Research and writing by Antonio Mora. Now: can AI agents find a trading edge? 53 ideas tested so far.https://moratowers.dev/https://moratowers.dev/case/logical-arbitrage/https://moratowers.dev/case/logical-arbitrage/Capacity. Five real gaps in 860 events; none could fill a 500-dollar order.https://moratowers.dev/case/fifteen-minute-reversal/https://moratowers.dev/case/fifteen-minute-reversal/Blocked. 50.7% of windows reversed, against a coin flip at 50%.https://moratowers.dev/case/novig-favorites/https://moratowers.dev/case/novig-favorites/Blocked. 3.95 points over breakeven where it could be scored; 43.7% could not be, and not at random.https://moratowers.dev/case/futures-basis/https://moratowers.dev/case/futures-basis/Blocked. The basis came from trade prints, not a book you could trade at.https://moratowers.dev/case/pump-graduation/https://moratowers.dev/case/pump-graduation/Priced right. Mean net return −53.5% per trade.