| title: DeepQuant | |
| emoji: 📈 | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: gradio | |
| sdk_version: 5.50.0 | |
| app_file: app.py | |
| pinned: false | |
| short_description: The AI-managed fund - researched, run and governed by AI | |
| # DeepQuant | |
| **The AI-managed fund** — its strategy was *discovered by AI research agents* through 150+ walk-forward-validated | |
| strategy configurations over 21 years of survivorship-free data, and it is *run and governed by AI* end-to-end on a | |
| live Alpaca **paper** account with no human in the loop. | |
| - **The secret recipe:** ~60 U.S. large caps selected monthly by a proprietary AI-researched quality signal, plus a | |
| crisis-only defense (40% into bonds + gold, only in confirmed severe bears). The exact signal is undisclosed; | |
| every trade and result is published live. | |
| - **AI governance:** a weekly AI investment committee (Analyst → adversarial Critic → PM on Claude Opus 5) reviews | |
| the fund against its own 20-year envelope; changes require walk-forward proof via a code validator. Failure → HOLD. | |
| - **Deterministic where it must be:** execution, risk rails (long-only, 10% cap, no leverage), reconciliation. | |
| - **MarketGuruLLM** narrates every move in plain English; daily AI self-diagnosis is published openly. | |
| Strategy v2, live since September 2026 (v1's full record is archived). Simulated research results, paper trading — | |
| not investment advice. | |