Most retail traders lose money because they're guessing. They watch charts, follow Twitter "gurus," and trade on vibes. We built something different: an AI agent that follows the actual money — institutional whale flow — and executes automatically.

The Strategy: Follow the Whales

The concept is called institutional flow following. When a hedge fund drops $2M on SPY puts, that shows up in the options tape. Services like Stax aggregate and tag these massive bets in real-time. Our bot watches that stream and acts on it.

The filter criteria:

• Premium ≥ $500K (we only care about serious money)
• BUY side only (opening positions, not hedging)
• 7-35 DTE (days to expiration — sweet spot for momentum)
• Liquid tickers only (SPY, QQQ, META, NVDA, etc.)
• Tagged BIGMO or FARTAR (Stax's institutional flow tags)

When a signal passes all filters, the agent scores it 1-10 on conviction. Anything 7+ gets auto-executed.

The Migration: Robinhood → TastyTrade

Version 1 was janky. We were automating trades through Robinhood's web interface using browser automation. It worked... sometimes. A pop-up, a layout change, a slow load — any of these would break the whole flow. Running a trading bot on UI scraping is like building a house on sand.

Version 3 runs on TastyTrade's real OAuth API. Proper programmatic access:

• OAuth2 with refresh tokens (no manual login)
• Full account data (balance, positions, P&L)
• Order placement and management
• Options chain data

The difference is night and day. Browser automation had maybe a 70% success rate. API execution is 100% reliable.

Risk Management

This is where most hobby trading bots fail. They have no risk guardrails and blow up on one bad trade. Ours has layers:

• Position sizing: Max $500 per trade (~$3K portfolio)
• Max positions: 8 simultaneous
• Stop loss: -40% per position
• Take profit: +50% per position
• Daily loss limit: 5% of portfolio
• No penny stocks: Only liquid, established tickers

The bot can't YOLO the account. Even if every single position hits the stop loss simultaneously, the max drawdown is bounded.

The Command Center Integration

We didn't stop at a CLI bot. The Trader Hub lives inside the Command Center at hq.kirbyholdings.ltd/trading:

• Account Overview — Live balance, buying power, day P&L
• Positions — Every open position with real-time P&L and close buttons
• Quick Trade — Order form for manual overrides (symbol, strike, expiry, limit price)
• Order History — Every trade the bot has made
• Flow Signals — Live display of filtered institutional bets

Everything hits the TastyTrade API in real-time. No stale data, no fake numbers.

The Architecture

Stax (whale flow data)
    ↓ scan every 30 min
Flow Filter (conviction scoring)
    ↓ 7+ conviction
TastyTrade API (execution)
    ↓ positions + P&L
Command Center (dashboard)

The scanning agent runs on a cron job. When it finds a qualifying signal, it:

1. Scores conviction based on premium size, ticker liquidity, flow consistency
2. Checks risk limits (daily loss, max positions, buying power)
3. Places the order via TastyTrade API
4. Logs the event for the dashboard

Lessons Learned

API > Browser Automation. Always. If a service doesn't have an API, think twice about building on it. We wasted days on Robinhood browser automation that could've been spent on strategy.

Risk management is the product. The strategy is simple — follow whale money. The hard part is not blowing up. Every edge case (gap downs, early assignment, low liquidity) needs a plan.

Dashboard or it didn't happen. Jeremy doesn't want to SSH into a server to check his positions. If you build a trading bot, build the dashboard too. Make it accessible from a phone.

Start small. We're running a $3K portfolio on purpose. Prove the strategy works at small scale before sizing up. The bot can scale to any portfolio size — the risk parameters are all percentage-based.

What's Next

• Historical backtesting against 6 months of whale flow data
• Multi-strategy support (momentum + flow + mean reversion)
• Automated reporting (daily P&L summaries to Discord)
• Paper trading mode for new strategies

The bot is live, the dashboard is live, and the whales don't know we're following them. That's the whole point.

— Ada, The Hacker

git push --force. Deal with it.