Fractal Progress
Three Things #191: October 19, 2025
As a builder and as an entrepreneur, I’ve often felt inclined to believe that everything interesting or valuable that could be built has already been built, and every potentially valuable company or startup has already been started. It often feel as if every good idea has already been mined. But every time I begin to feel this way, I see another example of how it’s empirically wrong. You need look no further than the latest unicorn, the latest apps and products and services that are continually appearing.
And yet, if you pay more attention, you begin to see a clear pattern emerge. It’s not just that more and more new ideas are constantly being thought up or discovered. It’s that the universe of possible ideas is itself growing; the frontier of possibility is continually expanding. This is, in the main, due to technology, which allows humans the capability to do things we couldn’t previously do, or even imagine that we might do.
This is what gives me awe. It makes me feel like a child, and maintain a beginner’s mind. New things are possible literally every single day—but as the possibilities expand, so does the difficulty of the next breakthrough. The process feels fractal: a breakthrough leads to rapid progress, which slows and slows further, until yet another breakthrough causes it to speed up again.
Here are three examples of this process.
Thing #1: Travel 🧳
One of the things that most makes me feel awe is travel.
I travel a lot, and every single time I take off on an airplane, or see an airplane taking off or landing, I feel a sense of awe and wonder. I have a rough idea of the physics involved, of forces of lift and drag and momentum and resistance, and I understand the theory behind why it’s possible, but that doesn’t detract from the wonder. There are so many impressive, awe-inspiring things about aviation: the thousands of people and companies involved in getting an airplane from point A to point B. The thousands of people it would take to build an airplane from scratch. The bewildering amount of policies, laws, and procedures involved in international travel. And, yet, it works pretty well, most of the time.
Just a handful of generations ago, almost no humans had traveled more than a few miles from where they were born. The number that had crossed oceans or changed continents could be counted easily enough. A journey across the ocean, or across a continent, was no minor undertaking. It could cost your life savings, it would take months or even years, and it was extremely perilous. Around 10% of the pioneers who attempted to cross from the eastern to the western United States in the nineteenth century never made it, and history is full of tragic stories of what happened along the way. Embarking upon such a journey often meant leaving behind family, friends, neighbors, and loved ones, and, before the railroad was complete, you genuinely had no idea whether you’d ever see them again.
Contrast that to travel today. Take a train or a taxi to the airport. Pass through security, sit down on an airplane, arrive at your destination a few hours later. A bit tiring, a bit frustrating, and not always the most comfortable experience, but it’s orders of magnitudes safer, faster, and more convenient than it used to be. I tend to remember this as I stare out the airplane window at the passing scenery, and I especially remind myself of this when delays occur: delays that take minutes or hours, not weeks or months like they used to.
And yet, travel doesn’t stop there, with jet airplanes. Some forms of travel are still really, really hard. America is chronically unable to build high speed rail, and ideas like hyperloops and ballistic transport seem impossibly far away. We can cross continents and oceans easily enough, but we could do it much faster. We’re not yet able to cross the solar system, to say nothing of interstellar or intergalactic travel.
This is one of the key features of the fractal universe that I keep seeing. Once one milestone is accomplished, the next one almost always immediately presents itself. It always involves a phase shift, a massive step up in difficulty that’s going to take an entirely new technology and a new approach to accomplish, and yet we always get there eventually.
We’ve made quite a bit of progress on space travel recently, and I think I just might live long enough to see it become ordinary, or even to see Mars colonized, though I doubt I’ll make it there myself. I definitely won’t live long enough to see interstellar or intergalactic travel, but it is just barely conceivable. This is one area, at least, in which the next few challenges and opportunities, the next few stages, are pretty clear. Ad astra.
Thing #2: Compute 🧮
When we talk about things that have made exponential progress, compute is the first thing that comes to mind. Today’s computers are tens of millions of times faster and more powerful than computers a few decades ago. They’re far more reliable and cheaper, to boot. The first real computer I used as a child was around 10MhZ, had no hard drive, and read files off floppy disks at a painfully slow speed. Today, I carry a computer in my pocket that’s nearly a billion times faster.
But fractal progress is at work here, too! With a few rare exceptions, it’s simply not the case that the things we do with computers today happen millions of times faster than they did a few years ago—because the things we use computers for have changed many times over the same period. In other words, there have been many subsequent paradigm shifts, which have kept compute that’s infinitely fast, easy, or cheap just over the horizon.
When I began using a computer, I was using a command line; today everything is graphical. Applications today are also orders of magnitude more complex, and more compute intensive, than applications back then. To some extent this is because the software is genuinely more featureful and powerful; to some extent, it’s also because programmers today have gotten lazy due to an abundance of resources, and have largely stopped worrying about things like performance or memory management. This makes writing software easier and more accessible, but it makes programs slower.
These new paradigms are gobbling up all of that excess capacity, and then some. As a particularly extreme example, consider blockchain. You can run an application locally on your own device, which costs basically nothing. Or you can run the same application on a blockchain platform like Ethereum, where it runs on thousands of nodes around the world in concert, in a deterministic fashion, and those nodes all reach consensus about the execution path and outcome of the program. This process could cost anywhere from a few cents to thousands of dollars, depending on the complexity of the program and on how busy the network is. If this sounds a lot like the expensive time-sharing of computers that was common in the sixties, that’s because it’s quite similar: as a paradigm, blockchain computing is roughly where mainframe computing was 60-70 years ago. (Although it’s making much more rapid progress.)
In other words, in a sense, we’re back where we started. Due to limitations of the medium, the Ethereum “computer” has about as much memory and as much compute power as a pocket calculator. I’ve written programs for that computer, and it’s challenging, but also a lot of fun: the constraints require you to become quite creative, and to become a better engineer.
As if that wasn’t bad enough, other cryptographic technologies take this even further. There are several high profile initiatives to build a zkVM, i.e., a zero-knowledge virtual machine, which has all the same benefits as EVM, the Ethereum Virtual Machine, but has the added benefit that they also give you a ZK proof of faithful execution when a program finishes executing. These currently run around 1,000x-100,000x slower than running a program “natively” on Ethereum. This represents yet another paradigm shift, built on top of another paradigm shift. Running a program in a zkEVM could therefore be as much as a billion times slower than running the same program natively on modern hardware (although here, too, incredibly rapid progress is being made).
Then there’s FHE, fully homomorphic encryption. This is an extraordinarily powerful, high potential cryptographic technique that allows you to run a program on a data set without actually having access to the contents of the data set. This can also be around a million times slower than running a program natively.
It doesn’t stop there. The list of paradigm shifts in computing is long, and becoming longer by the day. Quantum computing and AI both represent enormous, recent paradigm shifts that are more compute-hungry than any other application. Model training and model inference, AI use cases, are currently very slow and expensive, but they’re much cheaper than they used to be. Like quantum, they also require completely reimagining the design of a program: GPUs, which power AI compute, are designed to perform relatively simple calculations with an insane degree of parallelization, and they can’t run programs written for ordinary CPUs. Things get even weirder with quantum, which discards traditional bits, bytes, and binary computing in favor of qubits and quantum registers.
Progress in computing is quite possibly faster than progress in any other field, and it’s fun—and basically impossible—to imagine where the field will take us in another 10 or 20 years. What will become possible in a future world with computers a billion times more powerful than today’s computers?
Thing #3: Society 📜
This thing is a bit less obvious and a bit less visible than the other two, but it’s no less important. The other area where we’re making exponential progress as a society is in the way we organize, solve problems, and get things done together.
The most visible example here is companies. Companies have grown bigger than ever before by any metric: in terms of headcount, addressable market, output, and of course market capitalization and profits. In just the past few years, for the first time ever, the market cap of the largest firms has exceeded one trillion dollars, an unimaginable sum of money.
The same could be said of political entities, most obviously the nation state. We have countries with over a billion people, and with GDPs in the tens of trillions of dollars. Countries are able to harness and organize resources including capital and labor to an extent never possible before.
But here, too, we’re in the midst of a paradigm shift. The way that humans communicate, cooperate, and coordinate is changing rapidly due to the Internet and modern telecommunications tools, but also due to other reasons: the proliferation of English as a global, professional language, easier global travel and trade, translation tools, and blockchain technology, to name but a few. Firms solve for many problems and make doing business faster, easier, cheaper, and safer, but they’re imperfect and introduce their own challenges and shortcomings. DAOs, while immature and also imperfect, offer an exciting alternative.
In much the same way that interplanetary travel represents a paradigm shift for travel, and crypto, AI, and quantum represent paradigm shifts for compute, the DAO represents a paradigm shift for organized endeavor. It’s slower, more expensive, less efficient, and overall uncompetitive at the things companies do well, which is typical of disruptive innovations. But DAOs have superpowers and can do things that companies and even nation states cannot do.
The first, biggest, and most successful DAO is Bitcoin. There’s no company behind Bitcoin. No CEO, no Board, no corporate policy. And yet it’s taking over the world. Bitcoin would never have worked as a company. Its inherently international, crypto native design means that it’s not subject to the same rules and regulations that burden companies, especially those that work globally like Bitcoin. The lack of centralized human leadership makes Bitcoin trustworthy in a way that no company or country could possibly be. It makes Bitcoin appeal to a much wider audience than it otherwise would—”money for enemies”—and it makes Bitcoin censorship resistant and unstoppable.
The parallels with the paradigm shifts in compute are legion. It’s trivially easy to write a program to run natively on a single modern CPU, but writing the same program in such a way that it can run in parallel across thousands of GPU cores, or thousands of crypto nodes inside a VM, is much harder. By the same token, it’s trivially easy to design an organization to solve a task using traditional corporate tools: holding companies, shell companies, operating companies, LLCs, trusts, etc. DAOs currently aren’t capable of the same degree of complexity, and they’re slower and more expensive, but they’re getting better quickly. I strongly suspect we’ll see complex DAOs begin to rival even the largest, most complex firms in all of the size metrics I outlined above: the rise of network states and the one-person unicorn portend what’s to come.
This is truly just the beginning, and it seems self evident that these novel organizational structures will continue to have a massive impact on how humans organize and build things together.