Podcast notes – Vitalik Buterin (After the ETH Merge) – Epicenter
Host: Friederike
Guest: Vitalik
50% of blocks are OFAC compliant – “it’s a concern…but important not to overstate it”
Means non compliant transactions have to wait 2 blocks instead of one (on average)
Near-term solution – MEV boost will have transaction inclusion lists added – similar to what MEV-Geth (?) did before
Social slashing is a meme that’s gone too far – shouldn’t hard fork to delete censoring validators
“Optimistic that things will improve quite a bit”
Should validators have agency?
General idea is to make validators as dumb as possible – just run code – maximally dumb pipe-y – more predictable and easier to run a validator
Argument for making validators smarter / opinionated – could serve as second line of defense
Doesn’t trust threshold encryption – requires 50% honesty assumption
Suspicious of honesty majority assumptions – should have paths to recovery if dishonest majorities occur
MEV smoothing – Justin Drake’s fave idea – treat non optimal bid acceptance as invalid / non availability condition
How much of a problem is MEV?
Many kinds of MEV – some is a problem, some isn’t
One class of problem – outright exploitation (eg, Uniswap ETH-USDC, sandwich attack)
Some MEV is unavoidable / benign – eg, arbitrage if prices change during block confirmation, keeps prices synchronized
Don’t want proposers to need to update software to keep pace with MEV algorithms
Can’t count on dapps to mitigate MEV – “always gonna be dumb devs somewhere”
There are MEV minimizing architectures – eg, off chain order matching before sending to Uniswap / onchain DEX
Some (eg, Paradigm) argue MEV inevitable – and thus building solutions to capture MEV in somewhat decentralized way is necessary (eg, Flashbots)
Flashbots has prevented centralization of layer under them (the stakers) – but has turned into centralization vector itself
The Surge – danksharding requires trusted setup, should we worry about it?
Low probability (1 of 1000s?), but important to move away from trusted setup over time
Other approaches have too many tradeoffs
KZG now, roadmap for removing it as better snark tech catches up
Phase 1 – Proto danksharding – could be early to mid next year
Phase 2 – go to 16mb, split up data load – will be gradual nodes transition
Core development bandwidth limited, prefer solutions that are more distributed / third party
Examples: Account abstraction strategy – 4337; Rollups
Danksharding – benefit is split off development effort, core devs have simpler problem / less work, rest of work is on community, can get something out much faster
Concept of in-protocol fees going to specific dev teams – has been discussed before and rejected – trying to minimize governance
Core devs want role to be as technical as possible, avoid social value judgments
If can go back 8 years, pre-mine 3M ETH for long-term fund for soul-bound governance, maybe — but need to live with ecosystem as it is today
Centralized chokepoints in L2? Sequencers are centralized
Decentralizing sequencer is very important – multiple approaches and tools, balance complicated constraints (security, legal)
-In-protocol auction to buy up sequencer rights for future slots
-In-protocol proposer mechanism
$3B in hacks in 2022 alone – how do we protect normies?
Defi hacks have been in applications he doesn’t use and would never endorse using
Some in community have more aggressive ideas of what they wanna do onchain and will overshoot
“Best we can do is slowly expand frontier of what can be done safely over time”
Uniswap safe for long time, MakerDao, better DAO governance contracts – this safe space will grow
“Do better job of communicating difference between safe zone and crazy zone”
What must we get right in 2023?
“I’d still say scalability…there’s a limited time window”
If we don’t solve by next bull market, overwhelming chance that forms of scaling that sacrifice trustlessness will dominate, will be hard to come back from it