English · 中文
Equity derivatives trader turned crypto product manager. I've spent my career pricing and trading complex derivatives — swaps, notes, options and structured products on equities, rates and FX — and then moved into crypto, where I built structured products, portfolio margin, perpetual-futures risk engines, prediction markets and US-equity / RWA contracts.
These days I use AI to turn my own trading ideas into tools that actually run. I don't write production code — I direct coding agents, and everything below was built that way.
🔗 LinkedIn · 🐦 X · 📕 小红书 · ✉️ ymcharvey@gmail.com
📈 Running live on Arcus
arcus-v2-farmer — a dollar-neutral memory-basket spread that farms the Arcus point program on open interest instead of volume, with its leverage schedule, stop ladder and ADL reset, plus the research and the OI-time scoring model behind it.
Want to trade it? Start here → app.arcus.xyz/ref/SKHYGO (referral link — your fee tier is unchanged)
The ones that get used the most.
⚖️ onchain-arbitrage - The engine behind my own live book: on-chain equity tokens against perpetual futures, both legs priced by real VWAP at size, behind a data-quality gate that refuses to trade on a bad quote
📈 arcus-v2-farmer - Running live money on Arcus: a dollar-neutral basket that earns open interest instead of volume — plus the research, the risk rails and the points model that produced it
🌀 alphaloop (20 stars) - A human-agent investment research loop: you set the direction, the agent gathers the evidence, screens names and keeps the books — the human keeps the call
📚 OpenOrder - Turns every agent conversation into a permanent, compounding investment-research wiki. Inspired by Karpathy's LLM Wiki
🔌 trading-agent-mcp - MCP server for multi-market financial research: IDE-driven bull / bear debate at zero LLM cost
✅ claim-verification - Breaks tweets, news, research notes and other AIs' conclusions into atomic claims, checks them against primary sources, and labels what survives
🌐 stock-data-fetch - Vendor-neutral market-data skill with automatic failover: US stocks, A-shares, HK, crypto and commodities
📡 crypto-social-radar - Tri-track heat radar for any token (X + news + Binance futures), pushed to Lark every day
Skills, wikis and agents for doing investment research with a machine instead of behind one.
🌀 alphaloop - Direction from the human, verification / screening / bookkeeping from the agent, decisions stay human
📚 OpenOrder - An investment-research wiki any agent can read and write, so conclusions and mistakes compound across sessions instead of resetting
✅ claim-verification - Breaks tweets, news, research notes and other AIs' conclusions into atomic claims, checks them against primary sources, and labels what survives
🌐 stock-data-fetch - One data interface for every market, with multi-source failover so it keeps working on restricted networks
🔌 trading-agent-mcp - MCP server for multi-market financial research: IDE-driven bull / bear debate at zero LLM cost
📡 crypto-research-radar - Daily research radar across perp DEXs, exchanges, prediction markets, AI × crypto and DeFi
🧊 trading-discipline-coach - An AI coach that will refuse you: a trading-discipline skill for Cursor
Everything that wakes up on a schedule and tells me something.
📡 crypto-social-radar - Tri-track heat radar for tokens — X + news + Binance futures, pushed to Lark daily
🛰️ hk-ipo-radar - Auto-scans Hong Kong IPOs, runs an 8-dimension AI deep dive, and pushes the result to a Lark group with a cloud-doc link
🛡️ edgex-risk-monitor - Perpetual-contract market monitor with Lark webhook alerts
🐋 edgex-flow-monitor - On-chain flow monitor for L2 whale trades and L1 withdrawals
💧 edge-lp-monitor - LP position monitor for Uniswap V3 (Arbitrum) and PancakeSwap V4 (BSC), alerting every 8 hours
📊 liquidity-depth-comparison - Cross-exchange perp liquidity-depth dashboard: live order-book depth across 10+ venues, synced to Lark Bitable
📝 captain-trading-booking - Turns trade screenshots or pasted text into structured records in Lark Bitable / Excel
🎯 okr-agent-skill - Turns quarterly objectives into agent-executable tasks
Built during my time at EdgeX, now maintained in the open.
🖥️ edgex-cli - CLI for EdgeX perpetual and equity contract trading, built to be consumed by agents
🔌 edgex-mcp - MCP server for EdgeX perpetual contract trading: market data, account management and order execution as agent tools
Turning long-form content and ideas into formats that survive the trip.
🎬 md-to-bilibili - Markdown in, Bilibili video out: AI voiceover, AI illustrations and Chinese subtitles, published automatically
📖 youtube-to-markdown - YouTube videos turned into clean, translated Markdown notes, from human subs, auto subs or Whisper
🎨 frontend-slides-to-pdf (fork) - Turns frontend-native HTML slide decks into stable, high-resolution PDFs and stitched preview images