Published the Roadmap & Q2 agenda: Q2 is focused on double-stateless schemes for the small but highly mutable slices of the Ethereum state, network-agnostic architecture and reference impl for anonymity networks access, integrations of TorJS, auditing of Arti and early exploration of formal verification (Lean).
Spec and reference implementation of VIA finished via grant, an issue in its parameterization was discovered in the process, commuicated to the authors.
Started drafting an architecutre for network-agnostic access, the goal being: the edge (wallets, frontends, ..) connect to any anonymity network and validate their client-side binaries in the same way, eliminating integration overhead and switching costs at the edge.
Ongoing 2x integrations of TorJS 12 into light clients
Explored ideas on PIR for Merkle roots with Divyá but none were compelling to pursue further. STARKifying away these roots seems to be the most promising approach long-term.
Considered and rejected a grant proposal on GPU acceleration of insPIRe scheme.
Current focus double-stateless schemes:
One lesson from exploring new double-stateless schemes is that the required vRAM for a preprocessed DB is an important performance dimension for deployment. For instance, OnionPIRv2 can only fit a 4GB DB on an RTX-5090 (32GB vRAM), since the memory blow-up is 4x and working memory is also needed.
Switched focus to working on speccing and GPU-acceleration of InsPIRe PIR scheme, with some algorithmic optimizations.
Discussions and scope definition for a grant with 2x GPU specialists.
The O(N) server overhead of double-stateless schemes is significant, but (a) batching queries in a single sweep of N records on the GPU amortizes the cost; (b) these schemes handle DB updates cheaply and without involving clients, which is essential for highly mutable data slices (e.g., contracts of popular tokens); and (c) the server does not manage client-specific key material, which makes deployment easier and supports unlinkability between queries.
Ongoing discussions with 2x collaborators in academia on the design and GPU acceleration of doubly-stateless PIR schemes.
static config replaces consensus for easy testnet and future replacement
enable relay mode
Research into prior art then started drafting an architecutre for network-agnostic access, the goal being: the edge (wallets, frontends, ..) connect to any anonymity network and validate their client-side binaries in the same way, eliminating integration overhead and switching costs at the edge.
Testnet planning, low-priority in Q2 in favor of the network-agnostic arch
Investigated webrtc bridge design for decentralized RPC