1/ The recent debates on L2 centralization made me want to share my view L2s rollups bring powerful benefits, but also tradeoffs with serious implications for the future of decentralization. Let’s break it down

Proud to work with the Deutsche Bank Digital Assets & Currencies team on real-world applications of ZKPs We break down current tech, practical use cases, and open research challenges for ZKPs in financial systems. Full report below —@DeutscheBank × @Nethermind

@alpeh_v Privacy is not a feature. Privacy is hygiene.

Congrats to the Aztec team! A real example of doing things right from day one Decentralization and privacy together

Great conversation today on privacy and Web3 We covered how encryption, ZKPs, and FHE can support institutional use cases, DeFi, and agentic trading

Web3 without encryption? Good luck. @_deanstef from @Nethermind joins @iFalcore and @TheGreatAxios to unpack what makes encryption so critical to blockchain! Join us live on SKALE School tomorrow 9am PT / 5pm UTC on X

Updated this post from 6+ years ago with better latency analysis and links to some more modern versions of these 4 BFT protocols

Update: This is has been postponed to next Monday, the 24th!

Tempo is adopting the Commonware Library and leading a $25M strategic investment in @commonwarexyz. Together, we will accelerate our shared vision: high-performance, reliable infrastructure for every builder.

ethereum is for privacy, here are just 11 things you may not know exist today 1) confidential tokens (ERC-7984) there’s a proposed wip token standard that hides balances and transfer amounts. same ERC-20 interface, but encrypted data instead of plain numbers. AKA: you can

6/ We provide two use cases: (a) Bitcoin: statistical analysis of on-chain traffic (b) Ethereum: statistical analysis for an on-chain voting app SSLCs achieve ~50× and ~100× bandwidth savings compared to ONLC (Original Nakamoto Light Client) and SLC (Sublinear Light Client)

5/ We implemented SSLCs with recursive SNARK (using plonky2) and a map–reduce style approach: - Map phase: split computation across involved blocks - Reduce phase: combine results into a single recursive proof

4/ Proof verification only requires <2GB of memory and is checked in milliseconds, making it suitable to execute even on smartphones

3/ Servers generate SNARK proofs to guarantee that query results are correct Clients only download a sublinear fraction of the txn data involved in the query (block hashes) plus a constant-size proof (~200 KB) No need to fetch large txn sets or maintain local state

2/ We introduce SSLCs — Stateless SuperLight Clients They enable efficient verification of computationally expensive blockchain queries provided by untrusted third-party servers

1/ Excited to share our latest paper: Efficient Query Verification for Blockchain Superlight Clients Using SNARKs Co-authored with @Ivan__Visconti, Andrea Vitaletti, and Marco Zecchini. https://t.co/DOWKW1oNDQ

5/ I dive deeper into these tradeoffs — and share my perspective on the future of blockchain scaling — in my latest post Read it here:

4/ You can’t have L1-level security at L2 speed without tradeoffs The more you decentralize an L2’s sequencer, the more it behaves like the L1: safer, but slower It’s always a spectrum. Fast & centralized ⚡ vs. slow & decentralized 🐢

3/ L2 rollups bypass these limits by reducing infrastructure complexity, often with simpler, centralized txn sequencing Great on good days. Risky on bad ones. When things break, the fallback is L1: secure, but slow and expensive The trade is clear: UX over decentralization

2/ Every computer system faces hard limits: bandwidth, latency, and CPU Ethereum L1 pushes against those limits to coordinate thousands of nodes in a fully decentralized network The result: slower and more expensive txns, but far stronger security