Highlights

  • Sharded PIR design published. The sharded PIR design for the Ethereum state is now published on ethresear.ch.

  • New PIR scheme: LeanPIR. Keewoo is working on a new GPU-friendly scheme that could outperform insPIRe across all dimensions: ~30 ms server runtime for a 32 GB database, sub-second preprocessing (down from ~10 s), and <100 KB communication. It supersedes the earlier insPIRe GPU work and is being integrated into our Ethereum balance retrieval demo.

  • VIA spec approaching v1. Turan’s VIA spec implementation is progressing, covering VIA, VIA-B, and VIA-CB variants with reusable lattice primitives. The /primitives module is designed to be shared across all ReSPIRe-esque implementations.

  • Arti bootstrapping overhaul. A Brotli-compressed ~3 MB bundle now replaces fetching thousands of micro-descriptors — a fundamentally different and much faster approach than the standard Tor bootstrap method (albeit with that one-time fetch itself not onion-routed). Working demo at tor-js-gateway.voltrevo.com, docs at voltrevo.github.io/arti.

  • Ethrex collaboration for UBT. Kicked off a collaboration with Ethrex for client diversity on the UBT front, complementing ongoing work with Geth’s Stateless team.

  • Presenting at Stateless Summit on PIR and TorJS.

Details

PIR for the Ethereum State

  • Design post receiving feedback, targeting publication before April 1
  • LeanPIR, new PIR scheme— ~30 ms for 32 GB, sub-second preprocessing, <100 KB communication
  • Harmony integrated into the PIR benchmarks, h/t Ling Ren for sharing it
  • VIA spec v1 nearing completion
  • Internal PIR demo: sidecar pattern with GPU-accelerated insPIRe for real-time ETH balance queries (cold start data from Google BigQuery, 100–300 account updates simulated per block)
  • Sync with Ling Ren and group: agreement that doubly-stateless PIR is the right priority and that preprocessed schemes are best reserved for immutable archival data; we sniped them into VIA, they sniped us into Harmony

TorJS / Embedded Arti

  • Brotli-compressed ~3 MB bootstrap bundle replaces standard micro-descriptor fetching — fundamentally faster cold start
  • Giving JavaScript direct socket access (bypassing WebSocket) matches raw Rust Arti performance in Node.js
  • WASM compatibility code under review for upstream merge with Tor Project
  • Q2 scope shaping up: Ethereum-sandboxed anonymous network (WebRTC transport as first milestone), wallet SDK integrations, a closer look at messenger protocols/apps (+ potentially embedding Arti in some)

Verifiable UBT

  • Ethrex collaboration kicked off for client diversity
  • Geth integration waiting on Stateless team bug fixes and performance improvements
  • A UBT state checker utility

Looking Ahead to Q2

Q2 plan is being finalized (March 31). Key themes:

  1. PIR productionization — sharded design with LeanPIR, VIA, Harmony, OnionV2 as potential initial engines; GPU acceleration collaborations
  2. Arti integrations — Ethereum-sandboxed anonymous network, WebRTC transport, wallet SDK integrations
  3. UBT shadow chain — running provably MPT-equivalent shadow chain, overcoming UBT-MPT conversion performance blocker in Geth and accelerating Ethrex route
  4. Continuing some collaborations, and pushing some new ones forward (PIR academia, GPU researchers/practitioners)