How the Agent Works
A five-layer autonomous system that monitors markets, scores stocks, and manages a dual-thesis portfolio around the agentic AI revolution.
Data Ingestion
Every day, the agent pulls news and market data from multiple sources, filtering for agentic AI-specific signals: autonomous agents, job displacement, GPU infrastructure, compute demand, AI platforms, and disruption indicators.
Scoring Engine
Claude AI batch-scores all 70 watchlist stocks on a 0–100 conviction scale with sector-specific context. For bull-thesis stocks, higher scores mean stronger evidence of agentic AI tailwinds. For bear-thesis stocks, higher scores mean stronger evidence of disruption. Each score includes reasoning and is stored historically to track sentiment trends.
Dual-Strategy Portfolio
The agent manages two parallel strategies from the same conviction scores:
Long-only, $50B+ market cap, 10–20 positions. Bull-thesis stocks only. Managed via the Autopilot platform.
10-20 positions allocated purely by conviction score. No fixed long/short split — highest conviction wins. Full watchlist eligible including all market caps.
Bear-thesis stocks are scored and tracked now. Short positions activate once the platform supports short selling.
- Monthly rebalance: Recalculates all position sizes via quadratic conviction curve.
- Daily monitor: Watches for 20+ point score swings. Urgent signals queue for immediate review.
Audit Trail
Every decision is logged: the triggering article, score before and after, recommended action, and approval status. This public site shows holdings, agent activity, and score history — full transparency.
Human Approval
The agent recommends — the human decides. Every trade recommendation surfaces as an approve/reject decision.
Technology
Current Status
Disclaimer
This is an experimental, AI-managed thematic fund. It is not registered as an investment adviser or broker-dealer. Nothing on this site constitutes financial advice. Past performance does not guarantee future results. The AI agent's recommendations are based on news sentiment analysis and may not reflect fundamental company valuations.