Lifting Expensive Computation Off-Chain

There are lots of things that applications want to do on-chain because they require the security of L1, but they simply can't because they require too much data or are too computationally intensive.

With Maru, they can "lift" their computation off-chain, do it there, and then "settle" it back on-chain, where they can use the result however they see fit. Whatever the computation, it would look something like this:

  1. Pull in raw log / tx / state streams from every chain whose data we wish to incorporate in our computation if they haven't already been added to Maru.
  2. Write operators that filter for just the input data to our computation
  3. Write operators that perform the computation
  4. Subscribe to the output in one of the following ways:
    • Subscribe from an off-chain service (API, Browser, etc) and use them there
    • Build a merklized index (by address) of the result, and publish this merklized index to cheap, decentralized storage (e.g. Arweave or Filecoin). Browser clients and APIs can then interactively query the index.
    • Settle the result and proof to a smart contract on any chain whose applications wish to use it. In the case we use data from chains other than the one consuming it, this constitutes a bridge, which comes with extra trust assumptions discussed in the bridge use case (which is the case for (almost) any bridge).

Here are some specific examples:

Credit Scores

Many companies want to build on-chain credit scores for the following reasons:

  1. Lending on chain is currently over-collateralized. Allowing under-collateralized lending would be a massive capital unlock for DeFi, and that requires the ability to "predict" or "quantify" risk of default.
  2. Lots of projects (DAOs, social networks, stuff involving the "real world", etc) are investing in building reputation / "trust" scores for users. But it turns out that credit scores can be used as reputation scores as follows:
    1. Instead of using reputation scores directly, use bonded stake.
    2. Allow users to borrow from an undercollateralized lending market to pay for that stake. The higher their credit score, the lower their collateral ratio, and the closer to "free" their stake becomes.

While there's a lot of research going into the credit scores themselves, there's another problem that will arise - the sheer amount of data needed to scalably calculate the scores. Robust scores will likely require data from several ecosystems and may be resource-intensive to compute. Performing this computation on-chain is infeasible, but taking it off-chain into a "SaaS Product" presents a massive trust problem not unlike the credit rating agencies we use today. If we care about making scores transparent and replacing the trust we place on credit rating agencies today with cryptography, it's important to be able to prove that scores are correct.

Companies that would be interested in this include:

  • Lending Markets (Aave, Compound, Teller, etc)
  • Social Networks (Mirror, Lens Protocol, etc)
  • DAOs / Contribution Tracking (Gitcoin, Sourcecred, etc)

Off-Chain Pricing of On-Chain Assets

Bonds and Options are difficult to price on-chain, because efficient price formulas (e.g. Black-Scholes) are too computationally expensive and require too much data to perform on-chain. Being able to use better pricing strategies with the same security as L1 presents a large capital efficiency improvement without losing security.

Companies that would be interested in this include:

  • Bonds Markets (Element Finance)
  • Options Markets (Ribbon Finance, Synthetix, Zeta Markets, etc)