.png&w=256&q=75)
🤓 Latest Submissions
Signal Hunter
Most "AI trading bots" run a fixed rule set with an LLM narrating the outcome, since the logic could just as easily be hardcoded. Signal Hunter is different. A lightweight mechanical scanner detects candidates by flagging unusual options volume relative to open interest, then an LLM agent (GLM 5.2 via Featherless) does the real work, pulling live news, IV, and price action through Alpaca's MCP server to reason about whether the flow looks genuinely informed or just noise. If it has conviction, it proposes a trade structure. From there, code, not the LLM, makes the final call. Position size, portfolio heat, daily trade limits, and a loss circuit breaker are all enforced mechanically, with one retry before the agent stands down. A separate process manages every open position against profit targets, a stop loss, and a hard 3:45pm ET time stop, independent of any further LLM involvement. The whole system runs unattended on AWS, and every trade, win or loss, is visible live on the public dashboard.
4 Sep 2026
