[Podcast #7] Harry on Tonk and little internets

Watch the episode on Youtube or read the highlights below!

An introduction to Tonk

In our ecosystem, we don't think enough about users or regular people. Innovation is the most exciting thing, and solving problems in people's lives is the most important thing.

So when I shut my eyes and try to work on our copy and our story, I imagine my dad in front of me, my partner, or some of my partner's friends who aren't in the tech world.

It's very uncomfortable because it forces me to let go of words like ZK, blockchain, Web3, and platform overhead and explain to people in their own terms why we think what we're doing is exciting and why the internet is going to be okay rather than a fireball, a big raging inferno.

When it comes to how we want people to feel when they're using Tonk or when they're using products built with Tonk, we want the internet to feel little again.

You, as a human being, have agency. You are using the internet as a tool to interact with other human beings rather than being trapped within an endless doom scroll, rather than being trapped within a system that you are economically dependent on and that someone else essentially controls. When we say Tonk helps you build a little internet, that's what we mean.

Tonk is two things: a co-pilot-friendly framework and a network for building personal software on interoperable decentralized private data stores.

We're helping and trying to champion people who are building personal software for themselves, their friends, their communities, and their colleagues…

We help you build that without increasing your dependence on a platform, keeping your information private and close to you. We also guarantee that you can remix, play, and interoperate with your data however you want to in the future.

Why did you build Tonk?

Without giving the whole backstory of Tonk… Tonk's essentially been building in applied cryptography for about two and a half years. For much of that time, we've been like skunkworks, pumping out experiments and prototypes of games filled with ZK, research on MPC, little ZK Nintendo emulators, and social experiences.

We noticed a change in behavior in the world while we were transitioning into a very product-focused startup.

This has happened in the last four months. The behavior change in the world that we've seen is that more people are building more stuff themselves.

We notice this in ourselves, our friends, and the people we meet. A really simple example is that people are building ways for AIs to coach them based on their notes or personal meditation trackers.

Someone we met was building a little tool for calculating the quantity of sourdough starter he'd need to use and how much water and yeast to add to it to bake bread. It's called a bread calculator, and he just built it for himself.

We could see all these people building stuff for themselves. They were doing all this stuff because the world's changed. We can vibe code now because we have AI.

We can build more of our stuff faster. You can create games in the evening. You can develop your to-do manager in about an hour, and it can be more personal and more special for your needs than anything like Notion, Todoist, or Apple Todo could make for you over a year.

But that comes with a new problem. People who want to build personal software want to be able to have more control over the data they’re playing with. And they don't want to be bombarded by surprise fees or feature creep; they want to guarantee that they're going to be able to plug it into whatever they want to plug it into. They want to make sure they're not giving up all of their data; they want to feel independent. They want to feel private and independent and be guaranteed flexibility over time.

We're looking to help simplify the lives of people like that.

If you wanted to do anything relatively ambitious with your personal software, you would need to add all this extra stuff.

When I was at ETHDenver, I was in a restaurant meeting one of our investors, and we had dinner. Then, outside the restaurant, we bumped into this guy − as you do at a crypto conference. He was showing us this little app he'd built that auto-replies to all of his Telegram messages and WhatsApp messages and emails, or at least it drafts replies for him in a nice unified interface. It took him less than a day to build it because he just used Cursor. But he had this problem, which was, « I'm using everything at Postgres. I've gotta handle all these migrations. I've gotta handle all these, like, caching… How does it work across my devices? How do I do the authentication? »

If you want to make life as simple as possible for the person making the bread app, their personal financial planner, or their wedding planner, we need a new way to store information, read information, and synchronize information.

And that's when we were like, « Oh, perfect. This is where all of our insights from building applied cryptography over the last two years are going to come together, and we can build a stack that makes it super nice for everyone building personal software. »

A natural evolution for Tonk

The same motivation for why we're building the Tonk stack now is what got us interested in chasing down our previous ideas.

The thread through the whole thing has always been the same: the internet isn't yet finished. We have this massive thing called platform overhead that imposes costs and inconveniences on businesses and people using the web.

But there's a whole bucket of very exciting and very nascent frontiers: decentralized technologies, including, but not limited to, blockchain, programmable cryptography, and sync engines.

All of these make this nice toolkit and all these problems related to the internet, and they're going to come together in a really exciting way to create a new kind of internet that feels much more, cheaper, and more convenient, but also more liberated for individual human beings and individual businesses.

I never got into blockchain through crypto or money. I became interested in blockchain because I was more generally interested in decentralization on the web. When I was growing up, I never had a real aspiration to get into the tech world. I just wanted to create cool stuff, and I spent most of my time doing theater things.

It was only when I was in my third year of university that I realized that the world was changing a lot, and society's technological basis was driving it. That technological basis is changing so fast, and teenagers are building it in a valley on the other side of the world! I put all the arts to one side and decided to learn to code to get into tech, work with startups, and start my own.

I worked for Epic Games, an amazing company that created Fortnite and has an awesome CEO named Tim Sweeney. Two big things happened at Epic that really shaped my beliefs about the world and innovation.

The first is that my CTO was teaching me all this really great stuff about the history of the web and protocols and how you have these great men and women who kind of… live in the middle of Arizona somewhere. They're members of their local internet engineering task force, meet up, and are basically responsible for the protocols that power the internet.

But it just struck me how there was no HTTP company and no Internet company. It's like a shared protocol that we are using. At the same time, Epic was fighting a battle in the courts against Apple and Google for alleged monopolization of the app stores and the take rates that they were forcing on things like Fortnite.

And so it came together for me, and I realized that the internet is in a very platform-heavy part of the cycle. Still, protocols are a really exciting way of building a more liberated experience that diminishes platform risk and take rates. Not only that, but they also create bigger markets when you reduce individual corporate control of the market and get everyone sharing and building − more of a bazaar than a cathedral.

At the same time, at Epic, I was in charge of a product that would've been great to decentralize if the technology had been available. My job at Epic was to work on the child safety part of the organization. There are a lot of kids in Fortnite!

That imposes a responsibility on the company to ensure that we're being responsible and obeying the law.

There's a really funny part of Silicon Valley, the TV show, where they accidentally start holding kids' data in the US. You get enormous fines under this ruling called COPPA, which means you can pay hundreds of millions of dollars to the FTC unless you're correctly conserving that data.

The best way to keep kids safe when they play Fortnite is to identify the parent and ensure that the parent is in control of the child's experience. That imposed a challenge for us and the team. How do we tell who is or is not a parent, rather than an older sibling or the kid pretending to be a parent?

I was the product manager on the parent verification team, and my job was to distinguish between parents and non-parents.

We didn't want to own the list of all the parents in the world! We knew that Google, Facebook, and Microsoft all had the same problem with their gaming divisions. It would've been so much easier and better to share these lists of parents somehow and not even share them, but not even have custody of them in the first place. This was imposing all this extra cost on us.

That manifested from the inherently centralized nature of identity manipulation on the Internet. I got really into decentralized identifiers and verifiable credentials. Back in 2021-2022, I joined the W3C calls on finding new standards for verifiable credentials. These were basically a proto-zkID.

It was through that that I became interested in ZK as a cutting-edge technology that made decentralized approaches to identity management more feasible. It was through that that I was also introduced to the idea of blockchain, which I had always understood more as a money thing − I'd never really thought of it as a world computer thing.

At that time, I met my co-founder at Entrepreneurs First, the same incubator where Aztec and Gensyn started. We had a meeting of minds and a meeting of the soul as well. We really adore each other and are very mission-aligned.

What we noticed was that there was this really exciting thing called ZK. Still, also programmable cryptography, more general, and all the work was being poured into making the money machine go faster.

Our lead investor estimated that over half a billion dollars has been given to ZK teams. The vast majority of that has been given to research to speed up the protocols and make them more efficient and secure − relatively little money has been given to teams that want to push the application space.

I wanted to think of what this technology could be used for. Coming from my experience at Epic, I just had this intuition that there's going to be something so exciting here. So we built all these games and social experiences like Speakeasy, which is really supposed to be a matching engine that can index your whole digital footprint rather than the limits of any one platform.

And we built onchain games. What was so exciting to us about onchain games and autonomous worlds was the fact that you had a bunch of really creative people building software in a composable way on top of each other. And it was permissionlessly composable in a way that only Ethereum can really deliver.

That meant that the barrier to innovation was extremely low because anyone with a good idea could come up with it. We're inspired by what happened with the curve wars, Nouns DAO, and ConstitutionDAO.

We could see all these new forms of social coordination happening, but we could also see that the story here was so much bigger than that.

After a couple of years of basically accruing technical insights and user insights and getting better at getting a feel for building and launching, I resolved at Christmas time that for the Tonk Project, one of the resolutions for 2025 would be that we would be unbelievably user-focused and obsessed with users and their problems.

That led us to the path we're on today. We find this very niche set of people, and we can see how they need cryptography to do the things that they want to do. We are confident that we're one of the few teams in the world that can deliver that.

We really see it as our wedge to build and achieve this dream of starting a self-sovereign, private, and interoperable network of interoperable data stores that will be the natural place to build any information service on this new Internet.

How is data shared between models and people?

It's like food like someone's throwing a dinner party. You bring a cake, and then you eat the cake together. You could go around to someone's house in the afternoon, and you use some of their eggs to bake your cake, and it's all in your kitchen. And no Mark Zuckerberg is standing over your shoulder and asking you about the cake. I'm sure Mark Zuckerberg doesn't want to look over your shoulder, either. I'm sure that if he had better technology available to him, he wouldn't want to be there either.

So, I try to dream of the future. I think it's there. There are all sorts of architectural questions around: Is this going to be like Urbit? Is this going to be like BlueSky? Is this going to be like Ethereum? Is this going to be like some superpower database or a selection of databases?

Any application will be able to read data from any other application if the user permits it, and it should be as simple as just calling, like user.address or user.favourite_film. The user will simply have their data. The data will be a relatively unified blob that follows the user around the internet and feels like their own, and services and apps will ephemerally spin up.

They come in, and they play with your data if you let them. Then they leave, and your data stays where it is. So, you get a decoupling of applications and services from the underlying information that they rely on. And that is good for everyone. That's the dream. And there are just so many smart people building in that direction, and there are so many people who recognize the value of that new way of building.

And yeah, that's what we're so excited to be chasing.

What makes this possible today?

People have this feeling of, « Look. We've heard it all before. You know, you're going to tell us it's all going to be open and interoperable. But, you know, I was there at the start with Urbit, and now I'm jaded, and I'm unhappy. And stop giving me hope, man. You're going to, you're going to lead me down a toxic path. »

It is genuinely different this time. I understand the response, but it is different this time for two reasons. The first reason is that technology has changed, and the second reason is that the market has changed, too.

So, how has technology changed? It's changed in two kinds of ways. The first is that because everyone can vibe code now, this dream of malleable software, which has been around since the eighties and nineties, even back to Xerox PARC, decades and decades and decades ago, and Alan Kay, even like Apple HyperCard, Squarespace, and Microsoft Excel to an extent…

These are all basically driving at the same thing: how do you make it possible for regular people to build their computational environment rather than having to buy the one that's given to them from the supermarket? And AI changes that. English is the hottest new programming language.

And now anyone… well, now, people like me who are terrible at coding can build personal software, but in the next year, if AI keeps advancing in the way that it does, my dad will be able to do the same. He’ll tell his phone, « Look, I need you to make me a version of WhatsApp that integrates with my personal trainer's calendar and sends in my blood pressure on like a weekly basis. Can you do that, please? » and the phone will go « yes ».

The second thing is that this dream of Urbit, AT protocol, Ethereum, these decentralized protocols, the technology has just gotten a lot, a lot better at actually making them work at scale and solving certain scaling problems.

Because of the crypto industry, so much of this money has been poured into cryptography, and now you can literally do things that you weren't able to do before. One of the problems with Urbit is that it's not really multiplayer. You have all these different islands and all these different planets, but they don't really speak to each other.

And if you want to use a service that will tell you how to surf around the different worlds of Urbit, that service will have to be in a Web2, like, like platform. And that's changed now. There are ways to create virtual servers, such as experiences between different servers, without actually involving a third party with things like multi-party computation.

The technology has changed a lot. It's fueling the demand and the supply of protocols that can meet that demand.

The second thing is that the internet has changed. People are a thousand times more literate than they were 20 years ago about conscientiously choosing what information services they use.

So, in a world where both BlueSky and Truth Social usage is surging, the world's changed, and people are just waking up to the need to keep their digital worlds more indie and private. Who knows exactly what it's going to look like?

In my opinion, vibe coding is really significant because it's a platform shift. Every time there's a platform shift, everyone's built-up advantage goes away. I'm skeptical that Google has that much more of an advantage over Tonk in building in this new world because, relative to this new world, no one has an advantage.

It's like space. It's like the moon. No one knows what's going to happen, and we're going to see the weirdest stuff over the next few years, and that means it's a chance to hit the reset button. And so I think when it comes to how people create educational content, how people curate their info sphere and send funny videos of cats to their friends when it comes to how enterprises build, build businesses on top of like their data, on top of other people's data, when it comes to how individuals manage their health and their projects and their to-do planners.…

When it comes to creating tools, games, and events for our friends, I think everyone will have the tools of creation, which just opens up this question of where the state is going to live.

There are three outcomes.

Either the state will exist in individual buckets that don’t talk to each other. This is bad because the state is only valuable when it can contextualize itself relative to other states, where my dad's gym app can speak to his blood pressure app and so on.

The second outcome is that we all live under the sovereignty of Google Gemini, and everything in your Google Drive can speak to each other seamlessly. But then what happens when your Apple ecosystem needs to talk to your Google ecosystem, needs to talk to your Notion ecosystem, is to talk to your little indie health app?

It reminds me of the 1990s, when AOL, Yahoo, and Microsoft were like, « Ooh, the internet seems big. Maybe we could own it. » But it doesn't work. The public network effect is stronger than the private network effect, or at least that's the bet that we are taking.

We bet that there will be a maximally interoperable way for any data living anywhere to be fruitful and productive with any other bit of data and any other service coming in from anywhere without caring what ecosystem you are from… and that's going to have an unbelievably strong network effect.

While we're focused today on this niche of cyberpunks and indie devs, I think it is not going to be very long before the pure value of all the data intermingling becomes so obvious that it becomes the natural home in which to build your business, your application, and your internet space.

And this time is really different. That's why Tim Berners-Lee started posting again about this project for interoperable data stores that he’s had since forever: Solid. He started posting in the last few months because he can see the same thing that we can see. He's bringing it back because the world has changed. That's why it's so exciting and why I'm optimistic about the future.

Building the Internet that people used to dream

I was lying on the sofa on Sunday, exhausted. When I opened my phone, I saw that some dude in the Rockies had built an app with Tonk. We open-sourced our library last week, and people we don't know are already building with it.

He recorded a seven-minute video of himself using it and giving his feedback. He then showed us the little tool he used for watering his plants and sent it around to all of his buddies. People are already desperately trying to find this stuff.

This is us.

We had someone drop into the group chat yesterday, and they built a to-do planner. They were like, « How could I use this to share my photos with other people without going through iCloud? »

And so if people stop saying they want it, we'll go and do something else.

But there was this time in the 2010s when people were coming up with all sorts of bonkers ideas. What if we had Uber for pets? You know, what if we had Spotify, but it was like for pictures of the horizon taken from your bedroom window?

And we had this long, long tail of bad ideas that only mattered to a small number of people. And if you applied to Y Combinator, they'd be like, « There's no market for this. What are you talking about? »

There's this long tail of all the problems that were too trivial for the Google Maps product manager to prioritize on the Google Maps product roadmap, and the cost, as the cost of software, is just going to zero. It will continue to do so over the next year.

All the dumb stuff that was too dumb to build in the last 15 years? We're going to see it all. I'm so excited, and we've got a real chance. At Tonk, we're building a stack that makes it unbelievably easy to build little niche fun things for their friends, their families, their colleagues, their communities, and whoever.

What’s holding back innovation?

Fear is holding back innovation.

Every developer I've spoken to in the last couple of months can, especially in crypto or who's been building in kind of ZK or crypto-related stuff; they're all sad. They're all sad because they feel like crypto used to be like a big party. Three or four years ago, DeFi summer, Composable, you know, Nouns, all this random stuff was happening. But now it feels like the party's in AI. Now, I have to build casino meme coins, but I don't want to build casino meme coins. I want to build cool stuff.

No one is an expert in anything. There is no such thing as a sunk cost in your expertise or the skills that you've built up, like in robots, artificial intelligence, face, 3D printing, and decentralized networks.

There… there's so much. It's like a green field. Possibility. And although it can be easy to think that some people are further ahead, some people are further behind relative to the size of the space; we're all stupid. And I mean, all of us, like literally every single one of us, BlackRock, Sam Altman, Sundar… we're all beginners in this brave new world, and I think we all have a chance to create awesome things that solve problems for regular human beings.

We have a chance to do Web3 in a cool way that is rooted in human beings rather than imaginary problems.

A tangent about London

Being based in London is pure past dependency. I met my co-founder here.

There was a moment when Bill Gates was thinking, where do I build this office for this thing called Microsoft? And he knew that Silicon Valley, which had been going for decades with Intel, HP… He knew that was the obvious choice to go pitch up in San Francisco or Palo Alto. And he consciously decided not to do that. I think he chose somewhere else first, and then he decided to go to Seattle. Jeff Bezos saw the same thing a few decades later. That's why he didn't want to build Amazon in San Francisco or the Bay, either.

This is really important for crypto. After a while, the social innovation ecosystem can constrain your imagination and productivity for building products that users love.

Living in London has a strategic advantage: It's a very international city, the most similar place to New York in the world, and that gives us all the stimulus and access we need. But it also just lets us focus on what matters, which is the user, rather than, « What's the dinner party I'm going to? Have I been invited to Sequoia's offices this week? Was I seen with the right person in the right place? »

I don't care. I care about the user and the technology that improves their lives.

Tonk and Trust Infrastructure

Trust infrastructure is amorphous. It can mean lots of different things to different people. It predates what I wrote and what Arnaud wrote about it.

The best way to explain it is that it's a set of ideas that really try to do two things.

They try to show how much advanced decentralization technology that has emerged from crypto and other places over the last five years can improve human social interaction and coordination worldwide.

Imagine there's a big bucket of all the problems in the world. You cut out all the issues that markets and governments are really good at solving and take what's left. These are the things that neither governments nor markets are good at solving, and they are things to do with coordination.

They are moving very fast, and the governments are too big to catch up with. Whether it's coordinating over climate change and geopolitics or coordinating over, like, how do we make sure AI doesn't screw over like this school in this neighborhood of like West London… this is where there's an opportunity to build technology that improves how human beings get stuff done communally.

That's the first thing I would say about trust infrastructure. It's a new vision, like a network state vision. What coordination machines have we used to date, and what machines will we use in the future?

The second thing trust infrastructure is about is reminding crypto people that often in life, what matters is not creating completely trustless ways of interacting with each other. It makes it easier for small groups to come together, spin up, coordinate, and then spin out again.

It's a really compelling vision of our future. Most of these things are true in theory. The great challenge for innovators is to see how they come true in practice.

And so, yeah, that's what I'm focused on with Tonk at the moment. How do we know if we're moving in the wrong direction, and how fast can we figure that out?

Innovation is really hard. It's not supposed to be easy, but it's really fun, and it's a privilege to be creative with the things that constitute our world.

How to get involved with Tonk

The Tonk community is a Telegram group chat that has grown significantly over the last week and a half. At some point, I'm going to have to cap it because it's going to grow beyond Dunbar's number. So get involved now because there are really great people there who you can coordinate with and learn about what we're building.

We respond to feedback on the tool within 24 hours, including code support. We're in that really exciting moment, a product development way. You can be right in there and shape what it becomes as a tool.

There are two ways to use Tonk today.

If you're building a new app, we have a CLI tool that will spin up new Tonk apps for you. It's like creating a new app or a React app.

The second case is if you already have an application that you want to make more private, independent, flexible, and interoperable. In that case, we've got a way of using the Tonk framework to manage all of your data, state, and application lifecycle. This will massively improve your dev velocity as well because using Tonk to manage your state removes all the problems related to caching, migrations, and auth that you are used to from traditional development.

So yes, take a look at the Tonk library on GitHub. Use it to build stuff.

If you have any feedback, I have an open thing on the website where you can book a meeting directly with me. We'll see how long that stays up.

Get involved. It's just such an exciting time to build cool stuff.