Towards Iterable Institutions

A vision for how blockchains could create real change
Oct 27, 2023

Blockchains Are Almost Useful

The best decentralized technology so far is the Internet. But it’s not so decentralized anymore, and most of the value so far has been wrung out of it.

The second-best decentralized technology so far is blockchains. The value of a blockchain is that it allows people and organizations to make binding commitments about real actions without third-party contract enforcement. But the problem is that this isn’t very useful yet, because there’s not much to make commitments about. You can refer to on-chain information, but not much real commerce is actually happening there, because it’s still inconvenient to do business on a blockchain: it’s too expensive and not secure enough. People get their keys irrecoverably stolen, gas fees make large mainnet transactions untenable, bridges go up in flames with everyone’s money inside. Alternatively, you can refer to off-chain information. But although a patchwork of fledgling efforts is ongoing, there’s no clear solution yet to the problem of bringing complex, general, messy real-world data on-chain.

Progress is underway on both these fronts, though. On Ethereum, the expansion of rollups will allow computation to scale rapidly, and account abstraction will allow for more secure and consumer-friendly payments and finance. Interoperability efforts allow multiple chains to interact. All of this means much more economic activity will be able to happen on-chain.

And the building blocks for off-chain oracles that bring web2 and real-world data onto web3 are getting better: prediction markets, ZKML, proof of humanity, ZK Email, Schelling-point oracles like Kleros and Uma.


I think the really powerful thing about blockchain isn’t that we’re going to get a decentralized way to exchange money. I think it’s that we’re going to get a decentralized way to exchange general commitments.

A book that’s heavily influenced me is Thomas Schelling’s The Strategy of Conflict. It’s all about how most of mixed-motive game theory – that is, the theory of non-zero-sum games, which contain elements of both cooperation and competition – has to do with finding ways to alter your own payoff matrix by committing to certain actions beforehand. A classic example is the legal right to be sued:

Who wants to be sued! But the right to be sued is the power to make a promise: to borrow money, to enter a contract, to do business with someone who might be damaged. If suit does arise, the “right” seems a liability in retrospect; beforehand it was a prerequisite to doing business.

All of finance and trade is built on trust. You trust that your counterparty is good for the money, that they’ll deliver goods of sufficient quality, that you’ll have recourse if something major goes wrong. For a deal to be struck at all, each party must be convinced that the other has something to lose by failing to meet the terms of the agreement – usually either money, legal consequences, or reputation. The emergence of gig-economy and sharing-economy platforms wouldn’t have been possible without scalable trust systems: reviews and vetting processes mean you can book an Airbnb stay in someone’s home or an Uber ride in their car while remaining total strangers. Both sides not only want to be able to trust the other, but also want the other to be able to trust them.

Right now, the only way to make most commitments is through a central authority with the monopolistic power to enforce contracts. On marketplace platforms, the platform is the contract enforcer (in that it can ban anyone it wants). In most other contexts, that job falls to the government as the enforcer of legal agreements. But that’s not the only imaginable state of affairs. Schelling writes:

Consider a culture in which “cross my heart” is universally recognized as absolutely binding. Any offer accompanied by this invocation is a final offer, and is so recognized.

In a world like that, commerce would be extremely easy. You could do business with basically anyone, anywhere, knowing that they couldn’t walk back on any promise they gave. What’s more, commerce would be much more open and less gated; if you don’t need a central authority to enforce contracts, then you also don’t have to play by the authority’s rules.

Social Contracts

When I was a high school debater, there was this evergreen bit people would do. Some sweaty freshman would start in like “Well, according to John Locke, the government is legitimate because it’s like everybody agrees to a social contract, because they would all rather live under a democracy than in the state of nature –” and their opponent would stand up and go “Social contract? I didn’t agree to any social contract. Show me where I signed the social contract.” And then the first kid would have to stammer out something about how of course nobody really signed it but hypothetically of course you would if you could and your congressperson is sort of “signing” a “contract” for you and you’re sort of agreeing tacitly by receiving government services and not emigrating and so…

But what if we actually could make a literal social contract to sign? One of the cryptoeconomic ideas I’m most excited about is voluntaryist mechanisms for internalizing externalities, especially negative externalities. You could imagine a future where an entire system of government is specified on-chain, and individuals choose whether to bind themselves according to that system or another. The decision to fork a chain is a rudimentary version of something like this, but I imagine whole opt-in reputation systems, along with perhaps assurance contracts where you can agree to bind yourself to particular stipulations if and only if others do.

So okay, on an abstract libertarian-purist level, maybe this is better than a regular government, because it’s truly consensual. But on a pragmatic level, that doesn’t really matter to most people. Pedantic pocket-Constitution wielding teenager aside, no one goes around complaining that they didn’t sign the social contract. They’ve got gripes about the government, but the gripes are object-level, not “It doesn’t have the mandate of heaven.”

I agree. I don’t think ideological purity is the selling point. I think what’s actually going to be important is that decentralized commitment enforcement offers novel properties that centralized institutional commitment enforcement can’t.

Iterable Institutions

One such property is that a composable ecosystem of commitment layers – reputation systems, staking systems, proof-of-personhood, frameworks for individual contracts – can emerge. You could achieve the effects of regulation by committing to boycott companies that use particular practices (say, emitting more than a certain amount of carbon) if a certain number of other consumers do too, or by voting on the rules of a reputation system shared by many people, perhaps also committing to preferentially transact with people who also abide by your reputation system when deciding who they will transact with.

Another effect is that competition becomes possible. This has always been true of private companies, which can speedrun the lifecycle of an institution (get funded, develop, profit or die). Those selection pressures are why free markets optimize so effectively for producing the goods people want. But that’s only true of excludable goods. To generate public goods and allocate common resources, public-interest institutions are needed. But the much softer, slower selection pressures faced by governments aren’t enough to replicate anything close to the efficiency of private industry. So far, the state-of-the-art solutions for making government functional are

1) Think really hard from first principles and try to get it right from the start.

2) Do incremental change on existing institutions.

1 is impossible. Some people have put forth really solid efforts. The American Founding Fathers tried really hard to think of every possible exigency to build into the country’s governing framework, and did a pretty impressive job considering, but even they needed to scrap the Articles of Confederation a decade in to create a whole new constitution, and then they had to throw in the Bill of Rights after that, and even today the U.S. has problems the framers didn’t anticipate. No plan fully survives contact with reality. The world changes drastically over time, and even if it didn’t, central planning would just be too difficult to execute; there’s simply too much information for one party to absorb and act on.

2 is slow kind of by design: if it were really easy to make massive change to institutions, then everyone would do it all the time and it would turn into a power struggle, which isn’t the ideal preference aggregator.

But if we can make binding commitments about useful things before even creating a government, then we get iterable institutions. I call them that because they have a real iteration cycle. They can be born quickly: instead of waiting for an organization to become large, lindy, and reputable enough to be endowed with legitimacy, you can just trust the cross-my-heart commitment of an entity you’ve never heard of. And they can die quickly: the switching costs of adopting a new software-level application or protocol are much lower than the costs of, say, renouncing your national citizenship. A fork is easier than a coup.

Relatedly, in the status quo, commitment-enforcement services all have to be bundled together under the auspices of one large central enforcer (otherwise, who would enforce the enforcers?). Under a system of decentralized commitments, a model more like microservices could emerge. That further lowers switching costs, because you choose commitments a la carte instead of being locked into an all-or-nothing bundle. (For example, imagine if you could run a fork of the U.S. government that was identical to the current one except that it ran on an entirely different national ID system than drivers’ licenses and Social Security numbers.)

Together, these pressures open the door for a new kind of institution that hits the best of both worlds: public-interest, but subject to evolutionary pressures, because if it falls behind, a competitor can start up fast.

The vision is like: Parallel institutions. Markets for mechanisms. Governments that operate less like mainframes and more like virtual machines or even slim Docker images. Genuinely pluralist rule.

I think it’d be neat to have and I want to build it.