SmallTime: Replacing Bitcoin with a currency-free distributed transfer system
TL;DR: Bitcoin needs a currency because it relies on miners. Remove the miners, and the remaining system can be used to transfer almost anything for free. I propose a system that, if it works, is the next step for payment and transfer networks.
The first half of this post gives context: a simplified understanding of how cashless payments work; a spectrum to understand how cashless payment / transfer systems are changing.
The second half explains SmallTime, a proposed distributed transfer network. A more technical explanation of SmallTime can be found here.
Throughout, I have tried to avoid jargon and used as many illustrative examples as I thought helpful. Please direct all thoughts, comments, and criticism to @dbrgndl
All drawings courtesy of the fantabulous @bettinauxd, inspired by the one and only Randall Munroe
Without trusted third parties, cashless transfers are really complicated
Imagine that you have two people sitting in a room. Alvin owes Bartholomew money. The simplest solution is that Alvin pays Bartholomew the exact amount he owes in cash. Bartholomew performs some rudimentary checks on the money. He counts it, if it’s a high denomination maybe he inspects it to see that it isn’t forged. Without knowing anything about Alvin, Bartholomew can most likely accept the money and go on with his life.
In a cash-less society, this simple exchange becomes much more complicated. We don’t see it, though, because that complication hides behind a wall of trust. Let’s say that Alvin PayPals Bartholomew the money. Bartholomew waits until he receives a confirmation email from PayPal. Because he trusts PayPal, he accepts that Alvin no longer owes him money, and he goes about his life. This seems simple, but only because Bartholomew trusts PayPal to have taken some important steps on his behalf.
What steps does PayPal perform for us? Many of the steps PayPal and other trusted third parties take on our behalf are unseen. We can identify them by imagining a non-cash transfer without PayPal in the middle. The steps each party needs to take without PayPal is the service that PayPal provides us as our trusted third party. These are the same unseen services provided by banks, card companies, and other trusted third parties.
So, what does a non-cash transfer without a trusted third party look like? As before, Alvin would email saying that he had paid Bartholomew the money he owed. Lacking banks (trusted third parties), Alvin and Bartholomew track their money in excel spreadsheets. Alvin’s email says “I’ve reduced my cash by the amount I owe you, so you can increase your cash by the amount I’ve sent”. Bartholomew wants to verify this and asks to see Alvin’s excel spreadsheet. Sure enough, Alvin has decreased his cash balance by the correct amount. Alvin’s cash balance has fallen, but the money can’t have just disappeared. By increasing Bartholomew’s cash balance, the money has transferred to Bartholomew. Yes, that’s how cash-less transfers work.
But there’s a problem: it turns out that Alvin owes a lot of people the same amount of money. What’s to stop him from showing that same decrease in cash to Bartholomew and Cathy and Doris? Alvin is able to turn one transfer into many by convincing each of them that this decrease is Alvin paying them, not someone else. He is able to “triple spend” his money.
When we’re not dealing with physical cash, trusted third parties like PayPal stop this from happening. They run spreadsheets on behalf of Alvin, Bartholomew, Cathy, and Doris. PayPal keeps the spreadsheets synchronized, and makes sure that changes are properly sequenced. That way, Alvin can’t trick Bartholomew and co. into thinking a single transfer benefited each of them. Once Alvin has paid Bartholomew, Cathy only sees his newly-reduced account balance. This means that Alvin can’t “spend” the money again. Once you digitize money, all transfers are just increases and decreases in spreadsheets. It’s the job of these third parties to make sure that those spreadsheets stay synchronized. We just trust that they’re doing it well.
Trust and spreadsheet synchronicity: a mental model for payment systems
The way we pay for things is changing. You can think of these changes in terms of these two factors: spreadsheet synchronicity and trust.
In fact, these changes only happen to be about payments (and money) for now. The cells in a spreadsheet can hold many things: currency, bonds, the deed to your house, the right to play certain media, or even proof that you came up with an idea on a certain date. But let’s put that aside for now. It’s easiest to start thinking about these changes as changes in the way we transfer money. We can map these changes to a spectrum.
The spreadsheet monolith
Imagine if there was one global organization that dealt with all payments. They run one spreadsheet that includes the money of every person, organization, and government. That would be at the far left hand side of our spectrum (see later illustration). In some ways this would be awesome: money would move as quickly and easily as you can make changes to a spreadsheet. But, you would have to trust that entity to get things right. To not get hacked. That entity would wield an unimaginable power over the global economy.
The Federal Reserve banks
Let’s step into reality. To the right of that imagined spreadsheet monolith we have something like the Federal Reserve banks. Together, these banks hold an incredibly large amount of money. They electronically transfer money back and forth on behalf of their member banks. The system they have is a little more complicated than a single spreadsheet. There are as many spreadsheets as there are Federal Reserve banks. But there aren’t that many Fed banks, and they all trust each other a great deal. It isn’t all that complicated for the spreadsheets to stay synchronized. The vast majority of us trust the Fed banks to do this job well.
Traditional trusted third parties
To the right of the Fed, you have what I’ll call “traditional trusted third parties”. These are the banks and payment networks (card companies like VISA, Amex, etc., PayPal and Venmo). Like the Fed, each of these organizations has their own spreadsheet, but this time you start to appear on them. If you bank with Wells Fargo, somewhere your name appears on their spreadsheet. There are many more traditional trusted third parties than there are Fed branches. But, like the Fed, each of these organizations trusts the others to do a good job updating their spreadsheets. This trust is vital. During the Financial Crisis, banks questioned the value of assets held on each other’s spread [balance]sheets and consequently demanded more collateral or stopped lending to each other. The terrible ramifications of sudden distrust between banks are nothing new: Walter Bagehot wrote about them in 1873. Card companies are similarly mistrustful, relying exclusively on banks to issue their cards. Finally, while we all trust these third parties to keep in sync with each other, we still check, occasionally, that an important transfer went through correctly. Listeners to Startup will remember the wire transfer accident that left Chris Sacca’s money in an account in Eastern Europe.
Ripple
So now we leave the realm of traditional services and, moving right, find something like Ripple. Ripple is a protocol run by a network of organizations, including banks and non-banks. These organizations all know each other, and know how much they trust each other. The network is still in its infancy, but it’s growing. Someday, it will perhaps include more organizations than there are traditional trusted third parties.
If we compare the spreadsheet of one bank to another bank’s, they would be completely different. Each bank has their own customers. With Ripple, though, each trusted third party has a spreadsheet for the entire network. Each Ripple trusted third party’s spreadsheets will be identical if they are properly synchronized. This begs the question: how does Ripple keep the spreadsheets synchronized? Remember, there are many more Ripple trusted third parties than there are banks. If their spreadsheets aren’t synchronized, Alvin can double spend. To solve this problem, Ripple takes advantage of the fact that all the trusted third parties are known. Ripple trusted third parties vote to confirm that they have seen transfers as they are received. The synchronized spreadsheet only includes transfers with enough votes. This way, Ripple is able to have a much larger number of third parties, with less trust, and still keep the spreadsheets synchronized.
Bitcoin
Finally, we arrive at Bitcoin. Like Ripple, Bitcoin is a protocol run by a large number of organizations and individuals. The Bitcoin network is even larger than the Ripple network. Unlike Ripple, participants in the Bitcoin network do not have to know each other. This makes trust between participants impossible (or unnecessary). Instead, participants trust the Bitcoin protocol. When it comes to trust between participants, Bitcoin is a “trust-less” system. Like Ripple, Bitcoin relies on a subset of all participants to maintain and synchronize spreadsheets. Like Ripple, each spreadsheet is a spreadsheet for the entire network. The entities that maintain these spreadsheets are “miners”. The spreadsheet itself is “the blockchain”. Miners use the Bitcoin protocol to keep the blockchain synchronized across the whole network. This prevents Alvin from double spending. For an excellent deeper explanation of Bitcoin, see Ilya Grigorik’s post.
A side note: Bitcoin terms are often misleading. The bitcoin itself is not a coin or digital coin, but instead is an entry in a spreadsheet which associates an account value (e.g., 3 bitcoins) with an account owner. “Mining” sounds like you are digging for these coins. You aren’t. I’ll explain mining shortly, but it’s important to know that what you find when mining isn’t bitcoin and doesn’t have value (it’s not like gold mining). Instead, it is the Bitcoin network that compensates you for mining. Finally, the blockchain sounds like a bunch of interlinking blocks. It’s easier to think of as all the previous versions of the global spreadsheet, from the very beginning of Bitcoin until today, chained together. In this way, the blockchain is a record of all the transfers ever made with the Bitcoin protocol. Back to the post.
In Ripple, all the trusted third parties are known. This allows Ripple to rely on a voting system where each party gets a vote. Ripple uses these votes to decide which transfers are included on the global spreadsheet. But what happens if you don’t know the voting parties? What happens if you don’t even know how many parties are going to vote? You can’t wait for a 6/10 majority if you don’t know whether there will be 10 or 100 votes in total. I could cheat by voting multiple times. Or, I could cheat by having my friends join the network and vote. The Bitcoin protocol solves this problem using a “proof of work” system.
The Proof of Work (PoW) system asks miners to find a very specific number by repeatedly solving a math problem. It’s like hunting for a needle in an exceptionally large haystack. Imagine that you’re a Bitcoin miner. You need to find a needle in a haystack. You go and buy a needle-finding machine. You have 50 watts to run the machine. At 50 watts, the machine runs at full speed. Add another machine, and both machines will run at half speed. In this way, you can’t trick the Proof of Work by having your friends vote with you (adding machines). So you cheat by buying the fastest needle counting machine and the most electricity. The PoW system responds by increasing the size of the haystack. The system knows the probability of finding the needle in a haystack of a given size. This means that it also knows how long it takes to find that needle. The Bitcoin protocol adjusts the Proof of Work difficulty so that it takes 10 minutes on average, no matter what technology you use to find needles.
Miners collect transfers from the network, package them, and then race to find the needle first. The winner publishes their package of transfers as the newest version of the blockchain. All network participants download this new version, and the process repeats. Should two miners “win” in close succession, whichever version most participants download will eventually become the version used by the whole network. In this way, participants have “voted” on which version of the spreadsheet they will all use. By combining Proof of Work and this “vote-by-downloading” system, Bitcoin ensures that all spreadsheets are synchronized in ~10 minute intervals. It does so across massive networks of unknown participants. This is an incredible achievement.
Tying it together
Above, I’ve described a spectrum of transfer solutions. Moving from left to right, you have more participants. You have more spreadsheets which need synchronizing between these participants. Known participants are replaced by unknown participants. Trusting participants is replaced by not needing to trust participants. Instead, the concept of trusting a protocol emerges.
I’ve not mentioned one aspect: cost. The hypothetical Spreadsheet Monolith is so important to the system that it can charge whatever it wants for a transfer. But as the number of trusted third parties in a network grows, these transfer costs fall. Ripple is less expensive than traditional trusted third parties. Bitcoin is less expensive than Ripple.
Transfer costs are harmful because they slow the velocity of money. If a transfer costs $0.20, I am less likely to send $2 than I am $200. For example, if I am collecting $2 payments, I will wait until I have 100 before I send them to my bank. This means that $198 will sit uninvested until I can send $200 to the bank. The lower the transfer cost, the closer to real-time payments become.
Transfer costs are also important because they are exclusionary. If I am only capable of sending small amounts, I am less likely to use the transfer system at all. This is especially important when considering the over $400 billion in annual remittances. That’s $400 billion that workers in developed nations try to send their families in less-developed nations. Imagine what $0.20 costs you when you live on less than $2.00 a day. In 2005, that was ~40% of people.
While Bitcoin is less expensive than other trusted third parties, it still has transfer costs. The miners are willing to run their needle-finding machines because they get paid by the network to do so in two ways: 1. They are currently paid in procedurally issued bitcoins. 2. They can claim the fees paid by all the transfers they amass and add to the blockchain. I don’t want to get into the specifics of how fees work, but suffice to say: 1. The protocol/network automatically rewards miners with new bitcoins when they win the Proof of Work. This process will end in ~2024, when almost all new bitcoin have been mined. This date is arbitrary and reflects early monetary policy choices made by the Bitcoin community. 2. As newly-issued bitcoins taper-off and, in 2024, stop, transfer fees, also automatically paid, will increase. No one is quite sure how much they will increase, but they will rise. Although cheaper than other trusted third parties, to the people that need cheap transfers most, it could be very expensive.
I am thus compelled to ask: what sits to the right of Bitcoin?
Right of Bitcoin
I have already laid out the left-to-right trends of this spectrum. We can extend these trendlines further to guess at what lies “right of Bitcoin.” It will be less expensive, it will have more trusted third parties (they will be unknown), it will be trust-less, it will have to synchronize large numbers of spreadsheets. This all sounds good, but raises questions:
- How can it be less expensive than Bitcoin?
- How can it have more trusted third parties (miners) than Bitcoin?
- How will it synchronize among those parties?
To make a system less expensive, more inclusive, and still robust in my mind can only occur in one way:
Get rid of miners.
If on the spectrum’s left you have the Spreadsheet Monolith, on the right must be a complete spreadsheet democracy. Everyone has the global spreadsheet. Everyone synchronizes the spreadsheet. There are no “trusted third parties” at all. The protocol is the only thing you need to trust.
Three things happen when you get rid of miners:
- There are no network-enforced transfer costs, as there are no miners to charge them.
- The network is no longer limited to using an inherent currency (like bitcoin).
- The network becomes more robust to attack.
This second and third points bear explanation. While Ripple allows transfers of any assets, Bitcoin runs on the bitcoin currency. It needs an inherent currency because it needs to pay its miners. Without the miners, no one synchronizes the spreadsheet and Alvin can double-spend. The Bitcoin protocol can’t issue dollars (it’s not the Federal Reserve), so it needs a currency of its own. That currency is the bitcoin.
Removing miners removes the need for an inherent currency. The transfer system runs on distributed spreadsheets. That means that it can transfer anything that fits into a spreadsheet. As I said, that is anything from a currency, to digital asset rights, to messages. This network would no longer be about bitcoin. It would be a distributed, trust-less transfer network. Currency transfer networks like bitcoin are tightly regulated. Other transfer networks — like the internet itself — are not regulated. Because this transfer network does not have an inherent currency, it would not itself need to be regulated. Only currency transfers over this network would be regulated, which is how payments over the internet are regulated today. As a whole, this network would bear less of a regulatory burden.
It would also be more robust. Currently, bitcoin depends on having at least 2/3rds of its miners be “honest”. “Honest” means that they do not allow double-spending or ignore paid transfer requests. If there are many miners, it is difficult for 2/3rds of them to band together and be dishonest. If there are few miners, it would be easy for them to collude. Mining takes two things: electricity and fast needle-finding machines. You can always buy faster needle-finding machines, but the cost of electricity is hard to compete on. As a result, mining has already become very centralized around places with lower electricity costs. 55% of the network was briefly controlled by one group of miners in 2014. In the long term, economies of scale will drive mining to a few miners with very low energy costs. Collusion will be easier. Getting rid of miners would remove this problem.
Who replaces the miners?
Miners are crucial to the Bitcoin network. They spend electricity to run their needle-finders (also called “computers” or more likely, “mining rigs”). Their computers perform the Proof of Work. This synchronizes the global spreadsheet for all the new transfers that have taken place.
Let’s look at that process again:
Electricity -> computer
Computer -> Proof of Work
Proof of Work -> updates blockchain
Updated blockchain -> transfer
Electricity -> computer -> transfer
Who has a computer and would spend electricity to allow transfers to take place? That’s right: anyone looking to make a transfer themselves.
Miners are replaced by everyone. Everyone is a miner, so no one is a miner.
This sounds crazy, but stick with me. All network participants have four things in common:
- They have electricity
- They have a computer (even if it’s a smartphone)
- They have a network connection
- They want to make a transfer
(1), (2), and (3) are identical to what miners need in order to mine. In other words, (1), (2), and (3) together allow someone to do the Proof of Work. The Proof of Work allows transfers to occur. Taking (4) into account and you have someone who can do the Proof of Work in exchange for being able to make a transfer. This is true for every network participant. This means that between two network participants, you have a coincidence of needs. Each has a transfer that needs to go through. For that to happen, each needs another person to complete the Proof of Work. This coincidence of needs allows them to barter with each other. Alvin does the Proof of Work for Bartholomew’s transfer, so that Bartholomew will do the Proof of Work for Alvin’s transfer. Of course, this works better with a third party. Alvin does the Proof of Work for Bartholomew’s transfer so that Cathy does the Proof of Work for Alvin’s, and so on.
This works best if none of the parties know each other. In a large scale network, there will always be some unknown participant available to do the Proof of Work for you. In the end you have a network where the “cost” of a transfer is electricity. It’s a network where you exchange electricity to have your transfer made. It’s a free, distributed, trust-less transfer network.
Proof of Work without miners
Designing this new protocol is challenging. In Bitcoin, miners reward themselves for completing the Proof of Work. Their fees come directly from the update they publish to the blockchain. They are paid as long as they are first to complete the Proof of Work. In this new protocol, another person rewards “the miner”. When I’ve done the Proof of Work for someone else, another person does the Proof of Work for me. If I do the work but am not rewarded, the system fails. If I don’t do the work but am always rewarded, the system fails. Bitcoin bound together the Proof of Work and the reward. How do you build a system that binds people together?
It’s helpful to visualize this:
Imagine each transfer request as the person that sent it. You need to select a transfer request on which to do the Proof of Work. This would look like shaking someone’s hand. Then you need to do the Proof of Work by finding the needle in a haystack. Finally, you need someone else to shake your hand and do the Proof of Work for your transfer (find a needle in another haystack). So there’s a gigantic group of people standing in a field. You (your transfer request) are somewhere among them. In order for your transfer to go through, you need to shake someone’s hand. Then you need to be the first to find a needle in a haystack. Once that’s done, someone else needs to come and shake your hand. If you’ve found the needle but no one comes to shake your hand, the system breaks. If someone comes to shake your hand but you haven’t even looked for the needle, the system breaks. How can we build a system that works?
What if instead of shaking anyone’s hand, you had to shake a specific person’s hand? If you’re standing in a field of people, finding that person would be difficult. In fact, with enough people, it would be like finding a needle in a haystack. In the new protocol, the Proof of Work is finding the right transfer to publish to the blockchain.
But what is the “right” transfer? How can we make this “right” transfer provably random? How do we know that finding the “right” transfer is as provably difficult as finding a needle in a haystack? Here, I borrow from KARMA, a reputation-based network from Cornell.
Let’s imagine the field again. It’s filled with people (that are actually transfer requests). This time, each person has received a secret card when they entered the field. On the card is a set of coordinates in that field. They’ve gotten these cards from a machine called a Hash. The Hash gives each person random coordinates. But, the Hash machine is special. Once a person reveals the coordinates on their card, anyone can check if that person is telling the truth.
The system needs a way of telling you which random transfer you need to find in your Proof of Work. A way of telling you which needle you’re looking for in the hay. A way of telling you which random person you need to find in the field.
Once you’re in the field, your goal is to find the person whose secret card coordinates are next to your own secret card coordinates. You do this by going up to each person in the field and asking them what’s written on their secret card. This is like picking up a piece of hay and checking if it’s a needle. Like the Proof of Work, it is provably difficult. Like the Proof of Work, it is provably random. By knowing the size of the field and the approximate number of people standing in it, we know how long it will take to find the right person. This is equivalent to the Proof of Work.
When you find the right person, you shake their hand and shoot up a flare. Both of your secret coordinates came from the Hash machine. That means that everyone around you can verify that you found the right person. This also means that the right person was randomly chosen — not chosen by you or anyone else. Shortly, someone with secret coordinates near your own comes and shakes your hand. Launches a flare. Everyone else verifies. The system continues. Each time a handshake occurs, a new version of the global spreadsheet is published and spread. The transfer and the reward have been tied together.
*note: I’ve linked to a more technical explanation at the end of this post*
Some interesting aspects of this system:
- Scaling. When you want your transfer completed you must complete someone else’s transfer. That means that transfer requests and transfer resources are bound together. The more transfer requests there are, the more transfer resources there are. The network and its transfer resources scale at at least 1:1.
- Timing. The Bitcoin Proof of Work changes its difficulty so that a needle is found every 10 minutes on average. This new system is also based on difficulty, and can be similarly regulated.
- Pair anonymity. The Hash machine gives coordinates randomly. This means that whoever receives the coordinates neighboring yours is also random. This means that you can’t decide whose transfer to complete (or not complete). This makes it very difficult to conspire or cheat the system, making the system safer. Most importantly, all of this can be verified by any other network participant.
- Multiple transfers. Bitcoin incentivizes miners to include several transfers in each Proof of Work. In this new system, you can require several handshakes. Just ask the person to find [x] people whose secret coordinates lie within a certain range of your own coordinates.
- Connectivity. The better connected you are to the network, the better your odds of finding the right person. The Proof of Work incentivizes participants to build a stronger, more inter-connected network.
More important, though, is that you get rid of miners. Everyone does the work of mining. Better still, mining is not rewarded with currency, but with the ability to make a transfer. The new system is free to use. The new system can transfer anything. The new system is trust-less. The new system is “to the right” of Bitcoin, and may represent the future of payment / transfer systems.
A future…
The excellent folks over at Coinbase published a list of businesses / applications they want built on Bitcoin. We can imagine a future with these apps built instead on the new system I’ve proposed. I’m calling the system SmallTime. SmallTime can do everything Bitcoin can do, but with less hassle / cost. People talk about Bitcoin as being equivalent to HTTP in scale. HTTP lets you transfer things by duplicating them across the network. SmallTime lets you transfer things without duplication. And unlike Bitcoin, it lets you do this without a middleman. It lets you do this for free.
You can create micro-denominations of dollars on SmallTime. Suddenly, I can require someone to pay me a few hundredths of a cent to send me an email. No problem for a friend; costly for a spammer emailing millions of addresses at once. Advertisers could pay me micro amounts to allow their ad content. iTunes videos could bill by the second, so whenever I decide a movie is boring I have only paid for a fraction if it. Better yet, digital rights could be re-sold. I could buy the lifetime right to watch the new Star Wars movie and then sell it to a friend. The same goes for albums and artwork. I can gift digital collections to my ancestors without worrying whether they’ll keep paying for Spotify.
Messages could be sent directly over SmallTime. My data and social graph would no longer be a monetizable asset for What’sApps, Twitters, etc. Digital rights could be applied to my digital footprint. I could link my digital assets (photos, social graph, etc.) to a key. If I share that key with Facebook, they can access my assets. The moment I decide not to, they lose access. Privacy becomes a concept again.
The line between currency and asset blurs. It is just as free to pay in dollars as it is in another currency, or stocks, or credits to a wireless mesh network. Whatever the other person / company decides to accept, works. More importantly, trade accelerates. I settle expenses and realize revenue in real-time. I am no longer an unsecured creditor of a Venmo or PayPal because it takes time for money to move from them to my bank account. Small denomination transfers unlock global remittances. Money moves from developed to less-developed nations in a stream, not a lump.
We all use a nominal amount of electricity to watch videos, listen to music, communicate, etc. SmallTime extends that list to include making transfers. Not just transfers of money, but transfers of almost anything that can be digitized. Anything that can be represented digitally. Anything that you want to move without its being duplicated. The scope here is the scope of the internet itself.
Where to from here?
I have tried to make this explanation as intelligible as possible. I hope that I’ve succeeded. At least, I hope I have proposed a somewhat interesting framework for thinking of advances in payments networks. Whether it’s SmallTime or something else, I’ve made an argument about its likely characteristics.
Why SmallTime?
Why have I called this new system SmallTime? Partly irony: I recognize the stakes. But also because right now, this is strictly smalltime. I’m a finance guy, a payments geek, and a product manager. I am not a cryptographer, computer scientist, or distributed systems expert. I’m putting SmallTime out there as clearly as possible so that everyone can rip it apart. Show me where I am wrong. Show me my ignorance, show me my misunderstandings and mistakes. Maybe SmallTime will survive. Maybe it is truly idiotic. I recognize the stakes, so I am all too happy to suffer my idiocy in public. Have at it.
For a more technical walkthrough, see here.
Please tweet any comments to @dbrgndl
/end.
