Super excited for the Arc chain vision! In many ways it’s the improved spiritual next generation version of the suave geth TEE coprocessor we were working on at flashbots ~2 years ago!

Why does MEV spam persist on high-throughput chains? We study Base arbitrage and show that spam comes from two very different search styles: targeted searchers wait for known opportunities probabilistic searchers probe on-chain to discover them 🧵 collective.flashbots.net/t/t…

The Shape Rotatooors headed to @Princeton for the IC3 Blockchain Camp. They shared an overview of the program, their unique learning environment, and what they're working on:

Ethereum is the only network with 100% uptime

Updates since then: * Deepseek v4 is out. There *is* a 2-bit quant that can run within 90 GB ( huggingface.co/antirez/deeps… ), and it works, however it's only fast on Apple hardware (I've head ~35 tok/s). On AMD, it's ~7 tok/s. IMO actually taking the effort to properly support more than one hardware manufacturer is a great example of the difference between mere "decentralized AI" and genuine "CROPS AI". I hope we can become better at this. * github.com/vbuterin/messagin… also has alpha telegram support now. However, the path to adding your account is quite janky * github.com/Luce-Org/lucebox-… looks promising as a way to run "dense" models (eg. Qwen 27B) more efficiently. It's janky, but on my 5090 laptop it seems to be ~2x more tok/s than llama.cpp * VoxTerm (local AI recording, no third-party servers) continues to be developed github.com/dmarzzz/VoxTerm And there's a lot more projects coming on the horizon. One other thing that has been on my mind is that there's actually a lot of intersection between "CROPS ethereum access layer" and "CROPS AI". For example, we want a ZK way to make (paid) calls to remote LLMs. But if we have this, then it's just as useful for solving another problem: private RPC reads in Ethereum. Another example: application-specific finetuned LLMs. Leanstral ( mistral.ai/news/leanstral ; I get ~38 tok/s on AMD) fits into < 70 GB, but can hold its own against 1T models on writing Lean code. Things like this are a huge boon for writing more secure code ( vitalik.eth.limo/general/202… ). We should have models finetuned for Ethereum-related use cases as well.

Our paper on timing games in backrunning transaction sending, arxiv.org/pdf/2602.22032, co-authored with @0xBrMazoRoig and Christoph Schlegel (Flashbots), has been accepted at ACM EC'26! In this paper, we look into equilibrium strategies of capturing backrunning opportunities. A🧵

As of today, BuilderNet supports prioritised updates This means propAMMs and other designs that want to prioritise some calls over others are now feasible This Titan's announcement earlier this week mark the beginning of a new chapter in transaction execution on Ethereum with a whole new set of challenges, opportunities and unknown unknowns (definitely tighter spreads) Making this new market efficient and reducing the need to rely on trusted intermediaries will require a lot of coordination across the ecosystem As a first step, we want landing updates to be permissionless. So we're launching v0 of a registry format we hope will become an industry standard - one where builders don't need to trust the prioritised updater. We've been collaborating with other builders and DeFi teams on this behind the scenes - post coming soon. Landing updates is just the beginning, we're cooking on several related features. Some of them are predictable, some of them are not - stay tuned. Docs in next tweet

We just shared a post explaining some TEE security research we’ve been contributing Every security system has its assumptions, with TEEs that assumption includes “the guy with the glock” to protect a TEE from physical attacks. That assumption hugely varies depending on who the physical custodian is. Trusting randoms on the internet is a recipe for disaster. That’s why we are working on designs & implementation of attestation systems that go beyond attesting the enclave, and attest other properties like *where* the TEE is running or what *other software* is running on the same chip but outside the enclave

1/ “Just put it in a TEE” sounds convincing, until you ask: where is this actually running? TEE attestations can prove what code is running and on which CPU. They don’t prove where that machine is running. That missing piece is the physical-access gap. 🧵

Meet 0x841: a fresh sandwich bot lurking in the dark forest--currently attacking about 1500 users per day, while going almost unnoticed. What's particularly scary: most of the victims are using private mempools and should be protected.

thank you all for your contributions <3