s not as clean to capture how innovation compounds, but it does too.

I’m not going to do a better job explaining the idea than Tim Urban, so let’s go back to the Human Colossus (emphasis mine):

If an individual human’s core motivation is to pass its genes on, which keeps the species going, the forces of macroeconomics make the Human Colossus’s core motivation to create value, which means it tends to want to invent newer and better technology. Every time it does that, it becomes an even better inventor, which means it can invent new stuff even faster.

Humanity has spent its whole existence building things that become primitives for a future generation of builders. Every new invention is an output of centuries’ worth of innovations, and then voila, it becomes a simple, cleanly-packaged input for the next invention. Like a good API, each new invention abstracts away all of the complexity that went into its creation.

Let’s use tires, a relatively simple example. The wheel was invented in the 4th century BC in Mesopotamia. Ancient South and Central Americans used rubber to make balls for games as far back as 1600 BC. In 1844, Charles Goodyear vulcanized rubber, in 1847 Robert W. Thompson patented the air-filled rubber tire, and in 1888, an Irishman named John Boyd Dunlop commercialized the first successful pneumatic tire. 120 years later, in 2008, the Tesla team used an evolved version of that tire as one of many, many inputs, each with its own long and winding evolution, into its first Roadster. The Tesla team didn’t need to reinvent the wheel.

This same compounding can be seen everywhere in the economy. Every invention is the latest in a chain of compounding innovations.

The Time Crystal mentioned up top was only possible because Paul Benioff proposed the idea of a quantum computer in 1980 as a derivative of the Turing Machine, the idea for which can be traced back to Charles Babbage’s 1834 notion of a computing machine, then physicists at Los Alamos created a simple first prototype in 1998, Google developed its Sycamore quantum computer with no real use case in mind in 2016, and physicists brought them the idea of using it to create the Time Crystal, which itself came from its own long physics lineage. Now that we have functioning quantum computers and Time Crystals, new generations of inventors can take those as inputs into new and crazier things.

This overly simple graphic is also true: previous generations’ hard work makes it easier to build new things.

That’s what powers the increasing pace of innovation. Discoveries become inventions become building blocks become inventions become building blocks, ad infinitum.

Look at fintech. Amidst the venture funding boom, fintech has stood out from the pack. According to Pitchbook (via The Economist), VCs invested $34 billion into fintech companies in Q2 alone, 45% of the $75 billion invested. Makes sense -- The Economist points out that they’ve made $70 billion from fintech exits thus far in 2021, and that was before last week’s Robinhood IPO added $30 billion to the fintech exit scoreboard.

Source: The Economist

But we covered numbers in the last section. This section isn’t about the “what.” It’s about the “why.”

Fintech is a clear and early beneficiary of compounding innovation. Stripe, founded in 2009 and worth $95 billion, and Plaid, founded in 2013 and worth $13.4 billion, built much of the infrastructure on top of which a new wave of fintech companies can build more easily.

Robinhood doesn’t need to connect to your bank directly to get your money; it connects through Plaid.
Jeremy Smith, one of Equi’s co-founders, told me that they couldn’t have built their product without fintech infrastructure products like Plaid and Modern Treasury.
Not Boring favorite Ramp uses Stripe’s Issuing product to make and manage its cards.

At the same time, entrepreneurs continue to build more infrastructure. I’ve talked to two companies in the past week who are building card products on Lithic. Not Boring portfolio companies Unit and Embedded let other companies add banking and trading to their products.

Fintech infrastructure companies can create huge financial outcomes for their investors and enable the next wave of companies to build products that were previously impossible or at least very difficult, expensive, time-consuming, and imperfect. Their impact compounds.

I’ve written about the impact that off-the-shelf software has on new company development and speed before:

Logan Bartlett, a VC at Redpoint, guessed that just a decade ago, it would have taken seven or eight years for even a great company to build what Ramp did in two. And Ramp did it with a team of less than 100 people. Of course Ramp is worth more, faster today than a similar company would have been before.

API-first companies let other companies focus on their core differentiators instead of re-inventing the wheel. Young companies can get bigger, faster than ever before. You know the drill.

What I haven’t written about, or thought of until recently, is that being able to build faster and with fewer employees has compounding effects of its own.

For one, the faster a new company is built, the faster other companies can benefit from it. New primitives create new opportunities for innovation. Kind of obvious.

Less explored is the fact that if a new company needs fewer employees, because it can outsource whole functions to APIs and software, there are more talented employees freed up to go start their own things. If it would have taken 100 employees to build a customer-ready product before and today it only takes 10, then those 90 extra employees can go start nine more companies.

Even better, since many of the 90 employees in the only company would have been working on things like payments, bank integrations, authentication, scheduling, background checks, data labeling, and more, all of which was non-core and can now be plugged via API, it unleashes energy that can be used on innovation instead of execution.

In Andy Weir’s latest hit, Project Hail Mary, a character named Stratt described the past as “unrelenting mystery,” saying:

For fifty thousand years, right up to the Industrial Revolution, human civilization was about one thing and one thing only: food. Every culture that existed put most of their time, energy, manpower, and resources into food.

Mechanizing agriculture freed humans up to focus on other things. Technology has evolved quickly over the past couple of centuries as a result. I’m not comparing APIs to food, but a similar dynamic is at play: when people don’t have to focus on day-to-day survival, they can invent new things.

Back to our ten small companies. Not only can they build new things themselves, some of those ten companies will build things that become building blocks for new companies. Maybe their innovations will help bring the required team size down from ten to five, freeing up more innovation energy, and so on.

It also creates 10x the number of companies for investors to back and 10x the opportunities for big outcomes, without decreasing the potential size of any given outcome. Unless those companies directly compete with each other with practically-identical products, they might steal share from less tech-focused incumbents, expand the market, or create entirely new markets. Lighter-weight teams also lowers the cost of failure and lets people experiment more quickly on ideas before going all-in.

Building blocks let more new companies build faster and leaner.
Faster and leaner companies build more new building blocks.
Progress compounds.

This concept of compounding building blocks powers innovation well beyond fintech, too. Categories that seemed either out of technology’s reach or like straight-up science fiction just a decade ago now have infrastructure that makes building relatively simple.

It’s most obvious in crypto, where smart contracts are literally called “Legos” because of their composability. Everything is essentially open source, and new builders can snap in existing Legos to create exponentially more complex products. We’ve covered crypto a lot here recently, and you likely know what a big impact I think it will have, but if not, read this, this, this, this, and this.

NexHealth’s mission is to accelerate healthcare by connecting patients, doctors, and developers. They want to make it as easy to build healthtech products as it is to create fintech or crypto products today. There are trillions of dollars of value waiting to be unlocked in healthcare when new primitives make it easier to build new solutions.

Scale is building artificial intelligence and machine learning infrastructure that it hopes will bring AI/ML usage in companies from 8% today to all companies within the next two decades.

It’s literally impossible to predict the impact that widespread usage of AI and ML will have over the next few decades, particularly when they’re improving so rapidly. Tim Urban wrote the two-part Wait But Why piece I referenced earlier because, in his words, “It hit me pretty quickly that what’s happening in the world of AI is not just an important topic, but by far THE most important topic for our future.” The potential compounding impact of AI is what got Kurzweil so excited about the Law of Accelerating Returns in the first place.

Leaving our blue marble for a moment, space infrastructure has also reached the point where you can go to SpaceX.com and BOOK SPACE ON A ROCKET FOR YOUR SATELLITE more easily than you can buy a lot of older B2B SaaS. Seriously, I recommend clicking the link and going through the flow just to experience it yourself.

Source: SpaceX

That, of course, has compounding effects. Varda is stringing together a suite of commoditized space products the same way a software company would connect APIs in order to build space factories to make things that weren’t previously possible on earth.

SpaceX itself is using its launch and re-entry capabilities to bring satellites to space that will cover the globe in fast internet for $100/mo via its Starlink business. Akash Systems is going to do Starlink one better by delivering fast internet everywhere in the world for just $5/month thanks to its patented use of artificial diamonds to remove heat, driving down the size and cost of the satellites required.

These are absolutely mind-blowing accomplishments, only made possible by the Human Colossus’ compounding knowledge and innovations, but they’re also the building blocks for even more innovation in the future.

What is the world going to look like when everyone on the planet can access fast internet for just $5 per month?
What does it look like when you give billions more people internet anywhere and AI, healthcare, crypto, and financial tools infrastructure, all callable in a few lines of code?
What happens when you combine cheap, abundant internet access with tools like Replit that kids as young as 12 can use to learn code and build products?

It all compounds. Everything gets crazier, faster.

Even before accounting for cheap, fast internet everywhere, regions like Africa, Southeast Asia, and India are poised to experience enormous growth over the next couple of decades, enabling new innovators and billions more people in the global middle class. Increased digitalization, remote work, and crypto will knock down borders and level the global playing field in the coming decades.

And I haven’t even touched the Metaverse yet. In his Metaverse Primer, Matthew Ball touched on all of the disparate innovation that needs to take place to stitch together Metaverse, each piece of which is the result of a long chain of small, compounding innovations. Once the Metaverse is a reality and the next big platform shift, as Facebook certainly believes it will be, what craziness will that spur? What does the value chain look like when you’re dealing mainly in digital products?

What about when you combine that with crypto? The rise of games like Axie and the absolutely bonkers demand for NFTs shows that there is real demand for digital employment and status signals. Bernard Arnault’s ascent to the top spot on the Forbes list is proof positive that people are willing to pay to show off; the fact that it happened during the latest NFT renaissance is a beautiful coincidence. People have always and will always want to express themselves and show off. NFTs and the Metaverse open up new playgrounds for experimentation and status seeking, at a fraction of the production cost or environmental impact.

That progress isn’t linear. It’s exponential and entirely unpredictable. We’ve covered just a few of the countless innovations that will themselves spawn new innovations, and it would be impossible to model out the results of the permutations and interactions among the just the fields we’ve covered. Crypto barely existed a decade ago. What new categories beyond our anticipatory capabilities will emerge in the next decade?

All of which is to say: I know things seem outlandish right now, but it’s all just a tiny blip when you look right on the graph.

How to Play The Crazy

Here’s the rub: just because it’s likely that things will continue to compound doesn’t mean that there’s an obvious or easy way to play it.

First, there are the timing risks.

If the Fed jacks up interest rates, growth assets could tank short-term. Ham-fisted crypto regulation is a legitimate concern. Markets never go straight up. Even though global equities and the Nasdaq have compounded at a roughly 10.5% CAGR and ended up and to the right, there were a lot of bumps along the way. Please don’t mortgage your house and YOLO into risk assets. You need to be in the game to play the ride up.

Then, there are macro and tail risks.

What happens to the markets if the US Dollar somehow loses its status as the global reserve currency? The climate change threat isn’t going away any time soon. Will a further run up in the financial markets create such extreme wealth disparity that it causes violent unrest? Will governments tax investment gains, or even stifle innovation, in order to maintain the social order?

These being tail risks, there are countless definitionally unimaginable things that might happen. Perhaps the next global crisis won’t be such a boon for innovation. Unknown unknowns abound.

But let’s say that, broadly, over the next decade and beyond, the rate of progress continues to increase and the curve continues to steepen. Figuring out how to play that is still tricky.

Kurzweil and Urban both showed that progress increases exponentially, not necessarily that companies or markets capture value from that progress.
While global equities and the Nasdaq have compounded beautifully, their components come and go. It’s hard to predict which companies will capture the value created by compounding innovation.
If innovation continues to accelerate, there could be entirely new technologies or paradigms that knock existing leaders off. Mobile shifted the landscape dramatically; might crypto do the same to fintech? Is there another platform shift that no one’s anticipating that might dramatically alter the landscape?

Plus, I would actually be shocked if global public equities’ market caps actually reached $320 trillion in a decade, not because I don’t think that much value will be created, but because I doubt that so much of it will accrue to public equities. If crypto is as disruptive of a force as it seems like it will be, more and more of those trillions of dollars of value will accrue to token holders versus public market shareholders. If those token holders are the people creating value -- both makers and consumers -- that’s a fantastic thing, but it means that just closing your eyes and buying the S&P, Nasdaq, or other public market indices isn’t as safe a bet as it seems.

Concerns highlighted, what I am confident of is that dismissing what’s happening as a fad or temporary insanity will age poorly.

Everything is changing faster and faster. Everything keeps getting crazier. It’s impossible to predict specifically what’s going to happen.

So how do you prepare?

Open-mindedness will be important. You should continue playing the Great Online Game. Reading sci-fi will make the crazy seem more familiar (I just finished Project Hail Mary and it’s excellent). There will be abundant opportunities for people who are down to get a little crazy.

Zoom out. As Patrick Collison said when describing his admiration for Jeff Bezos to Ezra Klein:

There’s something quite deep about the notion of using time horizons as a competitive advantage, in that you’re simply willing to wait longer than other people and you have an organization that is thusly oriented.

Get yourself thusly oriented, jump in, and enjoy. The future is going to be crazier than you can imagine.

Not Boring Jobs

On Thursday, I launched Not Boring Jobs, the best place to find jobs at some of the fastest growing, least boring startups out there.

Today, I’m psyched to bring you a gem: Software Engineer at Cohere. The friend who told me about Cohere said, “Cohere is software that will change software forever.”

Thanks to Dan and Puja for editing!

Disclosure: some companies discussed are Not Boring Capital portfolio companies.

How did you like this week’s Not Boring? Your feedback helps me make this great.

Thanks for reading and see you on Monday!

Packy

See important disclaimer.