Startupland is full of venture capitalists and founders trying to figure out what next year is about, but there’s another kind of innovator who longs for the 1990s. He draws his inspiration not from to-the-minute forecasting, but from the original HTTP protocol, written when the potential of hypertext was untold.
Balaji S. Srinivasan is seated at the center of a long, shiny table, the kind he might be accustomed to as a Board Partner at Andreessen Horowitz. On this day, he’s on home turf at the offices of 21 Inc., where he serves as CEO. Comfortable in a heather-grey Champion sweatshirt, he’ll occasionally break eye contact to pull something up on his web browser and beam it onto the wall for illustrative purposes.
“There’s literally an error code in the HTTP Protocol that goes all the way back to the early nineties,” he says, loading the protocol spec on w3.org. It’s a stark, barren, HTML-era webpage. He scrolls down it until landing on Error Code 402: “It’s called, ‘Payment required.’”
Unlike the other client error codes, there’s no description of a hypertextual function beneath 402-Payment Required. Instead, there’s an intriguing note: This code is reserved for future use.
Bitcoin prophets interpret the existence of 402 as proof that the internet will become the future of payment. Somewhere along the way, the internet just forgot to honor this promise. Now, 21 is fulfilling it with open-source software.
“Our mission is to make micropayments a reality,” Balaji says. “This error code, ‘Payment Required?’ We’ve made that a reality, where you can go into your code and type, ‘Payment required: 5,000 Satoshis,’ and actually start requiring micropayments in your product.”
The end result is a network of machines that can demand and supply payments autonomously. 21 refers to this infrastructure as the machine-payable web. The implications of it will be huge for developers and consumers alike. And if any of this sounds a little sci-fi, that part is intentional, too.
“The page says this error code is reserved for future use,” Balaji says with a grin. “Well, we’re here to tell you the future is now.”
The machine-payable web
21 subdivides the internet into three distinct categories, two of which you’re familiar with already. The World Wide Web is a web of documents with hyperlinks to connect them. The social web is a network on which individuals are nodes, and where friendships, likes, and tweets act as the links between them.
And then there’s the machine-payable web, a network of machines bridged together by micropayments.
“One of the fundamental things that digital currency enables is for a machine to actually possess money,” he says. “That’s a very deep concept, but the best way to explain it is, if you hold a dollar bill, that is a sufficient and necessary condition for you to have a dollar. With a machine, if it has Bitcoin private keys, which is literally a string of letters and numbers on the device, it actually possesses the money.”
He’s quick to distinguish this from online repositories of money, the fintech logos you know. A computer with a PayPal API key isn’t technically a machine that possesses money. Ultimately, it’s linked to a human or corporate bank account, meaning that third parties can block the machine’s access to funds. Money that can only be accessed through a PayPal API key is subject to a whole host of constraints that a Bitcoin private key just isn’t.
Conversely, a machine can fulfill certain functions independently. Why machines need to enjoy laissez-faire freedom might be lost on some—it might even seem post-human—but 21 sees it as the answer to many of futurism’s hardest problems. Wired’s been theorizing about the Internet of Things for years that developers could’ve spent building it.
Part of the holdup for the Internet of Things may have been infrastructure like the machine-payable web. To prove this point, Balaji cites the example of self-driving cars.
“If a machine has a Bitcoin private key, it can send and receive money, which is one of several prerequisites for autonomous machine-to-machine payments,” Balaji says. “I came up with an example a while ago in which a self-driving car could autonomously pay cars to pass them.
“Andreas Antonopoulos improved on this with the concept of a self-driving car that can earn bitcoin and then go and refuel itself,” he continues. “It picks people up to earn bitcoin. Then it drives out to the gas station and gets some fuel. If it has an issue, its sensors say, ‘This part is broken,’ and then it drives into the mechanic and says, ‘Please repair me.’ It’s a closed loop, right?”
The PayPals of the world can’t program autonomy this way—nor are they really trying to. They’re still more focused on getting into people’s wallets, instead of giving machines wallets. To actually build an Internet of Things ecosystem, an underlying web of micropayments may be necessary.
But the physical world of self-driving cars is very hard to conquer. In the near term, 21 sees the machine-payable web mostly disrupting the virtual realm, beginning with the cloud.
Unto the grid
As a concept, cloud computing is by now familiar to the masses. Your grandma backs everything up on iCloud. Your business uses AWS. “The cloud” is just Mountain View marketing speak for a third-party data center that, in exchange for dollars, hosts your files remotely, instead of locally. The result of this is that your files are accessible anywhere from a centralized datacenter.
In contrast to the cloud, “grid computing” distributes resources across many machines outside of datacenters. And the flow of payment in grid computing is not unidirectional, but bidirectional: your computer can also earn money in addition to paying other computers for their time.
“The grid refers not to the full-time machines that are always running in data centers, but to the part-time network of weekend warriors—all the machines that are on laptops, phones, in your pockets, et cetera,” Balaji says. “On the grid, payments are bidirectional. Not only do you pay out to the grid, the grid pays you. Your computer may rent time from other computers, and in turn, it may also sell its own computer resources when it’s idle.”
The ideas behind grid computing have been around for a while, but the most visible example of them is Folding@Home. Folding@Home is a project that outsources power from idle personal devices worldwide to create a supercomputer that can simulate protein-folding. Its mission is nothing less than finding cures to diseases like cancer, Alzheimer’s, and more.
Volunteers lend their computer time to Folding@Home by running a screensaver, and earn status points on a leaderboard for each compute cycle they donate. Even though Folding@Home’s contributors donate computer time rather than sell it, the project has drawn attention to the untapped power of the grid. If users were incentivized with money rather than just status, perhaps much more collective computing power could be mobilized.
“Grid computing is about making use of all of these devices that are distributed through the real world, which most of the time are idle,” he says. “They have lots to offer, but their users have no incentive to offer that until micropayments enter.”
To that end, 21 has been building sample applications that flex the power of the grid. The simplest such application is Ping21. Ping21 allows developers to rent other computers through micropayments and collect a more accurate assessment of their own site’s uptime and latency. Sensor21 does something similar for environmental data: by paying Bitcoin to a distributed network of machines, you can collect and aggregate each machine’s individual data on barometric pressure, temperature, pollution, radiation, and measurements of this nature.
“You can pay a bunch of part-time machines and script them to ping a central server, or alternatively to give you access to their barometric data,” Balaji says. “One of the long-term interesting things with this technology is that you can now use digital currency to incentivize machines in a distributed fashion. A machine outside of a datacenter, localized in the real world, can now collect valuable sensor data and sell that to other machines.”
APIs are eating the world
For the futurists, the machine payable web makes a self-owning, self-driving car possible. For people interested in sensor networks, it’s a whole new way to incentivize distributed data collection.
But who else is the machine-payable web for?
“The machine-payable web is a toy for developers to experiment with cool new things that would be very hard to do otherwise,” Balaji says. “Chris Dixon has a saying: ‘The next big thing always starts as a toy.’”
To the consumer-facing part of the internet, hyperlinks and retweets still seem like the architecture of the internet. But if 21’s right, you won’t so much see as feel the machine-payable web emerge. And that could start soon.
A big reason is that APIs are eating the world (with apologies to Marc Andreessen). Or at least, they’re eating the software that’s eating the world. Every app wants to be the most quickly developed, the most feature-rich. To achieve that, it’s become increasingly common to augment your native code with paid remote API calls to machines running external services.
Paying for these API calls is complicated, though. For content consumed by a human (e.g. a newspaper article), the content creators have the option of selling it through either an ad-supported model or a paywall-based model.
In the ad-supported model, every surfer can view the content by spending their time looking at ads, without pulling out a credit card, doing what is essentially a micropayment in the form of their time.
By contrast, in the paywall-based model, surfers must pull out their credit card before viewing any content at all. While hybrids exist, these are two of the most common options.
For content consumed by a machine (like an API call), however, the ad-supported model is no longer an option. Machines don’t get distracted by ads in the middle of an API call, meaning they’ll never pay with their time by clicking on an ad, as a human would. Therefore, the only way to pay for API calls today is with a paywall. But that means a developer has to sign up for a service and enter in a human credit card before accessing it.
This is a problem, for the same reason that paywalls on the human-consumed internet are problems. Just like a signup form deters you from viewing a news website, so too a new API imposes a considerable fixed cost of developer time.
But Bitcoin micropayments might solve this. They would allow machines to directly pay other machines for access to their resources, without reference to any out-of-band humans, or credit cards, or billing relationships. Just the direct exchange of digital cash for digital services.
In this machine-payable world, the machine calling an API would directly remunerate the machine serving the API through micropayments attached to every call. This isn’t the grand fulfillment of some Skynet dream; it’s an everyday convenience for developers.
“Bitcoin in particular is good for payments that are very small, very large, very fast, very automated, or very international,” he says. “If you’ve got, let’s say, two or three out of those five, you’ve got a use case that does not fit the current financial system, but which fits Bitcoin.”
The machine-payable web also doesn’t mean the full-scale decentralization of companies. Contrary to the developers behind Ethereum (which Balaji compliments), 21 is more focused on applications that combine centralized business logic with decentralized digital currency.
“Ethereum is awesome, but it’s often challenging to fully decentralize application code,” he says. “For one thing, it becomes very difficult to update if you’ve written it to an immutable blockchain. Our basic philosophy is that for a while, to get useful applications with digital currency, you’re going to want to centralize everything other than the payment network itself.”
As Balaji puts it, “It required Satoshi-level game theory to get a decentralized wire transfer right, namely Bitcoin. How much harder will it be to decentralize a corporate charter with Ethereum? There are smart guys working on it, and it’ll happen, but it’ll take time. In the meantime, the 21 philosophy is to make Bitcoin a first class citizen in a Linux environment and integrate it with existing/standard tools, like Python, HTTP, apt-get, and the command line.”
Reserved for future use
Internet-scale payments may still be in their infancy. This seems counterintuitive, because Web 2.0 has done nothing if not flood the market with fintech companies. But to validate this claim, he launches into a thought experiment.
“There’s this great tweet by Vala Afshar, where if you rank the nations of the world, Facebook is number one with 1.7 billion “citizens”. Then you see China come in second with 1.3 billion citizens,” he says, jumping several tabs over to Twitter.
In the aforementioned tweet, Afshar has spliced the digital citizens of social networks with the actual citizens of nation states to create a fascinating sort of mashup: one half startup user base, one half national population.
“So, the first column of this table is population, but let’s talk about the second column, revenue,” he says.“Facebook only makes one-to-ten dollars, per person, per year, while China makes something more on the order of 1,000 dollars in tax revenue. The US is more on the order of 10,000 dollars per person, per year.”
It’s confusing at first, to hear GDPs spliced so loosely with companies’ returns, but as he continues along this line, his point comes into focus.
“So, literally, is it is possible to go up ten, a hundred, maybe even a thousand fold, in terms of monetization per person, per year,” he says.“That’s why I believe that internet monetization is just getting started. It starts with machines selling their time for digital currency with the machine-payable web. Eventually, it gets to humans selling their time for digital currency as well. And I think then we can then start doing more economically impactful things in these social networks besides just poking, liking, and tweeting at each other.”
His phone buzzes for the second time in five minutes. “That’s a much, much longer conversation, but I have to take this.”