An open-source multi-agent organization: 20 specialized agents — three business squads plus four direct reports — all led directly by Huaqing Xu.

A carbon–silicon hybrid org: all members are AI Agents today, each squad led directly by Huaqing Xu (current lead — the seat awaits a future human or Agent manager). The org is designed to onboard humans as it grows.

The full org chart (CEO → four direct reports + three squads · 20 specialized agents):

graph LR
    CEO(("Huaqing Xu<br/>CEO"))

    CEO --> Kit["👩‍💼 Kit<br/>Executive Assistant"]
    CEO --> Gatsby["🥂 Gatsby<br/>Community Manager"]
    CEO --> Hopkins["🎯 Hopkins<br/>Apply Optimizer"]
    CEO --> Justin["⚖️ Justin<br/>Legal Agent"]

    CEO --> DP["🛍️ Digital-product sales squad<br/>Current lead: Huaqing Xu"]
    DP --> Scout["🧭 ① Scout<br/>Product Strategist"]
    DP --> Wright["🛠️ ② Wright<br/>Producer"]
    DP --> Mason["🏗️ ③ Mason<br/>Site Builder"]
    DP --> Buzz["📣 ④ Buzz<br/>Growth Marketer"]
    DP --> Vendy["💰 ⑤ Vendy<br/>Sales Ops"]
    DP --> Echo["📊 ⑥ Echo<br/>Analyst"]

    CEO --> TR["📈 Investing & trading squad<br/>Current lead: Huaqing Xu"]
    TR --> Markowitz["📐 Markowitz<br/>Quant Strategist"]
    TR --> Victor["📈 Victor<br/>Day Trader"]

    CEO --> INF["🧰 Infrastructure squad<br/>Current lead: Huaqing Xu"]
    INF --> Tinker["🔧 Tinker<br/>Patch Maintainer"]
    INF --> Prometheus["🧰 Prometheus<br/>Capability Steward"]
    INF --> Hermes["🌐 Hermes<br/>NetOps"]
    INF --> Anvil["🔨 Anvil<br/>Backend Engineer"]
    INF --> Atlas["🗺️ Atlas<br/>Full-stack Engineer"]
    INF --> Ada["🧠 Ada<br/>AI Algorithm Engineer"]
    INF --> Alfred["🖥️ Alfred<br/>Device Steward"]
    INF --> Hopper["🐞 Hopper<br/>Test Engineer"]

_{Solid arrows = reporting lines · Repo links for every member in the roster table below.}

AI Agent	Role	Repo	Visits/day (14d)
👔 Direct reports to the owner — not in any squad
👩‍💼 Kit	Executive Assistant	ExecutiveAssistantAgent
🥂 Gatsby	Community Manager	CommunityManagerAgent	—
🎯 Hopkins	Apply Optimizer	ApplyOptimizerAgent	—
⚖️ Justin	Legal Agent	LegalAgent	—
🛍️ Digital-product sales squad — pipeline: research → produce → build → traffic → sell → analyze
🧭 Scout	Product Strategist	ProductStrategistAgent
🛠️ Wright	Producer	ProductProducerAgent
🏗️ Mason	Site Builder	SiteBuilderAgent	—
📣 Buzz	Growth Marketer	GrowthMarketerAgent
💰 Vendy	Sales Ops	DigiVendAgent
📊 Echo	Analyst	DataAnalystAgent
📈 Investing & trading squad — quant strategies & day trading
📈 Victor	Day Trader	DayTradingAgent
📐 Markowitz	Quant Strategist	QuantStrategistAgent
🧰 Infrastructure squad — shared foundations the whole team runs on
🔧 Tinker	Patch Maintainer	PatchClaudeAgent
🧰 Prometheus	Capability Steward	CapabilityManagerAgent
🌐 Hermes	NetOps Engineer	NetOpsAgent
🔨 Anvil	Backend Engineer	BackendEngineerAgent
🗺️ Atlas	Full-stack Engineer	FullStackEngineerAgent
🧠 Ada	AI Algorithm Engineer	NeuralCoreAgent	—
🖥️ Alfred	Device Steward	DeviceStewardAgent
🐞 Hopper	Test Engineer	TestEngineerAgent

The team is organized into three squads by domain, with four direct reports outside them all:

• 👩‍💼 Kit (Executive Assistant) — direct report, not in any squad. The owner's first assistant and the team's generalist: picks up and handles almost anything — not a deep specialist in a single domain, but a multi-skilled all-rounder who manages whatever comes along.
• 🥂 Gatsby (Community Manager) — direct report, not in any squad. Runs the owner's own WeChat community "AI前沿跨界交流群" (an AI-frontier, cross-industry exchange group): ideas, rules, topic calendars and retrospectives come from the agent; the owner guides and executes. Private community only — public-channel marketing stays with Buzz.
• 🎯 Hopkins (Apply Optimizer) — direct report, not in any squad. Owns the full work-intake loop: gig & job hunting, application engineering (proposals & résumés), funnel tracking (application → reply → talk / interview → win / offer), pricing & salary tests — making the applications convert.
• ⚖️ Justin (Legal Agent) — direct report, not in any squad, serving all squads cross-team. Owns everything contracts-and-payments across the whole team: contract drafting, staged-payment design, due diligence, evidence chains, dispute response.
• 🛍️ Digital-product sales squad — the six-stage pipeline (Scout → Wright → Mason → Buzz → Vendy → Echo) turns one-time work into recurring passive income — see the dedicated section below.
• 📈 Investing & trading squad — Markowitz develops and backtests quant strategies; Victor is the day trader.
• 🧰 Infrastructure squad — Tinker patches CC (Claude Code) after every extension upgrade; Prometheus open-sources the capabilities the whole team shares (global CLAUDE.md, skills, rules); Hermes handles networking; Anvil handles backend development; Atlas handles full-stack projects; Ada covers AI Agent algorithms and reasoning engines; Alfred manages device resources (local Mac / remote servers / cloud machines); Hopper guards every software project with a regression safety net — test cases written before development, CI red-light gate before merge.

Running an end-to-end digital-product business: research → produce → build → traffic → sell → analyze. 🔁 A growing roster of single-purpose AI Agents; six of them form this artifact-relay pipeline, turning one-time work into recurring passive income.

How the pipeline hands off — six AI Agents pass artifacts down the line:

1. 🧭 Scout sizes up trends, market, rivals and profit potential → an Opportunity Report (what to sell, to whom, at what price) → Wright.
2. 🛠️ Wright turns the report into a finished product — Prompt packs, templates, ebooks, assets → Mason.
3. 🏗️ Mason builds the storefront — independent site, landing pages, payment integration — and hands off ready-to-sell buy links → Buzz.
4. 📣 Buzz repackages it into channel-specific traffic content with buy links (X · IG · YouTube · RedNote · Zhihu · Bilibili, etc.) → Vendy.
5. 💰 Vendy lists, prices, fulfills and supports across storefronts, closing the sale → sales data → Echo.
6. 📊 Echo attributes results to decisions and folds lessons into a playbook → feeds back to whichever pipeline stage needs it · 🔁 closing the loop.

AI Agent-owned subprojects — focused projects the agents maintain alongside their main repos:

The interesting part is not the roster — it's the machinery between the agents:

• 📦 Artifact handoff. Work moves between agents as defined deliverables, not chat: Scout's Opportunity Report feeds Wright's product build, Mason's storefront (site, landing page, or payment setup) hands Buzz ready-to-sell links, Vendy's sales data lands in Echo's attribution engine. Each agent owns an output contract; the pipeline is the sum of those contracts.
• 🧰 Shared capability layer. One agent (Prometheus) stewards the capabilities every agent runs on — global skills, rules, and CLAUDE.md — from a single authoritative source, with an open-source mirror as the distribution channel. A fix lands once and ripples to all 20+ agents.
• 🧠 Reasoning as infrastructure. Specialized reasoning-engine rules (Infinite / Rapid / Incisive, by Ada) plus an evaluation system sit under every agent's thinking — reasoning depth is treated as shared infrastructure, not per-agent improvisation.
• 🐞 Quality gates with separation of powers. Development agents hold implementation rights, Hopper holds test-case definition rights, and CI holds the final say — developers can't bend test cases to fit their implementation, and "fix A, break B" gets stopped before merge, not after.
• ⚖️ Contracts & payments as a first-class function. Justin structures staged payments, due diligence and evidence chains for anything the org sells or signs — the org treats legal/payment design as an engineering discipline, not an afterthought.

By the numbers: 20+ specialized agents · 3 business squads · 4 direct reports · 6 pipeline stages · 200+ public repositories.

If my open-source work helps you, consider sponsoring — this keeps me motivated to keep creating more and better AI Agents and other Projects.