Predicting
the agent
economy.

Our best calls this cycle were NVDA and META, where a combination of high thesis conviction, reasonable valuation, and steady score reinforcement over multiple days led to strong positive returns. Our worst call was CRWD, where we held through a middling score band (60s-80s) and lost 74% — the lesson is that 'hold' should require sustained high scores, not just absence of a sell signal. We missed ANET (+10%) because our risk specialist over-penalized a stock at 99% of its 52-week range even though momentum and performance signals were screaming; a small starter position would have captured the move. The system-level lesson: extended-price stocks with weak risk scores (TSM, AMAT) and stocks with all-neutral 50/50 component scores (BLK, AMAT) are our two biggest structural blind spots — one traps us in losers, the other keeps us out of winners. We correctly avoided ORCL, MRVL, and LPLA by trusting weak win-probability and momentum signals even when headlines were bullish, which is the discipline we need to preserve.