## Protocol economics: the Ethereum protocol's perspective --- ### What are some protocol economics questions that matter? &nbsp; * What are the incentives facing users? * What are the incentives facing consensus actors (miners / validators)? * What is the (i) average-case, and (ii) extreme-case behavior of the chain, and how does that affect incentives at the application layer? * How do applications affect any of the above? ---  --- ### What are some risks to watch for? &nbsp; * Complex user incentives &#8594; wallet centralization * Economies of scale for consensus actors &#8594; consensus centralization * Incentives to participate in censorship or reorg attacks * Apps that work fine during average protocol behavior breaking in extreme cases * Incentive to have low latency &#8594; geographic concentration &#8594; jurisdiction risk --- ### Rise of defi <img src="https://i.imgur.com/T7JWzpf.png" style="width: 42%" /> <img src="https://i.imgur.com/6Eqvn6a.png" style="width: 50%" /> * Burst MEV existed before (eg. 2017 ICOs), but not enough to justify developing infrastructure to try to capture it; since 2020 it's permanent * Creates economies of scale for consensus actors --- ### EIP 1559  * Greatly reduced waiting times for users * Much less incentive to use complex gas price strategies &#8594; lower barriers to entry for wallets --- ### The merge &nbsp; * Completely changes the class of participants maintaining consensus (miners -> ETH stakers) * Makes reorg attacks vastly harder (see https://www.paradigm.xyz/2021/07/ethereum-reorgs-after-the-merge) * Very different incentive landscape re centralization and pooling --- ### Upcoming changes &nbsp; * Withdrawals enabled * Single secret leader elections * Different types of in-protocol PBS * Single slot finality --- ### Withdrawals &nbsp; * People can withdraw, so some will withdraw &#8594; less ETH staked * But... people can withdraw, so people become more confident they can get their money back if they stake &#8594; more ETH staked --- ### SSLE &nbsp; * Only the proposer will know that they are the proposer before the block is produced * Though they could voluntarily reveal this info * Intended effect: reduce DoS risk against proposers * Are there unintended effects? Particularly, might there be incentives for proposers to pre-reveal their identity to anyone? --- ### Single slot finality &nbsp; * Even more reorg security (see https://www.paradigm.xyz/2021/07/ethereum-reorgs-after-the-merge) * More bridge-friendly * Though with layer 2 pre-confirmations, maybe not de-facto single-slot... --- ### Layer 2 and protocol economics &nbsp; * Sequencing style: "based rollups" vs L2-controlled sequencing * What percent of L1 defi will move to L2s? * Fraud proof windows and censorship * Preconfirmations * On the whole, will the L2 shift make L1 economics "more boring" or "less boring"? --- <img src="https://i.imgur.com/Y5Cz5NR.png" style="width: 60%" /> With PBS, total anarchy ("based rollups") is much more viable! But what would the consequences be? --- ### Could layer 2s absorb "complex" layer 1 MEV? &nbsp; * If rollups are based, no * If a large part of L1 defi stays on L1, no * If rapid L1 tx inclusion helps L2 arbitrage in other ways, no * Otherwise.... maybe? --- ### Fraud and censorship &nbsp; * If you can censor a chain for a week, AND that chain doesn't socially reorg you, then you can steal from optimistic rollups * However, much shorter-term censorship attacks on defi exist too --- ### Preconfirmations  Even if have single-slot finality on L1, we still have pre-confirmations on L2. --- * Arguably, this is a point _in favor_ of SSF: it means that we could compromise on slot times (eg. to 32s) to make SSF easier on L1, and it would not hit users much * But still, we have two tiers of confirmation * This is fundamental: L1 is not willing to centralize as much as L2, and so L2 confs will always be faster * Note: preconfirmations are incompatible with based rollups --- ### Will L1 be more boring or less boring? &nbsp; * My hope: more boring. But this is not guaranteed! * All depends on to what extent L2 absorbs functionality and risk, as opposed to creating new and more complicated interaction effects with their own risks * Need more analysis of a fully post-L2 world!