On October 19, 2022, Aera hosted a live Q&A event to introduce the new protocol. This post is an abridged transcript of questions from that event, answered by Tarun Chitra and Shaan Varia of Gauntlet with Peteris Erins from Auditless.
How do you choose the instruments that are traded within Aera?
Tarun: The objective function, the thing that is used to grade the submitters and the universe of assets are chosen by the DAO when they vote to deposit assets into an Aera vault.
Right now at least, the universe of assets is fixed at the initial genesis of the vault. I think long term we've definitely thought about how to add and remove assets, but starting with a pure constant function market maker portfolio means you sort of can't do that. Like it effectively becomes kind of mathematically a lot harder to deal with.
But, you know, In the long run, we're definitely committed to trying to find ways to add and remove assets and especially as there are more derivatives in the protocol, so options, futures, et cetera having some kind of ephemeral assets will be pretty important.
How do you prevent collusion between vault guardians and arbitrageurs?
Tarun: This is the skin-in-the-game aspect of having the vault guardian stake some amount proportional to what they're managing.
In the paper, I think we give some sufficient conditions in the appendix, but there's actually a follow on paper that proves some sort of game theoretic properties of what happens between the two of them. But roughly speaking, you could basically think of it as profit beyond some sort of target threshold.
And the target threshold is like, how much is the DAO willing to pay in transaction costs? Because obviously there's no such thing as free rebalancing. So for every arbitrage, for every amount of arbitrage profit that's outside of that threshold, that gets passed on to the vault guardians via getting slashed.
And provided that the assets that are used to stake, can basically be slashed and that value is correlated to the value of the arbitrage profit, you effectively ensure, even if they're colluding, it's a negative. It's a non-positive sum game between the two of them.
And so that's sort of the long-term idea. Obviously, this relies on getting the amount of collateral that's placed to be correct. So you could think of this as sort of effectively choosing a collateral factor or LTV, like in the lending protocol, for the individual submitters. And if you look in the appendix it gives some of the sufficient conditions for that.
But yeah, we definitely spend a lot of time on how to try to minimize collusion. Another reason for using Shapley values is that they have this really nice property that it's actually very hard to collude as a cartel to manipulate them. They're very expensive to compute, but they have really good manipulability properties. And so that's, that's sort of another thing that's quite important.
How do you choose an objective function or what is an objective function, in this system?
Peteris: Yeah, I'll just add another point to the previous answer. In the spirit of over-determining security, we also have some smart contract protections, as well. So we definitely considered scenarios. If there's a single guardian and they're also an arbitrageur, how much damage they could do, and just making sure that their power is as limited as possible.
Now, in terms of the objective function, I think this is a question that is kind of really at the core of the value proposition. The way we think about it, the objective function is a combination of returns and risks. So each treasury has a specific objective. In terms of lending protocols, the objective ultimately exists to cover potential shortfalls. And so that drives kinda the risk component. Not to dive into too much detail, but what we're trying to do is driving the objective function to what shortfall could happen, as a result of the specific loan book positions that exist in a given lending protocol. We have methods of trying to effectively evaluate those positions and convert them into liabilities. And then we look at how those liabilities are distributed across tokens.
So not only do we expect these objective functions to be different per lending protocol. But also they very naturally, I think as one can realize, generalize to potentially other related problems as well.
What tools will you use to hedge risks, options or perps or something else? And will you only be using mainnet or do you plan to cover other domains?
Tarun: So the initial vault, which is live on Polygon, has only two assets right now, which are ETH and USDC. And the next thing that we're working on support for options.
And I think the idea is to help on-chain options protocols that don't have a lot of organic long term demand. Options are the easiest thing, at least for lending protocols, to write their exposure as I think perps are nice, but I think the funding rate volatility makes it a lot harder for a more passive vault to actually handle it versus an option, which is actually a really nice concrete way of doing it. And you asked about mainnet. I think our goal is any EVM chain, so we're definitely not locked to mainnet.
What is the minimum that a guardian needs to stake and for what time?
Tarun: So, there are epochs. You have to stake for a particular number of epochs and you could think of it as, the DAO trades into that portfolio and then holds that portfolio over the epoch. And at the end of the epoch, you're graded based on its performance relative to your benchmark. And your benchmark, again, might be how well the portfolio covered the loan book, how diversified the portfolio was, or maybe it was absolute return. It could be any sort of thing.
We've defined it in a way so that you have this optimization problem you're solving as opposed to a single concrete thing. So the precise amount of stake in dollar or ETH terms will depend on conditions at the time.
And again, if you go to the white paper, Page 68 to 69 covers how we're measuring the amount of collateral that is needed from the guardians to guarantee a certain amount of manipulation resistance. So there's sort of just a natural trade-off between how much capital is staked and how much assets can be managed.
I think another important thing to mention is there'll be a delegation system, so DAOs can delegate stake to members of their communities who are active and want to help. So I think hopefully that should ameliorate some of the staking cold start problems.
What is the minimum number of guardians?
Tarun: So yeah, there's sort of a trade-off. If you have a really large number of guardians per vault, you actually have this problem where the grading, the Shapley Values, actually get way harder to compute and your approximation of them can get very erroneous and potentially manipulable. So I think the target number is probably in the realm of like 10 to 15. Because that’s where you can actually correctly compute all of the attribution mechanisms and then correctly compute the attribution, like how much each person's portfolio contributed to the final output.
But another way of thinking about it is that a DAO could have many objective functions. One might be, to cover the loan book. One might be runway over this amount of time. And you can have many vaults with different objective functions and 10 to 15 guardians for each of them. And I think that's sort of it's the right amount of people where there's enough specialization for that particular objective, but also enough that it's actually quite hard to collude.
What would be the best way to diversify through Aera? Is it to specifically target diversification? Or is there some other type of mechanism?
Tarun: Aera is designed around the idea that portfolio management is an optimization problem, and it's an optimization problem that may not have a fixed goal. If it was a fixed diversification you could just basically put it in a fixed Balancer pool or something like that, and then, you'll eat the arb loss, but you'll keep yourself at the target forever.
But I think the idea is that people have objectives that are more easily specified by an objective function for a KPI of the DAO. So in the lending protocol, the simplest KPI is assets minus liabilities, and that is the objective function that you're measuring.
There may be certain times when you actually need to hold a bunch of options because the market's extremely volatile and you know it's better to do some volatility harvesting. There are times when the market is super mean reverting. You may just wanna hold spot assets.
So it's not about a fixed static diversification as much as it's a fixed KPI. And so the reason we started with lending protocols is that lending protocols have this very, very concrete goal, which is that their insurance fund needs to ensure the solvency requirement.
Peteris: And we're continuing to discover more ways of thinking about objective functions and incorporating things. And I think one of the tradeoffs in terms of coming up with a good objective function, is also that you have to consider the cost of retrieving that information, which is very real for vault guardians. They have to build ETL pipelines that can ingest and estimate those objective functions, so it's also a nuance. We're trying to strike a fine line.