Hello!

Over the past two weeks, I decided to pivot from writing to focus on creating contributions to the SSF space pre-Devconnect.

In practice, this involved me reviewing the most likely candidates for SSF adoption, RLMD-GHOST and Goldfish, and seeing if there were areas to improve in one of a few different ways. The first problem with RLMD-GHOST is that it doesn’t use subsampling, meaning that it doesn’t aggregate votes and instead requires every validator to vote in every slot. Currently, only ~32k validators vote in every slot, which is much more computation and bandwidth-friendly. Though this is beneficial for preventing some attacks, it’s also not necessarily feasible given our current computational power, and would likely require increasing block time (something I’m not opposed to, but is a serious consideration). An area of open research is how RLMD-GHOST could use subsampling and still retain its current properties.

Another path I went down was looking at EigenLayer, the new Ethereum re-staking primitive, and their claims about how it could be possible to create SSF through re-staking. This interested me for a few reasons. First, if SSF were to be implemented, doing it in EigenLayer first could be a good short-term solution, where we can see a mainnet implementation while also making it opt-in. In practice, I’m not sure how this would work, but it might look like re-stakers sending acknowledgment messages every round promising to not build on any block/slot except the one just published? Another reason I’m interested is because it’d be opt-in, meaning that users could potentially pay more for their transactions to be single-slot finalized. It may even be that users don’t have to pay anything extra, and validators run this service for a different incentive set. If it’s not clear, this is an early exploration, but one I find exciting and plan to dive into more!

Over the next couple of weeks, I plan to continue talking to the authors of RLMD-GHOST to discuss some nuances of the protocol, look at other potential candidates (Momose-Ren and Malkhi-Momose-Ren, both of which have some interesting design features I’d like to incorporate), and continue talking to the EigenLayer team about their current state of research.