L1-zkEVM Breakout #004
Transcript
- Manu
GM!
- Francesco Risitano
gm
- Ladislaus
Hat auf "GM!" mit 👋 reagiert
- Gary Schulte
gm
- Gary Schulte
is it quiet or do I have audio trouble>
- Manu
Quiet for me too
- Luca Donno | L2BEAT
some zklofi
- Andrés Ramírez
Reacted to "some zklofi" with 😂
- Gary Schulte
Reacted to "some zklofi" with 😂
- Kevaundray Wedderburn
Might be good to note that ZisK is below 400KB right now
- Luca Donno | L2BEAT
Reacted to "Might be good to not..." with 🔥
- Ladislaus
Hat auf "Might be good to not..." mit 🔥 reagiert
- Iván | ethrex
Reacted to "Might be good to not..." with 🔥
- Gary Schulte
Reacted to "Might be good to not..." with 🔥
- Ruben
Reacted to "Might be good to n..." with 🔥
- Gary Schulte
ethp2p?
- Kevaundray Wedderburn
Replying to "ethp2p?" For reference: https://github.com/ethp2p/ethp2p
- Manu
Sorry I have to go. Thanks all!
- Kevaundray Wedderburn
Reacted to "Sorry I have to go. ..." with 👍
- Francesco Risitano
Reacted to "Sorry I have to go. ..." with 👍
- Francesco Risitano
Thanks Manu 🙂
- Gabriel Trintinalia | Besu
Will the next release be bal-devnet-7 spec?
- Gary Schulte
can we share presentations links for easy access to hyperlinks? esp: Ignacio's
- Kevaundray Wedderburn
Reacted to "can we share present..." with 👍
- Łukasz Rozmej
Reacted to "can we share present..." with 👍
- ignacio
yep! https://docs.google.com/presentation/d/1xfMKpYMVVowi-tqX8Q0CxddGC24MMEgQervSi5JQFzc/edit?usp=sharing
- Gary Schulte
Reacted to "yep! https://docs.go..." with 🙏
- Kevaundray Wedderburn
Replying to "can we share present..." Links will also be shared under the pm issue
- Luca Donno | L2BEAT
Is there any interest in exploring making program hashes independent from the underlying zkVM? Or is it not worth the potential overhead?
- Iván | ethrex
I’m getting the third reviewer for #6560 😅
- Łukasz Rozmej
@ben come look at Nethermind ;)
- ignacio
Reacted to "@ben come look at Ne..." with 🔥
- Iván | ethrex
Reacted to "@ben come look at Ne..." with 🔥
- ben
Reacted to "@ben come look at ..." with 👍
- ben
Reacted to "I’m getting the t..." with 🙏
- Łukasz Rozmej
@ignacio we will be integrating with tests and then would love to integrate with benchmarks
- Francesco Risitano
https://frisitano.github.io/slides/presentations/optional-proofs-2026-05-13/
- Iván | ethrex
Reacted to "@ignacio we will be ..." with 🔥
- Francesco Risitano
Reacted to "@ignacio we will be ..." with 🔥
- Ladislaus
Hat auf "https://frisitano.gi..." mit 🙏 reagiert
- ignacio
Reacted to "@ignacio we will be ..." with 🔥
- Gary Schulte
Reacted to "https://frisitano.gi..." with 🙏
- Ruben
@Marcin Bugaj i will
- Han Jian
- Gary Schulte
Reacted to "https://docs.google...." with 🙏
- Francesco Risitano
How big is the size of the standalone verifier binary?
At a glance
Detailed notes
Cl Integration
- •Prysm now supports 3 modes: proof generation (via ZKBoost), proof verification, and EL-less verify-only mode00:08:13
- •Optional proof spec still based on Gossip; plan to migrate to Gloas soon00:11:33
- •Lighthouse and Prysm Kurtosis interop devnet running; devnet launch targeting next week00:11:46
- •Max proof size bumped to 400 KiB (interim); likely to exceed 1 MB before prover optimization; ZisK already below 400 KB00:15:52
- •Validator proof re-signing proposal deprecated; peer scoring deemed sufficient for DOS prevention00:17:04
Spec And Witness
- •Spec/test release 0.34 adds block test engine format and execution witness compliance dashboard00:25:17
- •Next release (imminent) adds schema ID prefix to guest input bytes, chain config in stateless input, and proper public key field population00:26:33
- •EIP-7709 proposed at last ACDE for Hegota; simplifies BlockHash resolution in execution witness00:30:39
- •Execution witness fixtures found a bug in a BAL implementation, validating test coverage value00:31:12
Guest Program Optimization
- •eFrex/ZISK profiling yielded up to 26% reduction in AIR costs: 21% from switching to precompile hash calls; ~0.5% from removing rayon/sync overhead00:34:13
- •SSZ serialization rewrite of execution witness planned to replace archive format00:36:27
Zkvm Guest Api Standardization
- •5 standards accepted: RISC-V target (RV64IM), I/O interface, Crypto Accelerator API, termination semantics, memory layout00:37:39
- •Two new standards in review: misaligned jump must cause abnormal termination (no rounding); vendor runtime spec (static lib + linker script)00:38:50
- •U256 acceleration benchmarked on ZISK: multiplication 2-6x gain, but addition/subtraction mixed or slower; impact on 500 real blocks negligible — not yet standardized00:41:10
- •Nethermind ZKVM effectively first to fully comply with unaligned-access standard per Cody's testing00:53:14
Infrastructure And Verification
- •ZKBoost config simplified in Kurtosis package: prover image and guest program links now resolved automatically00:46:07
- •Nexus verifier crate separated from prover and integrated directly into ZKBoost; prover container no longer needed for verify-only mode00:46:31
- •Formal RISC-V semantics verification of ZISK proving system in progress; external review targeted within weeks00:50:25
Action items
- •Execution-APIs PR for witness retrieval (withWitness boolean + executionWitness return field) needs review from EL teamsEL teams
- •Contribute to open issue on proof type encoding to support generic guest/ZKVM combinations (e.g., native rollup use cases)Community / L2 teams
- •ZKVM teams to review and opine on Keccak-F permutation accelerator PR vs. full Keccak function interfaceZKVM teams
Key decisions
Validator proof re-signing proposal deprecated; peer scoring is sufficient for DOS prevention
Peer scoring is sufficient for DOS preventionMisaligned jump standard: must cause abnormal termination — no recovery, no address rounding permitted
No recovery or address rounding permitted for misaligned jumpsU256 acceleration interface not yet standardized; further investigation needed on addition/subtraction and bitwise acceleration
Further investigation needed on U256 acceleration interface
AI Disclaimer: Some content or metadata on EIPsInsight may be AI-inferred or automatically compiled. If you find any discrepancy, please contact us at dev@avarch.org.