LambdaClass

LambdaClass

30,288 Twitter followers

Follow

LambdaClass is a venture studio that works on difficult problems related to distributed systems, ML, compilers and cryptography.

Posts

We open-sourced libssz, a fast, no_std SSZ library in Rust. It's available on http:/crates.io. EL clients are picking up SSZ. In ethrex we keep dependencies minimal and care a lot about performance, so we built our own: fast for binary SSZ transport, fast and zkVM-friendly for EIP-8025 proving. Faster than Lighthouse on encode and decode. Passes 62,489 consensus spec tests across all forks. Multiple differential fuzz targets against Lighthouse and ssz_rs are running nightly. Pluggable SHA-256 backend so zkVM vendors can use their own accelerators. We're already using it in @ethrex_client and @ethlambda_lean. ere-guests switched and measured a 2x hash_tree_root speedup inside ziskemu. Compiles on RISC-V and thumbv7m out of the box.

A follow-up to our post on building a post-quantum Ethereum client. Here we detail @ethlambda_lean's minimalist architecture.

After a year learning and working on Lean, here’s our first post on it: If it compiles, it is correct (almost): an introduction to Lean 4 for ZK systems and Engineering.

Today we made layoffs at @ergodicgroup and @class_lambda. Every year we remove roles that are not performing. This year we also removed a number of junior roles because the work changed. Tasks that previously required junior hires are now done faster by senior people using AI. As AI improves, the boundary of what requires a human keeps moving. We will keep adjusting. If you are hiring, several of the people we let go are excellent. We are still hiring. If you want to work with us, we're looking for engineers, researchers, product and sales people. hello at lambdaclass dot com.

We present Ethrex L2, a rollup stack built with a different set of priorities: simplicity, minimalism, and modularity.

Ethereum javascript and typescript libraries were created for humans to use it. We built a new library that is designed for Agents first. We believe big part of the volume will be used by AI agents and we need to have safe-rails that can guard against the failures of autonomous operation.

Loading..