I build useful things for strange, high-context corners of the internet.

Most of my work starts with the same problem: technology moves faster than the language, rules, and institutions around it. Sometimes that means teaching AI agents when they can move money. Sometimes it means translating a crypto meme without explaining the joke to death.

Previously Goldman PWM → OKX. Currently somewhere between agentic payments, onchain policy, and cross-cultural crypto infrastructure.

Agents can move money now. OmniClaw helps decide whether they should.

Pre-execution policy simulation · authorization traceability · real-time enforcement

1st place, Circle × Google Gemini hackathon · $20k · 1,000+ teams

An open-source Agent Skill for the gap between Chinese crypto and English Crypto Twitter.

It decodes slang, memes, tone, and cultural context in one direction. In the other, it helps Chinese writers publish natural English without quietly rewriting what they meant.

• when an AI agent spends money, who gets to define “allowed”?
• two wallets can look identical onchain and mean completely different things. how should systems tell?
• what gets lost when crypto ideas cross languages, timelines, and communities?
• can compliance become useful infrastructure instead of the final boss?

• SF Web3 Quest: 17+ Bay Area web3 companies mapped in Dragon Quest style
• HK Web3 Quest: same energy, different timezone
• US Crypto Reg Tracker: a weekly digest of stablecoins, market structure, and enforcement

clo = {
    "training_for": "usually something",
    "sports": ["triathlon", "skiing", "diving"],
    "default_state": "curious",
}