A recent crypto-economic design pattern that I’m very excited about is the Token-Curated Registry (TCR). It’s simple, elegant, powerful and has extremely broad application.
In this post, I’ll explain what it is, where it can be used, it’s potential limitations, it’s connection to my work on Curation Markets, and then ending of with some crazy extrapolation exercises.
Origins: The adChain Registry
The TCR formalization discussed in this blog post was codified for use in the adChain registry by Mike Goldin, Ameen Soleimani & James Young, who based their work off of an idea originally promulgated by Aventus: a blockchain-based event ticketing protocol. To avoid fraudulent events around ticketing, a registry is used to certify providers.
I quote parts from the adChain Registry introduction:
Introducing the adChain Registry!
Hello again! In our last post, The Ad Tech Landscape Today: Existing Problems As We See Them we addressed many of the…
In short the adChain registry is:
…a smart contract on the Ethereum blockchain, which keeps a cryptographically secure record of publisher domain names. The domains listed in the adChain Registry are, “accredited as non-fraudulent by adToken holders”.
It aims to address 3 concerns:
- Advertisers and Publishers Cannot Accurately Verify One Another Throughout the Supply Chain
- The CPM is an Industry Standard that Encourages Fraud
- Facebook and Google Have a Pejorative Duopoly
…and in using a TCR, it:
- It serves as an identity mechanism which allows advertisers to know something about listed publishers without ever having had a previous relationship with them.
- It serves as a rich source of premium publisher domains for advertisers that want to run campaigns with verified sites.
- The key value proposition of the adChain Registry is that it decouples the CPM incentive structure from adToken holders. adToken holders have one goal and one goal only, to vote in reputable publishers and vote out fraudulent ones.
The core, incentivized game being played through the token-curated registry system is to include reputable actors and exclude non-reputable actors. In the case of the adChain registry, it aims to build a whitelist of reputable publishers. Token holders earn from helping to curate this in-group against fraudulent publishers.
The Generic Token-Curated Registry
From this origin, Mike Goldin has written up a post/paper describing the TCR generically.
It’s a wonderful read, and highly recommended.
How It Works: An Overview
- Publishers submit a deposit (saying 100 adToken) in order to become a part of this whitelist.
- In doing so, they have to wait out a challenge period. If honest and reputable, none of the current adToken holders should dispute/challenge this publisher and they will become a part of the whitelist after the completion of the challenge period. Their deposit is then kept, staked to their listing.
- During the challenge period, if a current adToken holder feels that this publisher will degrade the quality of the whitelist, they can issue a challenge, by submitting an equal amount to the listing applicant’s deposit. This initiates a voting period.
- adToken holders then proceed to vote whether to include (‘yay’) or deny (‘nay’) this publisher to the whitelist. Any adToken holders can then come and vote.
- After the voting period concludes, tokens are settled as follows:
- If the challenger succeeds, the applicant’s deposit is distributed from them to the challenger and the voters on the winning side (‘nay’). With a 50% special dispensation, the challenger receives their deposit back along with an extra 50 adToken. This dispensation is a votable parameter by the TCR. The winning voters receive 50 adToken distributed by the amount of tokens they voted with. The voting process locks the tokens and thus is an opportunity cost on the token voters. They are reward for helping curate the registry and for foregoing the opportunity cost of voting.
- If the applicant succeeds, their deposit is kept.
Whilst an application is a part of the registry they may be challenged at any point in time. If the registry’s deposit requirement is increased and a listing is not “topped up”, the listed item can be instantly removed by any token holder.
The post from Mike describes it in more detail, along with other interesting details (such as potential attacks).
Dissecting the TCR feedback loops
As you can see, this model is fairly simple. There’s a fortuitous feedback loop occurring. As Mike Goldin describes: it contains 3 actors, namely consumers, candidates and token holders.
With this whitelist, it reduces the barrier to entry for the facilitation of a market exchange.
The consumers are looking for a curated set of actors that allows them to do less effort in actuating the exchange. Instead of verifying all the actors in the market independently, they verify the registry instead, and if they believe it has quality differentiation, it amounts to less work required.
The candidates, those who want to be a part of the list can get more customers if they believe that this list is worth the effort to enter. If they comply, then paying the cost will result in more market exchanges coming their way.
Finally: the token holders (whether it is candidates or separate holders), the curators, gain value in judiciously maintaining the integrity of this whitelist. If they do the work to ensure it is high quality it will attract both consumers and candidates.
This feedback loop will continue until it hits a point of marginal return: where the cost for an additional candidate to enter is not worth it for either the candidates or token holders. This cost, is the cost evaluate these candidates and the cost to maintain the list’s integrity over time. More on this later.
Re-using Token-Curated Registries
A lot of crypto-economic systems rely on using tokens and appropriate cost to avoid nefarious actors from gaming the system. By adding eustress in the appropriate locations, it captures the work being done effectively, and thus the value being added.
TCRs were invented in the context of facilitating exchange between publishers and advertisers, but as Mike Goldin describes (snipped a bit by me and bolded by me):
The product or output of a token-curated registry is a list. Humans have a penchant for list-making. Most lists can be abstractly classified as either whitelists or blacklists, and in both cases the contents of a list uniformly satisfy some criteria.
Useful lists are curated. Often by a single individual in the case of a grocery list, and perhaps by a committee in the case of a top-colleges list.
A token-curated registry uses an intrinsic token to assign curation rights proportional to the relative token weight of entities holding the token. So long as there are parties which would desire to be curated into a given list, a market can exist in which the incentives of rational, self-interested token holders are aligned towards curating a list of high quality. Token-curated registries are decentrally-curated lists with intrinsic economic incentives for token holders to curate the list’s contents judiciously.
I think this understates the value immensely. So much of what humans do are just effective list making.
When you hear of the TCR, you start seeing lists everywhere. Two recent examples I saw: seeing food standards ratings in restaurants, and tourism grading. Some of these ratings systems are regulated and some of them are independent bodies. But, with, a TCR, you unbundle it, and give the control of such a registry over to market forces.
Lots of new value can be generated, but also because the competition for lists is reduced, we can see more competing TCRs arrive. In effect, any classification, attestation, grading, rating system could be replaced with a simpler TCR… and when you think about it. It’s pretty broad.
It’s already started seeing application in other decentralized applications such as the District Registry.
TCRs as Binary Curation Markets
TCRs thrive in being binary systems: you are either in or out. It’s been interesting to compare it to the work I’ve done with Curation Markets. To compare, let me introduce you to Bo-taoshi.
It’s an interesting variant on Capture the Flag where a competing team has to boulder through a swarm of enemies in order to grab the flag at the top. It’s fun to watch!
With Curation Markets, one of the core goals has been for people to use tokenized signalling to curate information and reduce information asymmetry. What I love about TCR is that it is exactly that: using tokens and a crypto-economic game to reduce information asymmetry and incentivize a group of token holders to curate information (in this case: a list). It makes markets smarter.
In the case of a TCR, this list is like a city wall. It’s binary. There’s only an in and out. 0% or 100%. If you are inside, you are safe. In order to be a part of the benefit provided by the city wall, you have to prove you are worthy to be within it. Those within the city walls also want to make sure they aren’t letting anyone in that can destroy the city from within.
With Curation Markets on the other hand, it is rather a more continuous classification. It’s almost like a game of Bo-Taoshi. If you want to be safe from attackers, you want to be in the center as much as possible. The center is a 100% classification. As you progress outwards to the edge of swarm, the classifier goes down to zero. There is no wall, but you can still be safer if you are in the middle.
Thus, with a more practical example, if adChain used a Curation Market rather than a TCR, you would have a weight attached to each publisher signalling their reputability, rather than an in-or-out classification.
A Curation Market, however, is more complicated, untested, and brings potentially more complexity than is needed.
However, with a Curation Market, I see there being one core benefit due to having a reduced barrier of entry to classify: a larger economy size. There is no voting game. There’s only staking information.
Critical Mass & Resourced-Based Models
One of the most interesting unknowns at the moment with TCRs is just how large they can become. Mike Goldin touches upon it briefly in the paper, describing it as “economy size”.
What is the minimum size of an economy necessary to support the decentralized curation of a list in this way? Is it economical to decentrally curate a grocery list?
At some point the marginal return to maintain the registry is reduced such that token holders might not be incentivized sufficiently anymore to maintain integrity.
Early on, adding to the list grows the value and thus, the curators (token holders) are incentivized to add more members. However, adding another website when there are 100k of them, doesn’t add that much additional value. And thus, the mental costs (in verifying reputability) becomes larger than the benefit gained from doing the work.
The problem with this is that at this critical mass, the churn could be problematic if curators aren’t doing work anymore. The registry gets poisoned with unverified actors joining (no one challenged an unreputable actor). It then causes the value to go down. As the value goes down, actors are incentivized to redo the work (token value increases) and challenge the bad actors to get them to leave, but then marginal return kicks in where the work isn’t worth it again. And thus, this constant churn will come to exist.
This churn is reminiscent of the resource-based model from Butler that you find in online discussion forums.
At some point, this churn will inevitably happen. I touch upon this in more depth, in this post on eustress and complexity.
A way around this is to raise the deposit to maintain novelty at critical mass. BUT. The problem is that this critical mass is curtailed by the design of the registry system and how effectively information can be shared/vetted. Thus, the critical mass might exist at a level that’s not fit for various registries: eg, the size and amount of applicants might not be what the registry is aiming for. For example, we might want ALL websites in the world into adChain, but the registry can’t realistically maintain this.
There are ways to add additional weighting to such a registry, and Mike proposed a simple solution when I discussed this with them. When in doubt, add a prediction market. For example, add a market to determine if a publisher will still be reputable in 6 months. This adds more market signals, by adding a new weighting, and more financial incentives to do more work.
However, one thing that Curation Markets help with is that because it is ALL just a stake-based system, one gets access to a continuous vetting system (between 0 to 100%), as opposed to a binary one (only 0% or 100%). It thus reduces the barrier even more to curate information.
It is however not that simple (yet).
- It adds an additional vector of responsibility to the consumers, which is determining where they want to draw the line (is over 50% a reputable publisher? How do I know?).
- There’s more unknown complexity with Curation Markets that still need to be tested.
Extrapolating: Mitosis, Continuous Registries & Tokenizing Classifiers
One other way to avoid the above problem is to allow this registry to undergo mitosis at critical mass, splitting into more specific registries. It’s like a cell splitting into two. The market has clearly proven a demand for a certain set of criteria, and thus because it is not functioning properly anymore at a certain level, it can not only produce more novelty by having more specific criteria, but also keep the integrity of the registries by becoming more specific.
A way to more easily force the hand of allowing mitosis to naturally occur is to move towards using continuous token models with TCRs.
In order to participate in the registry the token is minted through a continuous token model. It’s value increases as more tokens are in circulation. At any point, the token holder can dispense their token back into the currency they bought it with.
Thus, with continuous registries, at any point, one can use that specific TCR’s token and create a sub-group within the larger registry, that has more specific criteria.
If these registries can undergo this more natural mitosis, then one could start extrapolating in interesting ways. It would then be the case that as soon as a market develops for more specific criteria, that that registry would want to start proving itself, and undergo mitosis. It could even be the case that a specific registry accepts currencies for token minting from other registries, meaning it has many parents, not just one specific one.
If that’s the case, then one could imagine that TCRs could eventually become market-based classifiers. They are incentivized to add tokenized tags to everything in the world. In a way, one creates a market for hashtags.
…but, who knows? We could just be underestimating the transaction costs involved in TCRs, and they are only broadly applicable in larger crypto-economic games?
I’m really, really excited to see TCRs in action. I think it’s a simple and powerful tokenized design pattern. I’m excited to work with people who are pioneering these amazing designs.
There are however some still unknowns, but I think over time we’ll see how they fare.
Who knows… Maybe these tags on this Medium post, might in the future be done through a TCR. The claps are just votes towards being included into the tags. ;)
Thanks to Mike Goldin & Niel de la Rouviere for review & feedback on this post.
Personal disclaimer: I’m an engineer at ConsenSys. I do not hold any adChain tokens.