What Makes a Well-Designed Cryptoeconomic Protocol?
TL; DR
Naval Ravikant proposes the following three criteria for a well-designed cryptoeconomic protocol: (1) the resource that the protocol is aggregating and allocating is provided by the network participants, (2) the resource provided by participants can be digitally verified and is fungible and (3) the network needs its own native token and could not more easily run on Bitcoin or Ethereum. By analyzing four different cryptoeconomic protocols, Filecoin, Tezos, Numerai and Steemit, we see that while Naval’s criteria are helpful, they are not always applicable to what may otherwise be a well-designed cryptoeconomic protocol. The analysis also exposes other features of cryptoeconomic protocols, such as unique incentive mechanisms and consensus protocols, that further elaborate on the above three criteria.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —— — — — — — — — — — — — — — — — — — — — — — — —
Introduction
In the recent Investor Field Guide podcast called Hashpower, in which Patrick O’Shaughnessy interviews leading cryptoasset investors about blockchain technology, decentralization, Bitcoin, Ethereum, ICOs, cryptography and hashing, Naval Ravikant discusses his own personal framework for evaluating well-designed cryptoeconomic protocols. After all, why can’t every centralized intermediary be disintermediated with a decentralized protocol? The three criteria he proposes are:
1. The resource that the protocol is aggregating and allocating is provided by the network participants.
This means that the actual organizer, which traditionally would be a centralized authority, does very little other than laying down rules in the code that are enforced when participants use the network. This is what makes the protocol decentralized and distributed. If the organizer is providing the resource, then you are better of with a traditional centralized service like Amazon’s AWS for cloud storage and processing power or JP Morgan for loans.
2. The resource provided by participants can be digitally verified and is fungible.
Digitally verified means the resource can be confirmed by computers on the network without having to access the physical world. Fungible means each unit of the resource is indistinguishable and therefore substitutable for any other unit of the resource (e.g., crude oil, shares in a company, bonds, precious metals or currencies). Without digital verification, some other authority would be required to verify the resource, which means that not all participants can verify the resource and the protocol should probably just be centralized. Fungibility is related to digital verification because it makes the verification process easier if there is no variance within the resource. For this reason, protocols that claim to be a “decentralized AirBnB” are less likely to work because AirBnB units are unique and physical in nature. Each unit is different and determining if a unit is available or occupied, the state of its condition or if it even exists cannot be digitally verified. In contrast, cryptoeconomic protocols based on storage space, compute cycles, bandwidth, solar power or the cellphone grid could theoretically be well-designed because these resources can be digitally verified and are fungible.
3. The network needs its own native token and could not more easily run on Bitcoin or Ethereum.
This means that unless there’s some reason the network can’t run using Bitcoin’s or Ethereum’s tokens, their incentive mechanisms and/or consensus protocols, then there’s a concern that the token itself is just being used as a fundraising mechanism that bypasses securities regulation through an initial coin offering (“ICO”). The Bitcoin network already has many of the desirable features needed for a cryptoeconomic protocol such as being peer-to-peer, borderless, fast, secure and cheap. Bitcoin has these features because of its (1) incentive mechanism that uses its own native token, “bitcoin”, and (2) consensus protocol, Proof-of-Work (“PoW”), which collectively incentivize participants to validate and record transactions.[1] But Bitcoin is limited in the types of transactions it can do because of its non-Turing complete scripting language and inability to maintain state (i.e., transactions can only be defined as spent or unspent without any other conditionality). Ethereum, in contrast, has Bitcoin’s features plus a smart contract system with a Turing-complete scripting language that can maintain state. Ethereum has its own native token, called “ether”, and uses PoW to incentive participants to validate and record transactions.[2]
In listing these three criteria, Naval is careful to point out that they are not “fully baked” and are by no means industry standard. Nonetheless, I found myself agreeing with these criteria and thought it would be interesting to look at a few cryptoeconomic protocols and analyze whether they pass the test. The protocols I have chosen are Filecoin and Tezos, which are the two highest grossing ICOs, and Numerai and Steemit, which are unique use cases of cryptoassets that have gained a lot of attention.
*Note to the reader: Because cryptoeconomics is such a new field, the below analysis does not intend to make an absolute claim that any of the protocols are “well designed” (although I would argue that based on VC investment, coverage and market caps alone they have at times been considered some of the top protocols[3]). Instead, the goal is to analyze whether they have the three criteria discussed and to discover what else can be learned about cryptoeconomic protocols in general.
Analysis of Cryptoeconomic Protocols:
Filecoin
Filecoin is a decentralized storage network that turns cloud storage into an algorithmic market. The native tokens in Filecoin are called “Filecoin”.
1. The resource that the protocol is aggregating and allocating is provided by the network participants.
Yes. The resource in Filecoin is cloud storage. The participants are storage miners, retrieval miners and clients that want data stored. In the protocol, clients pay miners Filecoin for providing cloud storage available on their computers to store client data and for retrieving that stored client data when requested. The Filecoin protocol itself consists of an algorithmic “Storage Market” and “Retrieval Market” that matches miners with clients and has code that can prove data is stored and replicated across the network. Protocol Labs, the creator of the Filecoin network, is not otherwise involved in mining or using the cloud storage resource.
2. The resource provided by participants can be digitally verified and is fungible.
Probably yes. Cloud storage can be digitally verified by confirmation that the client’s data takes up space within the storage miner’s hard drive. The Filecoin protocol uses Proof-of-Spacetime and Proof-of-Replication consensus protocols, which guarantee that storage miners have correctly stored the data they committed to store. Regarding fungibility, while a typical storage provider may be concerned with things like capacity, performance (e.g., throughput, latency), workload (e.g., read/write, block sizes, sustained or variable demand), availability (high availability, clustering, support, SLAs etc.), backups (e.g., snapshots, long term archiving, restore times) and security (e.g., location, governance, compliance), these are technical and business issues that can be abstracted away by the Filecoin protocol to make them irrelevant to participants (or at least make it so participants are not aware of the relevance). Therefore, we can say that cloud storage is fungible within the Filecoin protocol.
3. The network needs its own native token and could not more easily run on Bitcoin or Ethereum.
Yes. Filecoin use an incentive mechanism and consensus protocol that are different from Bitcoin and Ethereum. Therefore, it needs its own native token. The Filecoin token is unique from bitcoin and ether because: (1) clients pay Filecoin to store and retrieve data, (2) storage miners earn tokens by offering storage and (3) retrieval miners earn tokens by serving data, with the likelihood of a miner being selected to store or retrieve data (i.e., verify a block on the Filecoin blockchain) being proportional to the amount of their storage currently in use in the network. While bitcoin or ether could theoretically be used by clients to pay miners, Filecoin would not be able to create and compensate miners by generating bitcoin and ether based on their storage space since the Bitcoin and Ethereum networks only reward their respective tokens for showing PoW.
Tezos
Tezos is a platform that, like Ethereum, enables developers to build decentralized applications with smart contracts. Tezos is different from Ethereum because it has its own governance rules that allows participants to approve protocol upgrades that are then automatically deployed onto the network, thereby allowing the protocol to maintain network effects whenever a new cryptoasset innovation is discovered. The tokens in Tezos are called “Tezzies”.
1. The resource that the protocol is aggregating and allocating is provided by the network participants.
Yes. The resources in Tezos are computation of smart contracts and the Tezzies themselves. The participants are developers, nodes and miners. Like Ethereum, nodes offer their computational resources to run smart contracts and miners verify these smart contracts onto the blockchain. However, in contrast to Ethereum, the resource provided by miners is not computation, but rather the Tezzies they are willing to “stake” to verify a transaction. This is because Tezos uses a Proof-of-Stake (“PoS”) consensus protocol rather than Ethereum’s PoW.[4] The Tezos foundation is solely responsible for writing the code with governance features that allow participants to approve protocol upgrades and is not otherwise involved in computing smart contracts or mining.
2. The resource provided by participants can be digitally verified and is fungible.
Yes. Computation of smart contracts is digital and can therefore be digitally verified. The Tezos protocol uses a PoS consensus protocol in which smart contracts are recorded onto the Tezos blockchain by miners who stake their Tezzies for the chance to receive more Tezzies in proportion to their staked Tezzies if they are randomly chosen to create the next block. Miners are incentivized to act honestly because if they do not then they lose their staked Tezzies. Regarding fungbility, while the smart contract code is variable and therefore not fungible, the smart contract code will run so long as it gets computed and verified. Therefore, the computational resources and Tezzies staked by the miners are fungible.
3. The network needs its own native token and could not more easily run on Bitcoin or Ethereum.
Yes. Tezos uses an incentive mechanism and consensus protocol that are different from Bitcoin and Ethereum and can change its consensus protocol as more optimized solutions are discovered. Therefore, it needs its own native token. Tezos’ use of its native token also enhances network effects by allowing any value added to the network to be captured by the entire Tezos network. In Tezos, developers are rewarded with Tezzies if their proposed protocol upgrade is approved and added to the Tezos protocol. If developers were rewarded in bitcoin or ether, then the only value-add for the rest of the Tezos participants would be an upgraded Tezos network. The bulk of the value would go to the developer, which seems fair, but also other bitcoin or ether holders that may have no interest in Tezos.[5] By instead rewarding the developer in Tezzies, which is required to use the Tezos network, the value-add is captured and maintained within the Tezos network.
Numerai
Numerai is a hedge fund that offers a platform where data scientists compete in a tournament to build machine learning models based on historical financial data provided by Numerai. Before submitting their models to the tournament, data scientists stake Numerai’s native token, “Numeraire”, to show their level of confidence that the models will perform well in the real world. Data scientists are then rewarded based on the performance of their models and their level of confidence. The total reward is based on how much Numerai the hedge fund earns when it uses the models to make investment decisions.
1. The resource that the protocol is aggregating and allocating is provided by the network participants.
Mostly no, but yes where it matters. The resources in Numerai are machine learning models and Numeraire tokens. The participants are data scientists. In the protocol, data scientists are paid Numeraire by Numerai for creating financial models that perform well against historical financial data and against other models submitted by data scientists in proportion to their staked Numeraire.[6] Numerai, the company, is responsible for providing historical financial data, running the tournament in which data scientists compete, making investment decisions based on the best performing models, and paying out the data scientists. This gives Numerai more control over the protocol than other analyzed protocols. However, it is the data scientists that are providing the most important resource — in the form of their machine learning models — that is necessary for Numerai to function.
2. The resource provided by participants can be digitally verified and is fungible.
Mostly no. The machine learning models can be digitally verified because data scientists can check their models against the historical financial data to determine performance and adjust their level of confidence accordingly. However, the data scientists have to trust that Numerai is accurately comparing models in the tournaments and that they’re being appropriately awarded. Numerai refers to this entire verification protocol as Proof-of-Intelligence. Therefore, the resource can be digitally verified but participants must otherwise trust Numerai to be a good actor to operate the rest of the network. The resource is arguably fungible because any model can be submitted to the tournament, but there is obvious variance amongst models. However, because Numerai partially acts as a central authority, it can discern amongst this variance and act accordingly. In fact, the variance in models is crucial to Numerai’s operation because it increases competition, which increases the chances of producing financial models that will perform well against actual data.
3. The network needs its own native token and could not more easily run on Bitcoin or Ethereum.
Probably yes. Numerai uses an incentive mechanism and consensus protocol that are different from Bitcoin and Ethereum. Therefore, it needs its own native token. Numeraire is unique from bitcoin and ether because it is used to access the staking tournament on Numerai, which allows data scientists to show confidence in their models and earn more Numeraire. Interestingly, when Numerai originally launched, data scientists were rewarded with bitcoin. Numerai determined that bitcoin provided limited incentives and network effects because participants were discouraged from inviting others to join. There was a finite amount of bitcoin to give away in each tournament and so why would a data scientist want to invite more competition to potentially beat her out of winning bitcoin? In this sense, the value gained from each tournament accrued to the individual data scientist rather than the network as a whole[7]. To solve this problem, Numerai (1) issued their Numeraire token and (2) introduced the staking mechanism to participate in the tournament and show confidence. This resulted in better incentives and network effects because it meant that inviting more data scientists created more demand for Numeraire tokens (since Numeraire are required to participate in the tournament), which thereby increased the value of Numeraire and the Numerai network.[8] It is important to note here that the introduction of the token and staking together created the need for Numerai’s native token. The token needed to have a use case within the network. Otherwise adding more data scientists would not have resulted in increased demand for Numeraire and bitcoin would have continued to suffice.
Steemit
Steemit is a blockchain-based social network that allows participants to create content, promote the content they believe is good and comment on content, all while being rewarded in the process.
Steemit is a bit more complicated than the previous cryptoeconomic protocols and warrants further description before getting into the three criteria below. Steemit uses three different native tokens to create a unique incentive mechanism to encourage long-term investment in its network: (1) Steem: a liquid value token that can be bought and sold on exchanges but does not provide utility on the Steemit network, (2) Steem Backed Dollars (“SBD”): dollar-pegged tokens that are worth around $1, accrue interest and are used to protect content creators from Steem volatility and (3) Steem Power: a non-liquid token that cannot be bought or sold on exchanges but provides utility and gives holders influence over the Steemit network.
1. The resource that the protocol is aggregating and allocating is provided by the network participants.
Yes. The resources in Steemit are content and Steem Power. The participants are content creators, content curators (i.e., up-vote or down-vote content) and witnesses. To create or curate content, participants must use Steem Power. The more Steem Power a content creator the higher they can push their content and the more Steem Power a content curator has, the stronger their vote. The witnesses act as a type of miner with two functions: (1) verify blocks of Steemit content on the Steemit blockchain by staking Steem Power in a consensus protocol called Delegated-Proof-of-Stake (“DPoS”) and (2) publish a price feed of Steem/USD to the network to maintain the SBD value. Witnesses are compensated with Steem Power for each block they create. Steemit the company does not produce or curate any content and does not act as a witness.
2. The resource provided by participants can be digitally verified and is fungible.
Yes for digitally verified, no for fungibility. Content can be digitally verified on the Steemit blockchain using DPoS and each block contains information about the posts, comments, votes and currency transfers. This digital verification allows participants to trust the content because any data related to the content is recorded on the blockchain by witnesses that are discouraged from censoring content because of the DPoS mechanism. The content itself is clearly not fungible. However, fungibility does not seem to be a requirement for this use case because participants can select what content they want to create, curate and consume.
3. The network needs its own native token and could not more easily run on Bitcoin or Ethereum.
Yes. Steemit uses an incentive mechanism and consensus protocol that are different from bitcoin and ethereum. Therefore, it needs its own native token. Steemit’s use of three different native tokens alone clearly proves that it couldn’t be run with bitcoin or ethereum. But three tokens could just be an arbitrary complexity. More importantly, it’s the incentive mechanism in place that creates the need for three different tokens. Part of Steemit’s goal was to develop a cryptoeconomic protocol that not does fall victim to pump-and-dumps and other schemes that tarnish the long-term potential of cryptoassets. To do this, they’ve tailored their tokens to keep participants invested while also accommodating short to medium-term interests. When a participant creates content, they can be rewarded in 100% Steem Power of 50% Steem Power and 50% SBD. The Steem Power can only be converted to Steem and sold after two years. It is therefore a long-term capital investment, which Steemit encourages by making it the network’s utility token. SBD is more like a bond that provides consistent returns but can converted into Steem and sold after five days. The result is a stable cryptoeconomic protocol that can be used for both utility and speculation.In addition, Steemit’s DPoS consensus protocol has a unique payout structure that could not be accommodated by bitcoin or ethereum. For each block mined on the Steemit blockchain, approximately 2/3 of the Steem Power reward goes to the content creator, approximately 1/6 goes to the content curators and the remainder goes to the witness that created the block. Clearly bitcoin and ethereum could not be used for this. Therefore it is clear that Steemit needs its own native tokens.
Conclusion
From the above analysis we can see that while Naval’s criteria are helpful, they cannot always be easily applied to what may otherwise be a well-designed cryptoeconomic protocol. The above analysis also leads me to a few other conclusions regarding cryptoeconomic protocols in general:
- The organizer may play a large role in the operation of a cryptoeconomic protocol so long as the main benefit of the protocol (i.e., the resource) is provided by participants. For instance, Numerai the hedge fund is essential to the value of the Numerai network. Without them the data scientists would have no data to model and no means to compete against one another. But the main benefit that comes from the Numerai protocol — the machine learning models that lead to successful investment decisions on real data — is provided by the data scientist participants.
- Fungible digital resources can more easily be decentralized and distributed but they do not have to be fungible if the protocol can (a) abstract around this so that participants are not aware (as was the case in Filecoin), (b) if there’s a central authority to discern among the non-fungible resources and act accordingly (as was the case in Numerai) or © if participants are otherwise capable of verifying whether they want to use the resource (as was the case in Steemit). Fungibility makes a cryptoeconomic protocol easier to implement because if everything is the same then there’s less complexity in verifying the resource and less need for a central authority. But this is not an absolute requirement. After all, even bitcoins are not truly fungible because bitcoins may be tainted by illicit activity that stays with the bitcoin for all subsequent transactions.
- There can be multiple resources that are aggregated and allocated in a cryptoeconomic protocol. This seems to be the case in protocols that involve any form of staking mechanism (as was the case in Tezos’ PoS, Numerai’s Proof-of-Intelligence and Steemit’s DPoS).
- Incentive mechanisms and consensus protocols that are different from Bitcoin’s or Ethereum’s will probably require a native token. This was the case for all the cryptoeconomic protocols analyzed.
- Native tokens can be used for the sole purpose of improving incentives and network effects. This was the case in Tezos rewarding developers with Tezzies to ensure that the benefits of software upgrades stayed within the Tezos network and in Numerai rewarding data scientists with Numerai to encourage them to invite more data scientists to compete in the tournaments. The Numerai example also shows that protocols involving competition can benefit from a native token when participants would otherwise be discouraged from adding more competition.
In the future I plan to analyze other protocols to see what else can be discovered from different examples of well-designed cryptoeconomic protocols.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —— — — — — — — — — — — — — — — — — — — — — — — —
[1] Specifically, in a Proof-of-Work consensus protocol miners use their computational resources to compete to solve a cryptographic algorithm. The miner that wins (and therefore assigned to validate and record the current block of transactions onto the blockchain) is rewarded with newly minted tokens for its effort.
[2] Ethereum is scheduled to upgrade its consensus protocol from PoW to proof-of-stake (“PoS”) early next year. PoS is further discussed below.
[3] Tezos is currently facing class action law suits in California and Floridadue to internal governance issues. This has caused delays in the development of Tezos and it is unclear how it will affect its market cap in the future.
[4] Specifically, in a Proof-of-Stake consensus protocol miners are assigned to the ability to validate and record transactions deterministically in proportion to the number of tokens that they are willing to stake rather than based on a competition of computational resources to solve a cryptographic algorithm. The miner that wins is rewarded with transaction fees in the form of tokens (rather than minting new tokens) for properly adding a new block to the blockchain.
[5] This is because the Tezos reward would increase demand for bitcoin or ether and the value from that increased demand would accrue to bitcoin and ethereum users, respectively.
[6] Note that while data scientists “mine” the historical financial data for useful insight, this is not the type of blockchain mining referred to in the context of other cryptoassets.
[7] Like the Tezos example, value would also accrue to bitcoin holders because the Numerai reward would increase demand for bitcoin. As previously explained, this is undesirable when the value can be redirected to stay within the network.
[8] The addition of staking had the additional benefit of producing models that performed well on actual data rather than just being made to work against historical data.