I’m running an AI Agent experiment: Can an AI agent be more profitable than me on low timeframe prediction markets? And can it repeat its success while self-learning to eliminate losses. I think so. So, I made Oshi, a self-learning prediction agent. Right now, she trades 15m BTC/SOL/ETH prediction markets on @Kalshi. Next, she will spin off another agent to trade @Polymarket on @BullpenFi. Then another agent to trade leverage... But profitability is king, so I am currently testing, refining, and upgrading strategy and execution. We started out weak, but it’s still early. Current results: v1: 36% hit rate (dogshit, still had random errors) v2: 86% hit rate (less dogshit so far, working and improving sizing strategies) v2.2: just started I'm going to keep testing small. v4 to v5 is where it should really learn and improve. May eventually share the dashboards publicly, if I can get this thing profitable. Stay tuned, I’ll keep you updated.