Open-source tools and research in LLM-agent systems, developer tooling, and compilers.
A localized Aptos view-function simulator and CLI that executes Move view functions across mainnet, testnet, and devnet without requiring explicit entry-function declarations — giving blockchain developers fast local read-simulation of on-chain state.
An open-source library of out-of-the-box, composable agent “skills” targeting agent harnesses like Claude Code and OpenClaw. Harness users drop in ready-made capabilities instead of authoring them from scratch.
An agent-driven release-intelligence framework that tracks repository deltas and synthesizes contextually relevant, human-readable release notes automatically — bridging raw git history and meaningful changelogs without manual curation.
A background JIT compiler daemon for autonomous LLM agents. It analyzes agent execution traces and compiles stochastic, multi-step reasoning paths into deterministic, zero-token shell scripts — drastically cutting both latency and per-call API cost.
Upstream contributions microsoft/Trill tikv/tikv elastic/kibana
