TL;DR

The internet could be amazing but we're really up against it with two challenges: unreliability and hostility. This is really one problem - of how power is distributed on the web today. New technical progress in cryptography, game theory and distributed systems can solve both problems by creating web services that are optimised to be built upon. Such a web would be more humane and capable of supporting an internet many times more productive than ours today.

THE INTERNET CAN BE BETTER

And it will be better!

Despite the mess we’re in, I want us to fall in love with the internet again - in love with its connective potential to escape drudgery, overcome loneliness and spark creativity. This mission won’t succeed unless we pull together brains and hearts across society to manifest a dream of a better web.

The internet wasn't built for tech bros and venture capitalists. It was built for everyone, and everyone should be included in the dialogue about improving it. Whoever you are - this conversation is for you!

BUT IT'S NOT SCALABLE, AND KINDA NASTY

Here are two (in my mind, the chief) existential challenges facing the web:

1. In 2074, what foundations are required to sustain an "internet GDP" that is 10,000 times bigger than today's web economy?
2. In that future, what infrastructure will guarantee the continued humane treatment of internet citizens?

In economics, two helpful complementary metrics to track human development are (i) GDP, which expresses the raw productive power of mercenary capitalism, and (ii) HDI, which qualitatively tracks the value of things that markets can't price. You can think of these two challenges as the GDP and HDI challenges of the web respectively.

The internet is growing, but as with any economy, its growth strains and falters when three key traits are missing:

• Public goods are too weak to facilitate strong markets,
• Local intelligence isn't imbued with local control,
• Creative people don't benefit from the products of their labour, and economic value accrues to rentiers intead.

The internet's early, explosive growth was predicated on strong public goods that manifested these qualities, such as the hypertext transfer protocol, the domain name system and the simple mail transfer protocol - this latter being the world's first and most successful decentralised digital social network! For this reason, we saw growth that reflected the dynamism of a city like New York, rather than the centrally-planned, monopolised growth of a city like Disneyland.

However, as the dominant web platforms grew unassailable moats, they demonstrated excess power over key web goods such as the social graph, app stores and attention algorithms. Creative people and businesses of all scales are forced to pay rent to platforms that exert arbitrary control. My conjecture is that the rentier economic structure of Big Tech in 2024 most closely approximates the feudal system, and that we're no longer optimising our web to be built upon.

At the same time, the web feels increasingly inhumane. Toxic content, electoral manipulation and attention-decay thrive on today's platforms. The web used to feel like a library that connected you with the world's knowledge, and a meeting-place for grassroots communities that wouldn't exist otherwise. Now it feels increasingly like a firehose of information pointed squarely at your face, creating echo-chambers of self-satisfied rage that drive people apart as they doomscroll into oblivion.

Some say that you cannot achieve economic growth without sacrificing our humanity, or that the platforms have somehow earned the right to ruthlessly extract value. I disagree.

Instead, I propose that these goals - a great internet economy and a humane web society - are mutually compatible. Not only that, I propose that neither is possible without the other. A humane web is the most economically productive, and the most productive internet is necessarily more humane in the long term.

These two perspectives - the pragmatic perspective of raw capital seeking to replicate itself at an ever greater rate, and the ideological perspective that hopes for a more humane web - are as two great wings propelling The Internet Beast into the sky. Should we lose either, our grand computational network will death spiral towards the ground.

But things can be better - they will be better!

WE NEED INFRASTRUCTURE THAT'S OPTIMISED TO BE BUILT UPON

In meatspace, we rely on roads, passports, the judiciary, and much else to facilitate strong markets and economic booms. These public goods are the preconditions of private creative labour. Similarly, the early internet provided us with solid, neutral foundations (HTTP, SMTP, DNS, PKI) that enabled a burst of innovation in the form of social networking, e-commerce, e-learning and more.

My next conjecture is that the success of these early protocols is best explained by the fact that they were, at every level, optimised to be built upon.

What does that mean? I think it means that they are reliable - that, if you are a creative person coming up with ideas, when you choose to use those technologies, you have strong guarantees that they will continue to work as expected, and will not change against your interests, without your involvement. Often this is because they are modified by democratic committee, or because they are simply open-source and locally-deployed. Importantly, it's often because there is no HTTP Rentier Corporation that will turn around and adopt tenant-hostile design once they achieve unassailable commercial moats.

This is really important to understand because it cuts to the heart of why the internet feels the way it does today. It's not to do with Evil Executives in Big Tech MegaCorps - instead, it's an unintended side-effect of the incentive structures baked into the socio-economic composition of the web.

The internet platforms we all know and love tend to follow a shared pattern of growth, evolutionarily fine-tuned to be absolutely devastating in its memetic power. This pattern is sometimes referred to as aggregation and comes in a three-part cycle. As a member of the first generation to grow up on the web, it's a cycle with which I am all-too famiilar.

1. A product or service is launched, often free, and is designed to create as much value as possible for its members. Think early Facebook, early Twitter or the old days of Uber when rides were 5x cheaper.
2. As it grows, the product or service demonstrates a network effect, whereby the value of the network increases for every new person that joins the network. This means that the biggest network will be the most valuable. Two-sided marketplaces such as Uber or social networks such as MySpace are great examples of this - the network that has the most drivers, or most of your friends' social activity, is the one you're most likely to use. Entrepreneurs pursue these precious network effects with dogged intensity.
3. Once a network effect is entrenched, the network transitions into a "platform" that is very hard for members to leave, because no upstart competitor can create a similar network effect. This becomes user-hostile when the stickiness is not maintained by continued innovation, but the mere presence of an underlying passive asset: the network effect itself. At this point, the platform is compelled to extract rather than create value, and starts charging exorbitant rents, spamming users with advertising or adopting otherwise user-hostile patterns. Platforms consequently decay over many years into spammy, rotten ecosystems and there's not much members can do to fight it.

When platforms are too defensible, they become a victim of their own success in the long term. It's notable how distinct this pattern of growth is from the early days of the web, and indeed the traits of any hyperscalable economy. As mentioned above, it misses three key features:

• Public goods are too weak to facilitate strong markets: for example, many businesses rely on social graphs, but social graphs are owned and controlled by platforms. To illustrate - Zynga faced enormous platform risk as they grew, and their games thrived atop of Facebook's social graph. Facebook could make no reliable guarantee that Farmville could continue to rely on their API as expected, without surprise rents or censorship. As such, Zynga eventually had to vertically integrate and depart from their original growth partner. This was really expensive for Zynga, and impossible for most players.
• Local intelligence isn't imbued with local control: for example, many businesses rely on open-source code, and this works really well because any business can extend that code with whatever augmentations they need. Contrast this with Google Maps, where to update the product in line with your desired features you have to hope and pray that the relevant product manager at Google chooses to prioritise the market niche you represent.
• Creative people don't benefit from the products of their labour, and economic value accrues to rentiers intead: for example, creators selling their wares on TikTok today risk being shadowbanned without warning or explanation, and platforms can largely dictate fees. In the old days, you could just port your audience relationships, embodied by an email subscriber list, from one provider (AOL) to another (Hotmail). Today, there is no option to "leave" TikTok without starting from zero, somewhere new.

Another way to put this, is that a hyperscale internet requires reliable infrastructure, and our legacy socio-economic model of creating platforms has severe theoretical limits on reliability. So, for any technology, or any institution, we can ask the following:

• Are the incentives behind this technology optimised for its tenants to do their best work, and to what extent?

Don't be a dogmatist about this. Many "BIG TECH COMPANIES" (ew, gross! 🤮 /s) have designed their network effects such that rentier growth and tenant growth are mutally reinforcing. In fact, it's hard to see how you create successful innovation without having a decent protective moat. I think of Apple as a company that has just enough of a network effect (interconnected devices and priority access to the OS) to create a healthy enterprise, but not so much of a network effect that Apple can afford to stop innovating. After all, people replace their smartphone every 2-3 years and Samsung is always an option.

It also seems reasonable that there's a time and a place to optimise a technology to be built upon, and that in the early years, the rapid growth of a centralised but rug-pullable system may be preferable - especially before an ecosystem has matured. Consider the demise of RSS!

But, as I say, we need more reliable and humane web infrastructure.

Today, we've graduated past the original decentralised technologies that guarantee reliability, and instead have a new generation of key web technologies:

• Social graphs
• "Gig worker" networks
• Large language models
• Recommendation algorithms
• Moderation algorithms
• And more!

What would it mean to design these technologies in such a way that they are 100 times more reliable? That they are optimised to be built upon?

ADVANCED DISINTERMEDIATING TECHNOLOGIES

Over the last few years there's been a stonking amount of progress in what I call "advanced disintermediating technologies". You can think of advanced disintermediating technologies as the growing bucket of cryptographic and game-theoretic innovations that disintermediate centrally-operated web services with reliable peer-to-peer alternatives.

This includes, at least:

• Zero-knowledge proofs: a way for peers to make attestations to each other, about any arbitrary claim, while inheriting the authority usually reserved for trusted third parties;
• Conflict-free Replicated Data Types: a class of data structures that facilitate consistency in distributed systems by allowing multiple nodes to update a shared state concurrently without requiring complex synchronization mechanisms. CRDTs are an important component of the burgeoning "local-first software" philosophy.
• Blockchains: a way for agents that don't trust each other to come to agreement on some shared state of the world in an incredibly reliable way. This makes it exceptionally powerful for applications (such as money) where the stakes are enormous and you're interacting with 000s of network participants.
• Oblivious compute: the ability for a computer to verifiably execute some expensive operation on some data without seeing (i) the operation or (ii) the data. This relies on nascent techniques like homomorphic encryption.
• And more!

As is to be expected, there's a messy, imperfect mapping between these ADTs and the centralised web technologies I mention above, such as social graphs and moderation algorithms. However, CRDTs and blockchains will eat away at databases, zero-knowledge proofs will increasingly augment social networks, and oblivious compute will start to operate key web algorithms, in ways that are hard to predict at the micro level but inevitable at the macro level.

Inevitable, because they are infinitely more reliable. These disintermediating technologies cannibalise extractive platforms and the unassailable network effects they rely upon. They are often open-source, locally-deployed and feel more like public goods. As such they are solid foundations, optimised to be built upon, and capable of sustaining an internet economy in 2074 that is 10,000 times the size of today's. Despite their nascency, their flaws and all the work to be done, they are a classic example of disruptive innovation, and therefore worth investment.

IT WILL BE BETTER

These reliable technologies are in part reliable because they disintermediate services by replacing institutions with pure software and mathematics. As such, they sidestep platforms and put users and businesses in the driving seat. When algorithms and social networks are open-sourced and run locally, our web will feel more humane, because there is no unassailable moat behind which platforms can mercilessly extract value.

How will this feel in practice? It's hard to say, because technology doesn't tend to solve existing problems until it has first created new markets that we don't have words for yet. The future will feel weirder and more wonderful than we can imagine. However, here are some guesses:

• If you're a business that uses a social network to reach your audience, you can leave that social network without losing your audience;
• If you're using a social network and don't like what the CEO is tweeting, you can leave without losing your friends;
• If you're using a social network and you want a recommendation algorithm that promotes less divisive or toxic content, you can pick your own;
• If you're a parent and your kids are using social media, you can curate a recommendation and moderation algorithm for them, just as you curate their meals;
• If you're a gig worker, you can set up a network with other gig workers rather than rely on a costly platform like Uber;
• If you're a company that wants to integrate AI, you can train a model without revealing confidential information;
• If you're an AI agent, you can sail across different platforms in service of your owner by following standardised protocols instead of bespoke data-sharing contracts;
• And much more (for what it's worth, I present these ideas because they are the most legible to the uninitiated - but they are probably not the first that will succeed. Our own ideas at Tonk are way out on the spectrum of weird).

In such a way, the web will gently shed user-hostile patterns and rewild itself in the pursuit of more humane experiences. We will have a web that not only is a kinder place to live, but grounds out on solid foundations capable of supporting an economy thousands of times larger than that we can conceive of, today.

A LIBERAL REPUBLIC OF THE INTERNET

It’s been a while since our last networking revolution, and it’s understandable that we forget how to think about new worlds and get mired down in the cynicism of the present - but another future is possible. We must remember the key role of imagination for those who created the early internet, or the first personal computers. The alternative is a future not worth countenancing.

Decades have passed, and I'm still optimistic about startups as a vessel from which one can hijack the course of human history. Startups need to change, but I'm confident in a very particular kind of startup over the next decade - their time has not yet passed as an effective organisational entity. After all, we are but surfers on grand techno-social waves initiated by hearts and minds far greater than mine.

So get in the car, loser. It's going to be so fun.

- Baz