I think in systems, but I ship in TypeScript.
Agent Engineer · Building infrastructure for the agentic era
I build agent systems that actually work in production — not demos.
My focus: reliable execution loops, silent failure diagnosis, and agent-native infrastructure.
Currently deep in the z0 ecosystem and thinking about Agentic Primitives as the foundational layer of the next computing wave.
| Project | Description | Stack |
|---|---|---|
| z0 | Open-source Code Agent platform. Plan → Execute → Feedback, isolated Git branches per SubAgent (GitTree Model), checkpoint recovery. | TS · Next.js · Vercel |
| vite-browser | Vite runtime observation layer for agent workflows. HMR correlation, store-to-render tracing, mapped error diagnosis — no plugin required. The observe leg of an autonomous repair pipeline. |
TS · Node.js |
| loome | AI-generated creative video & visual effects. The project I care about most — where I go when I need to remember why I build things. | React · Tauri · Rust |
| CandyBomb | Web3 educational game. Minesweeper-style risk/reward mechanics on Sepolia testnet, AI-assisted gameplay. | Next.js · Foundry · wagmi · Vercel AI SDK |
Primary
Performance / Systems
AI / Agents
Infra / Data
- Agentic Primitives — the AWS-equivalent infra layer for the AI era
- Agent silent failure diagnosis & observability
observe → patch → apply → verify— autonomous repair pipelines- Sandbox abstraction as unified execution primitive (vite-browser · z0 · Rust CLI)
- MCP / A2A protocol convergence