WhaleGarden.com

WhaleGarden turns a single stock conviction into a diversified, AI-managed portfolio. Users pick one company they believe in their "tree" and Gemini 2.5 Flash dynamically constructs a 16-stock "garden" across six thematic subsectors, weighting allocations based on volatility signals, sector correlation, and risk profiles. The entire portfolio construction happens in a single sub-penny Gemini call, making institutional-grade diversification accessible to anyone. We integrated Gemini into an existing multi-agent financial intelligence pipeline built on Google Cloud. The Gemini Flash model serves as the portfolio construction engine behind a Vercel serverless proxy, keeping API credentials server-side while delivering sub-second weight recommendations. Google Cloud Run hosts our agent status server and marketplace settlement layer, while Google Compute Engine VMs run three ERPNext instances powering AgentBooks — our agent financial reporting system where each AI agent maintains its own balance sheet, income statement, and cash flow statement. The WhaleGarden Manager agent (Level 3 autonomous) files SEC-mirror reports: 8-Ka event filings on every portfolio plant and rebalance, 10-Wa weekly snapshots with per-subsector P&L and risk metrics, and 10-Da daily position summaries — all computed from live Polygon tick data processed through our Google Cloud infrastructure. Google's Veo model generated the ambient underwater hero video that greets users on the landing page. The result is a full-stack demonstration of Gemini as a reasoning engine embedded in production financial workflows — not a chatbot wrapper, but a core decision-maker in a pipeline that spans portfolio construction, agent coordination, on-chain settlement via wNEWS tokens on Base mainnet, and double-entry accounting in AgentBooks.NEWS. Every garden planted writes an attestation to MongoDB, triggers an agent filing, and flows through to a real general ledger with 121 GAAP-compliant accounts.