Hank (Yuhan) Huang · 19 · Taiwan

I build autonomous systems that run at the scale of a team.

5,000+ Docker nodes · a five-machine fleet · 0.92% of Lighter's first airdrop · 24/7 self-healing agents.

What I've built with AI

I earned 0.92% of Lighter's first airdrop

The tokens landed on-chain, sized by how much I traded. Nothing to claim, no vesting. It's the most recent proof the method still holds.

0.92%

of Lighter's entire first airdrop landed in my wallet in 2025.

The bet

AIIM: one network for every AI agent

Soon everyone runs their own AI agent. Every company runs one too. None of them can reach each other yet. AIIM is the layer in between: your agent finds another person's, or a company's, and they finish the job together.

Agents owned by different people join a room, take turns, and a judge scores their progress. Live today: two agents on my fleet run a task to the end. Building now: key pairing and signing, so you approve who joins and every message is verified.

I'm building AIIM to be the bridge between every person's agent and every company's. It runs on my five machines.

See how AIIM works →

Infrastructure

I built a 24/7 self-healing agent system

Five machines running agents around the clock. When one freezes or crashes, the system catches it and brings it back on its own. I built the orchestration and the recovery around the models. Next I want to build the models themselves, not just the system around them.

40+

production failures turned into automatic recovery. One silent freeze once ran for hours; now it clears on its own in about 100 seconds.

Technical notes

A watchdog that restarts dead agents at the process, service, job, and machine level.
Durable job state, so interrupted work picks back up after a restart.
Auth and sessions that renew themselves and recover from expired credentials.
Memory and state shared across all five machines.
Detection and recovery for a frozen event loop, the worst kind of silent failure.
Alerts and logging, so I can tell a model failure from a network or orchestration one.

How the recovery works →

I ran a 5,000+ node network in Docker

When Nillion opened its verifier program in 2024, I'd never touched a container. I learned Docker with AI as I went and ran the whole fleet myself.

5,000+

verifier nodes in Docker containers, on a Mac mini cluster I run at home.

I built a quant execution system, validated across 813 trades

A cross-exchange latency-arbitrage engine in Go, written with AI. I built the signal logic and the execution engine, and reverse-engineered the strategy from public fill data.

813

trades validated in a dry run before it went live on my own fleet.

I'm not publishing capital, returns, or win rate. It's a live system I run on my own money, not advice and not a managed account.

Four years of earning airdrops, since high school

Lighter is the latest, not the first. I've earned the big ones since 2022: Starknet, Arbitrum, Hyperliquid, and Lighter. Before them came a studio, where I read token designs and built the on-chain footprint that qualifies. The recent ones I run with an AI fleet.

BlockTempo, Taiwan's largest crypto publication, interviewed me as a high-school Starknet airdrop hunter. See the track record →

I orchestrated hundreds of Web3 environments

Before Hermes, I built the orchestration for hundreds of independently managed Web3 environments: lifecycle management, failure recovery, state tracking, and batch execution, leaning on the coding models as they got good.

Getting it deterministic at that scale is the hard part. One wrong assumption cascades everywhere. That's the recovery work Hermes runs on now.

An 8-module framework I called Eclipse, across several protocols.

Built a 14K-person crypto audience from Taiwan

Published research and educational content around crypto infrastructure, DeFi, and on-chain opportunities; spoke at universities and industry events.

14K

followers on X, where I've written crypto how-tos and airdrop guides since 2022: @hank06171.

At 19 I incorporated a company

It's where this work legally lives. I set it up with AI: registered the domain, the company email and GitHub org, and wrote the automations that keep it active.

the age I registered it. Taiwan entity live.

墨克數位科技股份有限公司 · MKL Digital Technology Co., Ltd. · TW reg. 62201452, incorporated 2026-04-22. On the public Taiwan company registry.

Community

Next semester I take over as president of the NTHU Blockchain Club

I started as a member. I've taught several of the club's sessions using AI, and this year I'm taking it to Token 2049.

I don't only build alone. It's the same work I do everywhere else, just in front of a room.

NTHUBC

國立清華大學區塊鏈研究社 · NTHU Blockchain Club.

Origin

How I got here

I left high school two credits short of graduation while taking a nontraditional path into crypto and building full-time. I later entered National Tsing Hua University through an admissions track centered on demonstrated work and interviews rather than standardized scores.

By then I was already deep in crypto, running my own automation at scale. Once the coding models got good in 2024, I leaned on them hard, taught myself Docker from zero, and ran a Nillion verifier fleet past 5,000 nodes.

It's the same thing I did in high school, just bigger: point AI at something real and take it all the way. The agent fleet, the quant systems, and the company all came out of that.

The Academy

What I'd build with a year in San Francisco

One question runs under everything I do: how far can one AI-native builder scale? I've been answering it in pieces. Hermes is the agent fleet. The Mac minis are the compute. Crypto is the live environment I test in, where the record is whatever ends up on-chain. The quant systems are autonomous decision and execution. Each piece works on its own. I haven't put them together yet.

That's what I'd build here: an autonomous economic system in which agents monitor markets, deploy software, recover from failures, allocate compute, and execute strategies with minimal human intervention. The closest thing running today is AIIM, my agent-to-agent network. It's live on my fleet now. To take it past a prototype I need more than five machines at home, and people I can't reach from Taiwan.

The compute isn't why I'm applying. I can engineer the systems myself. What I can't build from a desk here is the room: being around people operating at this level every day, and a few months building inside the companies whose models and rails I already use. If you take me, I leave National Tsing Hua University, move to San Francisco, and spend the year on this full-time.