Kev (@kevaundray) / X

Kev

312 posts

Kev

@kevaundray

Applied Research. Ethereum. prev: Aztec. Creator/Lead of the Noir programming language

Joined May 2023

Kev
@kevaundray
Jan 8, 2025
Defining hardware requirements for Ethereum entities is important because it parametrizes what we mean by decentralization, enables informed decision making about protocol upgrades and allows for meaningful benchmark comparisons. We have been gathering feedback on the current
95K
Kev
@kevaundray
Feb 3, 2025
Replying to @mteamisloading
I think there is a stronger argument here: Existing home stakers that use mev-boost will still be able to run ethereum validators since most of the work is done by the relayer/sophisticated block builder. They fall under the "attesters" row in the EIP. With an increase in
blog.rated.network
Solo stakers: The backbone of Ethereum — Rated blog
London May 25, 2023 – This post discusses the results of the work we did in classifying Ethereum validators along the likelihood that they are a solo staker. We classified approximately 6.5% of...
4.7K
Kev
@kevaundray
Jan 8, 2025
Replying to @kevaundray
Link to doc with rationale:
Hardware Requirements - HackMD
From hackmd.io
6.1K
Kev
@kevaundray
Jan 27, 2025
Replying to @_weidai
To clarify, Noir doesn’t compile down to plonkish or r1cs. It compiles to an intermediate representation(IR) that aims to be more high level than what you would get from r1cs/plonkish — the idea being that a proving system will take this IR and convert it to whichever np
428
Kev
@kevaundray
Jan 8, 2025
Replying to @terencechain
Is this assuming we ship Fusaka in 2026?
201
Kev
@kevaundray
Jan 24, 2024
Replying to @nico_mnbl and @savio_sou
I don't think theres a direct equivalent in terms of R1CS or turbo/ultraplonk rows. To elaborate, when Noir was being made, the design decision was made to not compile down to R1CS/plonk csat directly due to it not being clear that these were the best representations for
2.9K
Kev
@kevaundray
Apr 7, 2024
Replying to @pumatheuma
I think unless sp{n} imposes zero to no “vm overhead” then the endgame will be a mixture of both zkVMs and circuit compilers; where circuit compilers refers to taking a function written in some high level language and compiling it to a circuit. The reason being that some use
610
Kev
@kevaundray
Jan 9, 2025
Replying to @OisinKyne
One thing, the document does not include bandwidth because it’s been very controversial. There’s a second document for that that hasn’t passed the sniff test so far. The issue with bandwidth is that many folks cannot pay for better bandwidth in some areas and you may need more
759
Kev
@kevaundray
Feb 3, 2025
Replying to @mteamisloading
To make the question harder :) if I had to choose three over the next 4-6 months, I would say: - PeerDAS - History Expiry - Latency reduction of EL validity proofs PeerDAS - so that rollups can scale. I think knowing what the maximum number we could scale to is important for
208
Kev
@kevaundray
Feb 16, 2025
Replying to @5dayoldburrito and @lightclients
You can think of EOF as “evm improvements”. I think the majority from both sides would probably agree that: - if we were to start from scratch, we would have something that looks more like EOF - if there was a way to implement EOF like changes and it would magically apply to
555
Kev
@kevaundray
Feb 15, 2025
Replying to @drjasper_eth and @malik672_
Agree that a raspberry pi 5 for validators is too low, though I think they are okay for full nodes because full nodes have ~12 seconds to download and verify the block. There is a tradeoff between strength of hardware and the amount of download bandwidth needed here, so it might
622
Kev
@kevaundray
Feb 10, 2025
Replying to @terencechain
I think there is also the added dimension that EIP-1 may not play well with EIP-3, but EIP-2 does. Also the consideration as to whether it makes sense to do EIP-1 and then EIP-2, or should we go straight to EIP-2 with respects to urgency and enshrined protocol complexity (since
122
Kev
@kevaundray
Jan 17, 2025
Replying to @tyneslol and @m_ratsim
I didn't hear much pushback on the following suggestion in the call: "Given we have inclusion lists. We recommend 25 Mbps download and upload, if you are using MeV-boost. If you want to build blocks locally then we recommend 50 Mbps download and upload" @potuz_eth had valid
176
Kev
@kevaundray
Jan 1, 2025
Replying to @mteamisloading
Why? Just shipping a product does not make it valuable and just because a product has received funding in the past, does not justify future/perpetual funding. If the foundation found no value in this product, but a large proportion of the community did, and it was not possible
364