Tahitian Women on the Beach | Paul Gaugin (1891)

2:1 Fitness Landscape

Imagine yourself on an island in the Galapagos surrounded by hundreds of different hummingbird species. Each bird is uniquely adapted to thrive on a single isolated island, slowly shaped by millions of years of collective trial and error. 

Amassing enough nectar day to day is a great way for your average hummingbird to stay alive. However to thrive, attracting a mate who has also amassed enough resources to stay alive to mating age ensures that their uniquely well adapted genes survive to the next generation. 

On this one special island, hummingbirds with especially long beaks have an evolutionary advantage, as they are the only sub species of hummingbird capable of reaching the nectar produced by an especially long stemmed flower.

Female sword-billed hummingbird (right) compared to a buff-tailed coronet with shorter beak

Female sword-billed hummingbird (right) compared to a buff-tailed coronet with shorter beak

Evolution acts as a crude but effective system selects for defining "fitness", or how well adapted individual agents are to an ever changing environment. Agents in this simplified example being both the long stemmed Hummigbird, and the long stemmed flower which both mutually co-evolved to survive together in a beneficial relationship.

Advantages (unfortunately) do not last forever, no matter how well adapted you are to your current environment. One sea captain, arriving with just one ship, containing just a few rats, can be enough to upset the entire evolutionary order.

On an island full of Peregrine falcons, such an event would barely register as a blip on the evolutionary radar. Within days, the falcons would spot the helpless rats scurrying about and that would be the end of that. 

On hummingbird island sans apex predators, there is a drastically different story that unfolds something like:

  1. The rats gorge themselves of hummingbird eggs and exponentially multiply

  2. Eventually there are no more hummingbirds left to lay more eggs

  3. The rats descend into cannibalistic warfare until they all die, or a few super rats survive adapted to their new egg-less landscape.

  4. Many complex feedback loops then continue to cascade across the island, including the extinction of long stemmed flowers as they are no longer capable of being pollinated.

  5. At this point you must be thinking: what does this have to do with distributed ledgers and distributed applications?

Rat Island

If we can determine a rough proxy for what evolution selects for (in our landscape of databases that run payday loans, credit default swaps, supply chains, academic records, etc.) we might be able to narrow our focus to select what a winning distributed ledger rat might look like. 

In a dynamic fitness landscape, agents are constantly looking for ways to amass resources as a ploy to more effectively pass on information. Humans are unique in that some of us care about more than passing on just genetics, but also have a deep need to pass on knowledge, art, political prestige, and other more esoteric pieces of information. For this reason, we must not just look at the technical merits of each ledger project, but also the socio-cultural elements.


Regardless of motivation, the fitness selection process to determine what information gets passed on happens in two main ways. 

The less risky strategy

Where agents try to climb higher up the current fitness peak. This Hill Climbing algorithm incrementally searches for small optimizations that help agents become more fit over time.

On hummingbird island, this is the slow process of adapting ever longer beaks over millions of years. On art island, L'Académie des Beaux-Arts incrementally made paintings more and more realistic over hundreds of years to achieve an optimal local peak of perfectly capturing the nuances of an aristocrat's wig. 

But what happened? How did the impressionists succeed in making such a bold new form of art favoring the emotional impression a scene conveyed, over such rigid realism?

The more risky strategy

Eventually you'll reach a local fitness peak and have no where to go but down. 

So a few brave agents (like Manet, Monet, and Degas) boldly jump off the current fitness peak, deeply unsatisfied by the status quo. This jump into the abyss is a terrible idea for most people, as jumping from safety into the unknown gives a much higher probability of not passing on any information at all. (as the numerous test pilots killed trying to break the sound barrier, or settlers trying to across the Atlantic in rickety boat can attest)  

While being de-fueled this Bell X-1-3 exploded, destroying itself, the B-50, and seriously burned the pilot Joe Cannon

While being de-fueled this Bell X-1-3 exploded, destroying itself, the B-50, and seriously burned the pilot Joe Cannon

If more than a small handful of the population employs the more risky strategy, there is a large chance the population will cease to exist, as they have all jumped off their evolutionary life raft, and into the abyss. 

This is why low cost index funds are so great for most people. There is a market capitalization selection bias that constantly culls the herd. New and successful companies enter the index, while failing ones exit. As an investor, you can "safely" park your money in an index and be as abstracted as possible from the gyrations of individual fitness peaks, by owning the collective basket of most fitness peaks. (E.g. if Xerox becomes noncompetitive and drops out of the index, Apple or another newer successful player will replace it proportional to their market cap weighting)

Borderless friction reduction: or How the West was Won

With only smalls grains of salt needed, we can expand our heuristic around the small jump/big jump fitness landscape much further. This elitist, culturally insensitive, and impossible to prove empirically story goes something like: 

  1. A small portion of the European population was enticed by an entire "new" continent of natural resources. The jump across the abyss to a new local fitness peak required massive sacrifice. Not only did they risk death by crossing the Atlantic ocean on rickety sailboats, they often gave up their life savings for a ticket, or worse subjugated themselves into indentured servitude to make the trip.

  2. Fast forward a century or so and the process repeats again, this time leaving everything behind on the east coast of the United States to risk it all crossing the country in a rickety wagon in search of rich farm land and literal gold.

  3. Fast forward another century, and the survivors of the survivors of the survivors have created a hot bed of innovation risking their excess capital on new business ventures with an overwhelming likelihood of failure.

The major difference between the first two waves of risky jumps to new fitness peaks with the third "Silicon Valley" jump, is not that the evolutionary process is any different, but that the speed at which information flows is speeding up the entire process.  

Rather than pulling talent in from local or even national borders, a global competition is a foot in the 21st century to fund the best ideas.

When Peter Thiel spoke to a group of Stanford students in 2005, they asked what the best strategy was to secure a spot working for the next Google. Thiel replied that the best strategy was to look within a 5 mile radius of the classroom they were in.

Shockingly accurate, unbeknownst to anyone in the room at the time, just 1.7 miles from Stanford campus in a small Palo Alto garage, Facebook was solving real identity in the social media space for the first time at scale, and amassing a treasure trove of freely given user generated spy data. While Thiel didn’t know about Facebook specifically at the time, he did understand the fitness landscape enough to know Silicon Valley was THE local fitness peak circa 2005.

Ask Thiel the same question in 2018, and he is much more apprehensive about declaring where the next Facebook will come from, putting the odds the next earth shaking success will come from within the confines of the bay area at much less than 50%.

For the fitness landscape in a decentralized distributed ledger world has truly shifted. Those brave souls looking to jump into the Abyss can do so from the comfort of their own homes anywhere in the world. As our titular hero Xoana from the introduction to our book shows, anyone with an internet connection can participate as full owners in this new world.

When Thiel put his $500,000 dollar check into Facebook in 2005, the concept of an initial coin offering was still more than a decade away. If the venture capital market at the time was truly liquid and competitive, anyone in the world could have competed with Thiel to get the first checks into Facebook.

However, at the time massive barriers to entry precluded all but the most blue blooded (accredited) venture capitalists from being able to participate in such high-risk high-reward value creation.

Today, despite a lack of even the most basic investor protections, for the first time anyone on earth can participate in the probable downside / possible upside of any distributed ledger token project by buying a portion of the project directly.

Aggregation Hoarders

Where does the story go from here? If your good idea is good enough, isn't it copied down to the pixel by Facebook (cough cough Instagram stories)? Or acquired by Google? How can anyone hope to compete with the death star that is Amazon relentlessly squeezing every drop of margin out of existing legacy retailers? 

These massive technology companies are the latest rats to feast on hummingbird eggs, but are by no means the last. 

Remember the value creation chain from the beginning of Part I where raw materials are turned into finished products? Computers allow companies that can best harness the powers of automation (re: big data spying) to succeed where others fail, by drastically increasing the efficiency of turning raw materials into finished goods.

Facebook and Google combined employ around 100,000 full time staff, a tiny fraction of the numbers of employees traditional newspapers needed to run their business. They are also the first companies in history to take advantage of aggregation theory on a global scale, something their newspaper ancestors could only dream of. 

Aggregation theory states that the gatekeepers of information stand to benefit the most from distributing content.

Take an individual Pizza shop owner placing a coupon on Groupon. The pizza owner sacrifices significant margins in hopes of driving new customers to her door. The margins she sacrifices get collected by a single centralized intermediary which has comparatively little overhead, as they only employ a few programmers to reach a global audience. All tech companies from Facebook, to Google, to Netflix employ similar variations on the same theme.

Put another way, aggregating content drives all producers towards a commodity. What bargaining power does an individual local newspaper (or pizza shop, or movie producer) have to dictate the terms of negotiation with a Facebook, Google, or Netflix?

As quasi-monopolies, they can extract large tolls from each individual provider of content, as there is no other place the individual can go to get his or her message out to the world. (Yet)

  • In this system, you end up with a Google News feed that sucks out text from each individual newspaper, then places their own ads on top of a curated feed without paying the creators of the content.

  • In this system, you end up with pizza shops selling 20 dollar coupons for 10 dollars, then splitting 50% of the coupon with the aggregator, effectively selling 20 dollars worth of pizza for 5 dollars.

There must be a better way. Say a hypothetical magic technology that replaces monolithic centralized value suckers with a transparent and nearly free way of transacting value between any two peers. Since the days of the earliest empires from Mongol hoards to the Greeks and Romans, aggregating toll fees at the top of a hierarchy seems to be baked into human nature. We’ll see if the shift to DLT once and for all breaks this pattern of human misery, or accentuates it with the continued hyper concentrated ownership among early adopters.

Crypto Bourgeois returning the Means of Production

Hang tight, we know mentioning Karl Marx can illicit strong reactions, and the following is in no way an endorsement of his worldview. While Marx might have reached some crazy conclusions, one thing most people can agree on is that he knew how to deconstruct the way capitalism works. 

The basic theory is the owners of capital (the Bourgeois), are enriched by the workers (the Proletariat) who run their capitalist machine in exchange for wages. Importantly, wages ultimately get fed back into the machine through economic activity.

Under this lens, tech companies are the latest version of the Bourgeois that are evolutionarily better at sucking wealth out of the economy than their predecessors. (Remember, the pizza shop owner is going to raise prices indirectly to cover the marketing costs paid to FAANG)

Is the targeted ad cheaper and more efficient than the shotgun blast that is a traditional newspaper? Of course. The crucial difference between the crypto bourgeois and the centralized bourgeois is in the abstraction of how capital is returned.

Remember in a centralized system, value is hoarded by the owners of the black box, while in a decentralized system the only middleman income is generated from micro fees owners of the core protocol tokens make when transactions are registered onto the chain.

The Crypto Pay Day Loan Use case

Rather than continue to pick on FAANG companies, let’s quickly examine how pay day loans can be made drastically more efficient with distributed ledger technology to drive home just how fundamental the shift to ledger based value creation can be in this new fitness landscape.

What a worse way to pray on the poor than to charge exorbitant rates for short term bridge loans when people can’t make ends meet. However, as any student with a basic corporate finance class under their belt can tell you, these exorbitant rates are perfectly rational for unsecured non-recourse debt. As every defaulter drives the rates everyone else pays for unsecured debt higher, investors must be compensated for the additional risk that their capital might not be returned.

In this respect, the current fitness landscape looks something like a local high for those that have access to economies of scale capital that allow them to amortize the cost of bad debt across enough repayments.

But what happens to their business model when everyday people wanting a higher return on their capital can efficiently compete with the centralized incumbents?

In our potential distributed ledger future:

  • Scheduled payments between employer and employee are not ACH payments to a bank account, but rather token transfers from the employers wallet to the employees wallet via a smart contract.

  • As such, automated garnishment of wages and other assets becomes much easier.

    • Today, pay day loan companies want title to something with value to secure the loan if possible be it car, boat, or even furniture in case you default.

    • Tomorrow with distributed ledgers, a future could exist where this process becomes vastly more efficient. Things like digital scarlet letters become achievable where title transfer of secured assets automatically takes place when the terms of the loan are breached, then any additional income income made is automatically siphoned off to repay debts after an automated bankruptcy proceeding.

You can imagine how surreal your self driving car repossessing itself must feel.

If we believe such automated logic is possible, we can arrive at a future much like our dBay example from chapter one, where a centralized intermediary like a credit card company, bank, or pay day loan outfit now has to compete with the general public to provide the lowest cost access to capital.

If this sounds remarkably similar to existing peer-to-peer lending services you would be correct. Like all DLT innovation, the business logic of a centralized peer-to-peer lending platform, vs a decentralized platform is identical.

Both would have a marketplace of loan seekers and loan investors posting the terms they are looking for. If there is a match, then the loan is generated. The key difference is the centralized platform:

  • can unilaterally charge maker/taker fees set to whatever the market will bear

  • can theoretically front run orders & modify terms at will

  • get hacked, shutdown, co-opted by state actors, enter into hidden price fixing collusion with competitors, etc.

A decentralized platform could try the same tricks, but would eventually be caught by the paper trail of digital receipts or more quickly by raw capitalism. This gets at the core value add of distributed ledger based systems, fees tend towards zero as there is endless competition to more cheaply facilitate provably secure transactions. Sure one loan platform can be shady, but its code can be quickly tweaked for more equitable sharing that attracts more users than the corrupt platform.

The real question is at what price point per transaction is the value add of a decentralized pay day loan platform so great and the network effects so high, no further competition will come in to erode marketshare from one distributed platform processing the transactions to another better/faster/cheaper platform.

While Groupon intermediaries, Google news intermediaries, Netflix intermediaries, and pay day loan intermediaries, might on the surface seem like separate problems to solve, they all look remarkably similar from the vantage point of replacing centralized fallible black boxes, with provably secure transparent boxes.

Reinforcement Learning

In Part I, we learned the basic heuristics around what technical properties a good distributed ledger platform might have. In Part II, we are adding new heuristics around the socio-cultural properties a good distributed ledger might have.

In this way, we are in effect using the same strategies an artificial intelligence uses to make decisions. This strategy called “reinforcement learning” is a remarkably effective system for winning in an ever undulating fitness landscape.

In traditional reinforcement learning, an artificial intelligence is given a task, say make the score of Pac-Man as high as possible.

To accomplish this task, the A.I. begins exploring its environment by randomly inputting values to move the Pac-Man character around. Over time, the system learns to avoid Ghosts, and go for dots to maximize the score. However, it is very easy for the A.I. to get stuck on a local fitness peak, unable to make a cognitive leap to a better strategy outside of the local fitness peak.

For instance, in Pac-Man, say hypothetically the best players in the world have developed heuristics to:

  • NEVER go into the lower right quadrant of the map during end game play.

Through years of trial and error, they have learned that strategy will get them hopelessly trapped in the corner and thus lose them the game. With an A.I. we would not want to artificially limit the exploration space in this way, even if the A.I. develops similar logic to always avoid a certain losing behavior. Instead, we want the A.I. to occasionally explore this space as it might come up with a novel new strategy. Unlike people, computers have time to waste limited only by the computational resources provided.

To combat the aversion to seeking new fitness peaks, every once in a while programmers introduce a logical override that intentionally replaces what the A.I. thinks the best strategy is, with a random value, or different goal to optimize for. While the vast majority of the time random mutation will end in failure, every once in a while the shake up is helpful towards finding a better fitness peak.

Of course, constantly applying random strategies is a poor strategy for resource efficiency. We (like the A.I.) want to find the best strategies as soon as possible to out-compete everyone else in the fitness landscape. To not limit ourselves to closed ways of thinking, our valuation framework will be open ended, yet needs structure to prevent squandering resources on projects with defects that can hamper their ability to succeed in the market.

Our basic A.I.-like heuristics will be used to uncover the best places to apply distributed ledgers to existing centralized blackboxes, and use deductive then inductive reasoning to work backwards which distributed ledger solutions will work best to replace or augment those legacy systems.

In the next chapter, we will explore how this strategy will change greatly depending if open internet-like ledgers succeed over (or in conjunction with) private intranet-like ledgers. As in the original internet, the same fundamental technology underpins both public and private networks, making the choice for one over the other a socio-cultural issue, rather than a purely technical issue.

2:2 Public Ledgers Matter ->

Nobody phrases it this way, but I think that artificial intelligence is almost a humanities discipline. It’s really an attempt to understand human intelligence and human cognition.
— Sebastian Thrun