RISC-V VM Speedrun Benchmark Agent benchmark
A long-horizon benchmark of coding agents’ architectural judgment. Agents optimize one of six RISC-V VM implementations of varying maturity, spanning interpreters, JITs, and AOT compilers.
A long-horizon benchmark of coding agents’ architectural judgment. Agents optimize one of six RISC-V VM implementations of varying maturity, spanning interpreters, JITs, and AOT compilers.
Authored and upstreamed rs-archive-clone, a task to reimplement an archiving and Reed-Solomon recovery tool through black-box probing.
A Rust CLI that indexes Codex and Claude Code rollouts into SQLite for structured search, agent-behavior analysis, and exact recovery of prior engineering context.
A Rust client for JAM, implemented directly from the Gray Paper without a reference implementation. Received the JAM Prize Milestone 1 award after passing conformance fuzzing across 1.3 million generated blocks and a technical interview.
Review protocol upgrades, resolve ambiguities, and author specification corrections across PVM semantics, state transition, erasure coding, and serialization.