Why Not Build on Bitcoin?

December 16, 2023 by Alan Szepieniec

Why Not Build on Bitcoin?

Why are we launching a new layer-1 instead of building on top of Bitcoin or integrating our technologies into it? In short, Bitcoin does not currently support trustless two-way-pegged sidechains, and it is unlikely to adopt our technologies unless we prove their worth independently. Also, we want to build something that does not inherit Bitcoin's vulnerabilities.

Introduction

The question inevitably pops up when we explain what we are building to Bitcoiners:

Why aren't you building on top of Bitcoin or something to improve Bitcoin?

It's a fair question. After all, we are fans of Bitcoin and the architecture of Neptune is informed by what we perceive are its flaws. Bitcoin has an established track record and network effect going for it. Surely the distance to "Bitcoin, but better" is shorter from "Bitcoin, as it is now" than from yet another altcoin that needs to start from scratch. Right?

And that's just the practical aspect. There is also Bitcoin's outstanding cultural immune system, in whose eyes anything that's not Bitcoin is a scam looking to gullible people to defraud. That heuristic classification might be a valuable work-saving device when it works accurately in 99.9% of cases, but it doesn't help identify the 0.1% outliers. More importantly, it tends to turn people who would otherwise be receptive to the values espoused by Neptune into non-player characters.

Integrating (Some of) Neptune's Technologies into Bitcoin

Neptune features a bunch of technologies that could in principle improve Bitcoin when measured in simple straightforward metrics. (Part of the complication is of course that simple straightforward metrics do not paint the whole picture and there is no widespread agreement about which set of metrics are adequate.) For instance:

  • Confidential transactions hides UTXOs behind cryptographic commitments and proves the integrity of transactions (i.e., that the sum of all inputs equals the sum of all outputs) in zero knowledge. This construction benefits privacy.
  • Transaction aggregation using Triton VM replaces the signatures on a bunch of transactions with a single proof authenticating the entire batch, and potentially increases the on-chain transaction throughput capacity.
  • Post-quantum commitments and proofs – the commitments to UTXOs, the signatures that unlock them, and the proofs that replace the signatures, are all built using cryptography that promises to resist attacks on future quantum computers in addition to attacks deployed on present or future classical computers.
  • Recursive block validation also using Triton VM requires that new blocks prove the correctness of the previous block, obviating the need for the costly initial block download (IBD) and verification tasks upon first connecting to the network or reconnecting after some time spent offline.
  • Anonymous accumulation using Mutator Sets compresses the UTXO set to a short cryptographic digest that can then be used to prove membership of UTXOs to nodes that do not store the whole UTXO set, thus reducing storage requirements. Moreover, in order obtain this cryptographic digest nodes do not need to compute it from scratch (which would require downloading all UTXOs) nor do they need to trust other nodes on the network (which would require trust assumptions) because its most recent value is present in every block. Lastly, the anonymity of this accumulator benefits privacy by making it difficult to link transaction inputs to transaction outputs, a task that would be trivial otherwise.

Benefiting privacy, increasing transaction throughput, reducing storage and bandwidth requirements, are in abstract terms, unequivocal improvements. The challenge is that they induce controversial tradeoffs in practice. For instance:

  • All these technologies rely on the security of the Tip5 hash function, whereas Bitcoin currently only relies on the security of SHA-256 and RipeMD-160 (in terms of hash functions). Tip5 is not provably secure and is unlikely to be proven secure in the future (a feature it shares with SHA-256 and RipeMD-160). Integrating any one of Neptune's technologies into Bitcoin therefore implies introducing a new and relatively under-tested hardness assumption.
  • Triton VM in particular and multi-round proofs in general are only provable in the random oracle model, a simplification of the real world that replaces concrete hash functions with an idealized version for the purpose of producing a security proof. So even if Triton VM were to switch from Tip5 to SHA-256, its integration into Bitcoin would imply introducing a new cryptographic assumption – not a cryptographic hardness assumption, but rather the assumption that the random oracle model is adequate.
  • Even if the security proof in the random oracle model is palatable, Triton VM is instantiated using parameters targeting conjectural soundness, a regime of parameters where the soundness proof is not known to hold. Nevertheless, no attack is known and for all we know a proof for this conjecture may one day be discovered.
  • Even if all the security proofs did go through and in the standard model, that leaves the question of implementation errors. Vulnerabilities have been discovered even in popular cryptographic libraries years after they were introduced.
  • While verifying proofs instead of individual signatures or blocks reduces the workload and resource requirements for one user, this reduction is only possible if another user produces the proof to begin with. Without a change to Bitcoin's current incentive mechanism, there is no incentive for the proof-producer to work; and so ensuring the production of proofs requires changing one of Bitcoin's most delicate and worst-understood components – the incentive mechanism.
  • The use of accumulators requires users to keep the membership proofs of their UTXOs synchronized as the accumulator is updated. Synchronization requires users to be continually online or else spend effort upon rejoining the network to catch up before they are in a position to initiate transactions. (This task is relatively benign in terms of complexity but the point is that its complexity is nonzero.)

And these are just the technical tradeoffs. The pathway towards an upgrade to integrate some or all of these technologies is fraught with difficulties as well.

  • The integration of any one of these technologies would require at least a soft fork, and thus risks bifurcating the network or the community or both.
  • Any new technology introduced into Bitcoin is there to stay because phasing them out would mean eliminating code relative to which historical blocks are valid. Any new technology has to be maintained forever. As a result, maintainers understandably want 100% certainty that the proposed upgrade is the right solution and will not ever need to be revised (even as new science is developed).
  • The upgrade would need to enjoy overwhelming support from a large variety of developers, miners, influencers, and bitcoin holders. Previous hard forks have shown that this level of consensus is exceedingly hard to attain except for the most un-controversial changes.

Given that the technologies are controversial, they are very unlikely to ever be integrated into Bitcoin. The strategy has a very slim prospect of success and no route to monetization in the event of success. Under those conditions, who would fund this project?

And to the stereotypical Bitcoiner who answers reflexively: "charity!": even if your goal is to integrate these technologies into Bitcoin, the best possible argument in favor of their security and desirability – and therefore the likeliest pathway to success – is the deployment of a successful cryptocurrency that demonstrates their viability and features.

(That – and also: we await your check.)

Building a Sidechain on Top of Bitcoin

If integrating some of Neptune's technologies into Bitcoin's layer-1 is infeasible, why not build them into a layer-2 on top of Bitcoin instead? We could copy the bulk of the blockchain architecture and even tie the consensus mechanism to Bitcoin's in order to avoid a class of problems coming from bootstrapping a proof-of-work system in an environment full of parties knowledgeable about mining and laden with idle hardware.

In fact, this might be a case where proof-of-stake makes sense: when miners have to earn block production rights by staking Bitcoin. They are incentivized to stake and produce blocks because new blocks generate new Neptune coins. Neptune coins, in turn, are valuable because that's the currency in which the price of the staked bitcoins is declared and against which third parties have the option of buying the staked bitcoins at the declared price. (The system whereby taxees self-declare and tax man has the option to buy the goods at the declared price is known as a Harberger tax.) This mechanism has the potential to reel in the tendency of proof-of-stake systems to devolve into a self-reinforcing oligarchy.

The trouble with this strategy is that Bitcoin does not support trustless two-way-pegged sidechains.

(That statement is bound to attract some controversy. So we explain below in exact terms what we want and why existing proposals do not qualify.)

The same difficulty as laid out in the previous section applies – we cannot to rely on an upgrade to Bitcoin that may or may not happen. Additionally, tying Neptune's consensus mechanism to Bitcoin's in this way would sacrifice Neptune's claim to crypto-economic security in the world where quantum computers are built. The following scenario cannot be ruled out: the quantum attacker dumps stolen bitcoins, the price drops, and miners switch off their hardware. In this scenario, a side chain Neptune is just as vulnerable as Bitcoin is to reorganization attacks.

Validia Chain

John Light's excellent report on the topic calls the sought-after construction a validia chain and even identifies sub-categories that are out of scope here. The user experience should be some variant of the following:

  1. The user deposits bitcoins into a specially designated Bitcoin address for escrow, by making a regular transaction that sends the funds there.
  2. Once the deposit is confirmed, it unlocks an equivalent amount of bitcoins on the side chain and assigns them to the user's control.
  3. On the side chain, the bitcoins can be exchanged in a way that benefits from the features that the side chain offers (such as privacy and scalability).
  4. When the user wishes to withdraw his bitcoins to the main chain, he burns them while announcing a Bitcoin address that.
  5. On the main chain, an equivalent amount of bitcoins are unlocked from escrow and sent to the announced Bitcoin address.

The important feature distinguishing a validia chain from a validity rollup (from the same report) is that a validity rollup uses the main chain for data availability whereas a validia chain comes with its own data availability solution (and the details about how this custom solution works maps onto the various subcategories of validia chains). Validity rollups shrink the footprint of transactions on the main chain, but inherit the main chain's size limitation. In contrast, the marginal cost of one extra transaction on the validia chain, as measured by the main chain, is zero.

The common feature shared by both validity rollups and validia chains is that the bitcoins in escrow are attached to a UTXO that knows or somehow commits to the state of the side chain. This state can be updated only in accordance with the evolution of the side chain and only with a valid STARK proof (or transparent SNARK) attesting to this fact. Therefore, given the evolution of the side chain, the commitment on the Bitcoin blockchain evolves integrally without trust assumptions.

In order to support validia chains, Bitcoin requires at least one soft fork upgrade to enable two distinct features. First, UTXOs need to hold state and bind the ways in which they can be spent so as to ensure that the state persists and evolves in accordance to rules set out at the beginning. The technical name for this feature is covenants, and it can be achieved with OP_CHECKTEMPLATEVERIFY. Second, UTXOs bound by a covenant need to admit updates to the encapsulated state contingent upon a zero-knowledge or succinct-verifier (or both) proof. So Bitcoin Script needs a new opcode to verify such proofs.

BitVM

An exciting recent development due to Robin Linus called BitVM shows that contrary to popular opinion, Bitcoin is in fact Turing-complete1 already. The caveat of this true statement is that the computation is simulated by an interactive protocol between two parties. This protocol resembles the dispute resolution protocol in Lightning channels: the challenger, by exposing the fraud perpetrated by his counterparty, can take all the bitcoins locked in the channel. In this analogy, BitVM establishes that a state shared off-chain between two or more parties can be updated in accordance with pre-defined logic with crypto-economic integrity. Specifically, if one party updates the state fraudulently, or refuses to supply witness information, the other party can initiate a dispute resolution process that will end in the fraudster's loss of collateral.

The disqualifying feature of sidechains based on BitVM relative to validity rollups or validia chains is the crypto-economic versus cryptographic integrity guarantee. Specifically, when a zero-knowledge or succinct-verifier proof is required to update the state (as it is for validity rollups and validia chains) then it is impossible to update the state to an invalid one unless the updater manages to break the cryptography. By contrast, with BitVM it is possible to update the state to an invalid one; the updater just risks losing their escrow. And since escrow needs to be involved, this updater is in effect a centralized sidechain operator. Moreover, with BitVM, the sidechain operator can collude with miners to censor the disputation process. And it might be rational of miners to collude if their promised share of the spoils is significant.

Drivechains

Drivechains is an older proposal due to Paul Sztorc consisting of two BIPs, BIP-300 and BIP-301. The idea is that miners vote on withdrawals. With enough votes, a withdrawal is approved, and the escrowed bitcoins become liquid again. The voting process takes some 6 months so that there is enough time to act on malicious votes through the social consensus layer – for instance by banning blocks that include them or naming and shaming the perpetrator.

The disqualifying feature of drivechains is similar to that of BitVM: crypto-economic security as opposed to cryptographic security. Miners can collude to steal withdrawals if they are prepared to pay the cost. This effectively limits the volume of bitcoins that can be withdrawn securely per unit of time.

Furthermore, BIPS 300 and 301 have not been approved yet and it is not clear if they ever will be.

Liquid

Liquid involves depositing bitcoins to an escrow address controlled by a federation of authorities. A withdrawal goes through when a quorum of authorities sign off on it. Although Liquid denotes a specific side chain – the one spearheaded by Blockstream – it is nevertheless feasible to build another side chain with a similar setup.

The disqualifying feature is that trusted third parties are not even crypto-economic security mechanisms – they are attack vectors. Spreading key shares among double the number of parties does make an attack more cumbersome but not doubly so, and its effect is a far cry from that of doubling the security level. Moreover, the authorities have to stay in sync with $n-1$ other authorities, so in practice the breadth to which key shares can be spread is limited. Lastly, spreading key shares among trusted third parties obfuscates the fact that trusted custodians are still needed. History is replete with violations of that trust, and Bitcoin was supposed to deliver us from them.

Conclusion

Embedded in the heuristic of Bitcoin's immune system, which classifies all altcoins as scams, is the implicit allegation that founders and proponents of altcoins seek to enrich themselves at the expense of Bitcoiners. In this narrative, Bitcoin brought about the first fair distribution of money and any attempt to emulate or redefine this distribution is undermining it and thus unfair. Builders of alternative currencies are particularly guilty because they could have elected instead to build on top of Bitcoin and reinforce the fair distribution rather than undermine it. This allegation is particularly easy to direct towards founders of cryptocurrencies devoid of technological merit. (Are there really 10 000 technological innovations necessitating whole new blockchains, or are some of them just money grabs?)

In the case of Neptune, this allegation is irreconcilable with an important piece of countervailing evidence: the size of the premine is 1.98% of the asymptotical limit of the token supply. This premine is barely enough to fund the development of the project, let alone enrich its founders. (To anyone contesting this claim: we challenge you to fundraise for a new cryptocurrency with double the premine.)

In general we applaud founders who become rich as a result of their project's success, but we also make an exception specifically for money. Credible neutrality is an essential quality of good money, and so it has to be a precondition for success. A lopsided distribution, whether favoring founders or early investors, undermines that credible neutrality.

The reason why we are not building on top of Bitcoin is purely technical: Bitcoin does not currently support the features we need. In the future maybe, but we cannot afford to stake the success of our project on convincing a critical mass of skeptical Bitcoiners.

The narrative of the first and only fair distribution of money is just that – a story. And it could have been more enthralling if Bitcoin's limitations in practice did not drive beneficial technologies elsewhere. The true story is the tragedy of Bitcoin's first-mover disadvantage: a flurry of science and technology was developed in the wake of its success with the hope to improve it, but that success has made it too ossified to benefit from those improvements.

Acknowledgements: Special thanks for Louis Guthmann for fruitful discussions and suggesting the Bitcoin-staked Harberger tax mechanism.

Neptune citadel
Fig.1: The Neptune citadel will not be built on top of the Bitcoin citadel. Image courtesy if Midjourney.
1

In fact, BitVM shows something strictly stronger: Bitcoin, in its present state already, can decide undecidable functions. After all, there is a lookup table that maps algorithms to the Boolean value that indicates whether or not they halt. This lookup table has a Boolean circuit, and can thus be encoded into a BitVM contract. The disqualifier in practice is that we have no means to distinguish Boolean circuits that compute this function from Boolean circuits that do not.