zerosync.org/zerosync.pdf
Interesting state of play update on this zerosync project.
The most important high-level idea to understand here is: the proofs created here are fixed length, and very cheap to verify, independent of how much history you prove. This mindblowing feat is accomplished (without trusted setup, mind you!) at a cost: the computation required to actually *create* the proof is enormous. It's easy to see why that tradeoff is particularly attractive for initial sync of a bitcoin node - the history is a single, global (very big) thing.
Their current code apparently takes 3.5 hours for 3K transactions (i.e. approx 1 block), so they are very far from breakeven (10 minutes per block). Also there are other big limitations still: they haven't yet done segwit, let alone taproot (so can only look at pre-2017), and they haven't implemented witness validation, so can only do a `assumevalid` type of initial sync. They are relying on utreexo but that itself is not much of a limitation.
As much as that list of limitations might seem comical, this feels to me like the start of a slow paradigm shift.
The final section of the paper talks about a rather speculative but fascinating `zkCoins` protocol that is arguing for a much more powerful version of token exchange in a client-side validation model (they claim it's better than existing RGB/Taro stuff and I'm inclined to believe them though of course it doesn't actually exist yet!).
Minting would need either federatiion, or on-chain snark verifying for proper two-way peg, or burning btc (I'd argue perhaps locking, also could do the trick - time value of money etc).
