The “Unbundling of Trust”: how to identify good cryptocurrency opportunities?

Decentralization and centralization are two ends of a continuum. Look for opportunities to disaggregate “bundles of trust” to identify good opportunities in the cryptocurrency space

There are so many potential uses for cryptocurrency technology. But how do you know if any of them are good ideas? Blockchain-mediated financial exchange? I have a good feeling about that one. A bank-sponsored local currency system for small businesses? My sense is that it’s probably a terrible idea. But short of going out and building it, how would you know?

So are there any test you can apply beforehand to figure out if a blockchain is a good technical solution for a given problem?   And can you turn a bad idea into a good idea?

It’s a topic that comes up regularly when I present to audiences on Bitcoin and cryptocurrencies. Here are some slides I often use in these discussions. Slide 15 is where I discuss this topic.

Slides I sometimes show when presenting on cryptocurrencies. These represent my views, not IBM’s. But they are Copyright IBM. Please do not reproduce without asking permission first.

For me, the key to deciding if an idea is good enough is the one I’ve summarized on page 15 of the deck: this space is all about decentralization and if your problem isn’t about centralization then this technology may not be for you.

That may sound obvious. But internalizing this point is the key to understanding what a good cryptocurrency use-case looks like. And how to turn a bad one into a good one. Because even if your problem looks centralised, there may be portions that don’t need centralised trust and unbundling those components could be the key to doing something valuable. Here’s what I mean…

Go back to the beginning: what problem was Bitcoin designed to solve?

Bitcoin was invented as an answer to a decades-old question:

How do you come to consensus about some facts with a large group of people when you don’t know each other and some of you are cheating?

In Bitcoin’s case, the “facts” are “who owns what?”

And one answer to that question is, of course: “we all agree to trust somebody (e.g. a bank) and now we don’t need to trust each other”. But the obvious problem is: you have to trust the bank and that’s a potential point of failure. The breakthrough of Bitcoin was in showing us how to answer this question in a way that doesn’t require us to trust any single third parties.

We say the system is “decentralized”, as a shorthand for this concept.

(As an aside, I explained Bitcoin from first principles in this post on how the counter-intuitive genius of Bitcoin is that it works by going slow! For those who want to go even deeper, I share a way to think about the confusing “Unspent Transaction Output” concept in Bitcoin through an analogy of land.)

This is why Bitcoin is often positioned as being a decentralized equivalent to the centralized banking system:


Bitcoin allows us to agree who owns what without having to know each other or trust anybody else. This is the opposite of the traditional system where everybody has to trust their bank

Bitcoin-as-envisaged isn’t what we have

But there’s a problem: Bitcoin-as-envisaged isn’t what we have today. Phenomena such as mining centralization and the use of SPV Wallets mean that Bitcoin isn’t completely decentralized. It’s not currently a problem but one can already see its effects. For example, some miners refuse to mine certain types of transactions. The effect on average confirmation time for these transaction types might be marginal but it exists, nevertheless.

So Bitcoin-today is somewhere in between. It’s not 100% decentralised yet nor is it centralized.


Bitcoin today is neither fully decentralized nor is it centralised

So it seems reasonable to consider that centralization may actually be a continuum rather than an either/or phenomenon:


Are centralization and decentralization actually two ends of a continuum?

This way of thinking can be helpful because it allows us to think about other innovations in this space, such as Smart Property.

Smart Property

I’ve written in the past about a decentralized securities systems being built “hidden in plain sight”. The key idea here is that you can use blockchain platforms (like colored coins or counterparty, etc) to track the ownership and transfer of real-world assets. What distinguishes these platforms from Bitcoin itself is that they have to bridge to the real world: the asset could be a bond with a corporate issuer, being kept safe by a custodian bank, for example.  So there are several real-world entities on whom you depend.

I’ve written about how smart property allows these two roles to be merged (the issuing company could do both) but somebody has to do it – let’s just call them issuers.

So this system has points of centralization (the issuers) and points of decentralization (the ownership tracking and exchange). So perhaps it sits somewhere here on the continuum:


Perhaps there is value in different “degrees” of decentralization for different business problems

You can have more than one type of decentralization in a single service

But it’s actually more interesting than that. Because not only do smart property systems sit somewhere on the decentralization continuum, the key point is that different parts of the systems sit in different places:

  • the ledger, exchange and transfer system use the underlying Bitcoin consensus system – so they’re all pretty decentralized. No need to depend on any trusted third parties.
  • But you do, of course, have to trust the issuer.  That part of the proposition is centralised

So the important thing about smart property systems is that the all-or-nothing “trust bundle” is unbundled: you need to trust a specific issuer but the ledger, exchange and transfer functions are decentralized in their operation.

The unubundling of trust

And I think that’s what gives decentralised consensus systems some of their power: you can now break down products and services into their constituent elements of trust and implement each one with the most appropriate degree of centralization. For smart property, perhaps the picture looks something like this:


Different degrees of decentralization can exist within the same service: trust is being unbundled

So what?

Of course, none of this tells us whether smart property or cryptofinance is a good idea. But it is a way to think about whether a particular service is doing anything particularly novel.   Think about it the other way: if somebody proposes a cryptocurrency business idea that doesn’t meaningfully unbundle any trust in an existing service, is it actually doing anything valuable?  Likewise, take any real-world centralised service and ask yourself: what are all the things I need to trust for this to work? Which components have to be centralised? Which could be decentralised? Does that lead to lower risk? Lower cost? More opportunities for competition? Reduced friction for consumers? If the answer is “yes” to those questions then you could have an interesting proposition on your hands.

Unbundling trust in payments

A similar analysis works for systems like Ripple. Ripple’s architecture is more distributed than the traditional payments systems but less so than Bitcoin (at least as envisaged) so perhaps we may place it somewhere like this on the scale:


Ripple is another example of a “trust unbundling”

But, just like in the Smart Property example above, in the Ripple system there is a “trust unbundling” going on: the ledger is fairly decentralized in its operation whilst you necessarily need to trust a specific gateway.  So it actually looks like this:


Different degrees of decentralization can exist within the same service: trust is being unbundled

To see why this is important, recall how current payment systems work. I wrote a simple explanation of it here. As the article shows, you have to trust a lot of moving actors and the point is that you have to take this as a bundle… it’s all or nothing. You trust all those parts of the system or you can’t achieve your objective.  With a Ripple-like system, you only trust the minimal set of actors you have to – namely, the banks who issued liabilities.  Everything else can be decentralised to some degree.

Unbundling trust in contract execution

One last example: a similar argument applies to financial contracts. Projects like Ethereum (and Counterparty!) are exploring the decentralized modeling and execution of law. Gavin Andresen has written about how something similar could be achieved on the base Bitcoin platform.

You can think of this in terms of “trust unbundling” too: the decentralized platform ensures the integrity of contract execution and you can use n-of-m oracles to provide reliable external data. You only trust who you have to, to the minimal degree possible.

Using “trust unbundling” to turn bad ideas into good ideas..?

So now we can put this model to the test. Does it help us spot the silly ideas? Even better, does it help us turn the silly ideas into good ideas?  [UPDATE 2014-11-15 this section was heavily reworked)

Antonis Polemitis commented on an earlier version of this article:

Here’s what I think he means:

A better airline miles system?

As Antonis points out, airline miles systems are highly centralised: the airline is the issuer, redeemer, owner of the ledger, setter of the rules and controls everything else too.

So imagine an airline were to announce that their new airmiles programme was to be based on a fork of Bitcoin. Perhaps they would create their own Blockchain, issue the miles on top, secure it themselves and distribute wallets to all their customers. Brilliant…  an airline miles programme with all the benefits of Bitcoin!

Really? From a consumer perspective, surely this system would be indistinguishable from a traditional system and what is the argument that says it would be better in any meaningful way?

But take a step back and think about airline miles again and think about the trust bundle. Which parts of the system require you to trust the airline?  Issuance and redemption of the miles, for sure.  And setting of the scheme rules.  But storage, exchange and trade doesn’t need to be done by them.

And perhaps there’s a cost saving for airlines if they offload that work to a decentralised network and a benefit for customers if it gives them additional utility – perhaps new ways of swapping miles between competing programmes to accumulate enough points to book a flight?  Some very interesting possibilities emerge if multiple airlines base their systems on the same platform or if third parties can build new services on top of a platform like this.

Suddenly you might have something interesting: an interoperable, multi-provider airline miles storage, transfer and redemption platform. Now it could be a terrible idea – these schemes only work because most miles are never redeemed, after all! But the thought process is important: who are users expected to trust to use your service?  And what are they trusting them for?  What if a component was decentralised? What new possibilities would that enable? What risk could it mitigate?

Now the real world is more complicated than this. But the key insight remains:

  • if your cryptocurrency idea requires users to trust only you, you’re missing the point
  • but if there’s something in the value proposition that can be usefully decentralized or shared with others, you could be on to something

A decentralized securities trading and settlement system is being built hidden in plain sight

Colored coins, chromawallet, coinprism, NXT Asset Exchange, Mastercoin, Counterparty… tens of projects are working on asset tracking, transfer and exchange systems. What are they doing? Will it work?

I wrote a piece last year explaining how today’s securities trading and settlement systems work. The full picture of participants is pretty complex:

Figure 8 csd

There are surprisingly many parties involved in the safekeeping and exchange of securities. What would the picture look like in a “decentralized world”?

At core, I think the system is all about assuring “performance”. That is… it’s all about making sure that people actually deliver on the promises they make when they enter into a trade

Recent controversies might make this seem hopelessly naïve – and they show that ensuring fairness in exchange is important – but assuring performance is the core of the aspiration.

And to deliver on this aspiration, today’s system is based on a closed, centralized model. I talked about it here and also argued  Mt.Gox model was even more centralized than the mainstream system.

We’re now seeing serious projects work on this problem. Perhaps revisiting the fundamentals will help us predict which of these projects will prevail?

Why do we have exchanges in the mainstream world?  There are lots of valid answers (liquidity, fairness, …) but none of this matters if you can’t be sure a trade you make will be settled. After all, what’s the point of agreeing a trade with somebody if they can just change their mind afterwards if it suits them?

In the mainstream world today, the general model for a stock exchange is one where it has members, who are the only entities allowed to trade on that exchange. These members are subject to strict rules. For example, the London Stock Exchange’s rule book has over 100 pages:

Rule G5000 sums captures the critical function of the exchange for me:


“Obligation to settle: A member firm shall ensure that every on Exchange trade effected by it is duly settled.” Obvious, perhaps… but it needs to be said!

So the exchange helps ensure an orderly market by vetting and monitoring its members. This gives participants confidence: they don’t need to worry about who is on the other side of their trade. They know the trade they agree to will get settled. But other exchanges employ different models:

  • Prefund: Mt. Gox asked everybody to deposit their Bitcoins or fiat with them before they could trade. It guaranteed that trades executed on Gox would settle. Unfortunately, it only guaranteed they would settle on the books of Mt.Gox. As many people discovered to their cost, a settled trade on Gox was not the same as cash the bank or Bitcoins in their wallet
  • Escrow: The model I outlined in my piece earlier this year was essentially an escrow scheme. You place your Bitcoins beyond reach and they are either delivered back to you when your bid/offer expires or are delivered to the buyer. The trick here is in choosing the escrow “agent” (or agents…) carefully.
  • Clearing: This is how the The London Stock Exchange does it. In certain situations, members don’t even need to own the securities they’re selling at the time they trade them; they just need to make sure they deliver them as promised on the day of settlement. This model works because there is a closed group of trusted and well-known entities. However, there is clearly a risk: what happens if one of the participants goes bust between trade and settlement? That’s what a clearing house is there to solve, amongst other things. It keeps a close eye on its members, requires them to contribute to a “default fund” and steps in to make the other members whole if one of them fails.

Now, when we look at some of the most vibrant projects in the Bitcoin and cryptocurrency world, we see something interesting: a large number of them are working on representing non-crypto assets – such as securities – on the blockchain – They’re building out the vision of a decentralized general-purpose asset ledger.

There are two concepts we need to understand:

  • A token – something that represents an asset. Perhaps 100 shares of IBM Common Stock or ownership of a particular car.
  • An issuer – somebody that makes a promise to confer the rights and benefits associated with that asset to whomever holds it at any given point.

A concrete example: imagine I owned 10000 IBM shares (I wish…). I could issue them onto one of these platforms and publish the definition so others could see it and could see it was from me. I would, in effect, be making a promise:

“I will convey whatever benefits I enjoy through my ownership of these shares to whomever holds the token”.

So if I receive a dividend cheque, I pay it to the holder of the token. If you trust me to be good for this promise, you might be willing to purchase the token from me for $2m or so… the price of the IBM shares… owning the token would be just as good as owning the shares… and you could store it in your Bitcoin wallet and not have to deal with your broker any more!

Now, it is unlikely that you’d trust such a promise from me. But if was made by a major custodian bank you might. But note: you do have to trust the issuer.

So why bother? Why bother going to the trouble of building a decentralized asset ledger if you have to trust somebody at the end of the process?

For me, the answer is that this approach might allow increased competition between issuers. Furthermore, moving disparate asset registers (custody records, vehicle registration databases, etc) onto a common architecture might enable innovations we haven’t yet considered.   It’s too early to tell so we can all be grateful to the pioneers who are experimenting so we don’t have to.

I think there are three broad camps:

  • Coloring Bitcoins. Projects such as chromawallet and coinprism are working on systems to “tag” Bitcoins so that they can be tracked across transactions
  • New Protocols Running Over Bitcoin. mastercoin and counterparty piggy-back on Bitcoin’s peer-to-peer network, double-spend protection and consensus system but their tokens are essentially independent of Bitcoins. A counterparty token is not simply a “tagged” Bitcoin.
  • Entirely Separate Protocols. NXT and ethereum fit into this camp.

I have no particular insight into the structure of any of these projects so let’s assume they’re all run by capable, honest people and further assume that we’ll see a future where assets of all types, including securities, will be represented on a blockchain-like decentralized platform.

Then what? Presumably people will want to buy and sell…. To exchange.

And that’s where things get interesting… because we have to solve the performance problem.   We’re now in a decentralized, pseudonymous world… how do we ensure somebody who offers to buy an asset for a given price actually goes through with it and pays up?

What is the crypto-ledger rule G5000?

Is it possible to build a decentralized exchange on any of these platforms that has the strong performance guarantees we need? Can we build a decentralized exchange where a matched bid and offer inevitably lead to a settled trade?

It we look at our three models from previously, “clearing” isn’t going to work (it is, by definition, centralized and reliant on trusted identities). “Prefunding” is also problematic – what happens if the entity you sent your assets to disappears? So it looks like “escrow” is the only game in town.

Now, part of the solution already exists: we can construct “atomic” asset transfers using the Bitcoin protocol today. So I will assume exchanging payment and asset in a single transaction (“Delivery versus payment”) is achievable today on any of the platforms discussed above. But we need to get to a point where creating a valid transaction like this is inevitable once a bid and offer are matched.

Here’s where I think the state of the art is with the three approaches and it’s surprisingly different:

Coloring Bitcoins. The systems I’ve looked at don’t route bids/offers over the Bitcoin system so any matching will be done external to the platform. So it seems to me that “decentralized exchanges” on this model will have to require those posting bids or offers to demonstrate that they have placed the corresponding colored coins/Bitcoins in escrow with one or more acceptable third parties. There’s nothing that will do this automatically. So, it’s worth watchin firms like Xapo in the US and Elliptic in the UK. Professionally-run Bitcoin “cold storage vaults” such as these feel like “proto custodian banks” that could perform this function. The question is: can they devise a service that is sufficiently decentralized yet which still allows them to earn an income?

New Protocols Running Over Bitcoin. My understanding of these systems is that they embed bids/offers in the blockchain and have a protocol definition that means matches can be determined unambiguously. Furthermore, the act of making a bid or offer locks the associated assets until the trade is resolved or a bid/offer expires… automatic escrow, if you like. Assuming I am right, then this does appear to offer the “inevitability” promise that I think is so important. But it is at the expense of polluting the blockchain with bids/offers. It seems inelegant to me that one would store transient data (time-limited bids/offers) in such a permanent form of storage. But perhaps there’s no other way?

Entirely Separate Protocols. My working assumption is that NXT, too, works on the basis of bids/offers encumbering the associated assets until the outcome of the trade is resolved.  With Ethereum, the answer to every question is, of course, “it’s Turing Complete so of course you can do it” but I need to dig a little deeper to be sure….


Where is this going?

I think we’re going to see a market test: the colored coin approach is, in many ways, the most elegant as it uses the blockchain solely for storing/transferring the asset.   It means a range of exchange types can be trialled (escrow, pre-funding, reputation-based?)… but none of them will deliver full “inevitability” of settlement.  Perhaps consumers will care. Perhaps they won’t.

Projects like mastercoin and counterparty look able to deliver on the “inevitability” promise but will it be at the cost of blockchain bloat?

It will be an interesting few months ahead.


A final thought… What if we simply don’t worry about it and price it instead?!

The other approach is completely radical… instead of trying to force performance, why not model it as an option? We can think of somebody who posts a bid/offer but who then reneges as exercising an option to renege. This option clearly has value – if they would lose money by completing the trade as agreed, the option payoff is at least as much as they stood to lose! So is it possible to model the value of the option to renege and force participants to pay the option value up-front in order to post a bid/offer?

Unanswered questions: to whom would the price be paid? Is there any precedent for modeling the “option to renege” in this way? What would be the liquidity implications?


I said at the start of this piece that a new financial infrastructure is being built “hidden in plain sight”. For the reasons outlined above, I think the “exchange” aspect of this infrastructure still has a long way to go but we’re about to witness a fascinating experiment.