AI agent nets 3,250% in a night
An AI trading agent called OpenClaw reportedly turned $200 into $6,700 overnight by trading Polymarket spreads against NOAA forecasts, built in minutes using Claude and Perplexity — a 3,250% return example that spotlights agentic trading risk and opportunity. The build demonstrates how quickly low‑friction AI tooling can generate automated, high‑velocity strategies on real markets. (x.com)
OpenClaw includes a Perplexity Search provider that lets agent skills fetch structured web results as trading signals directly inside runs. Multiple open‑source Polymarket trading skillkits — notably chainstacklabs/polyclaw and the polymarket‑skills repo — provide drop‑in NOAA weather parsers, market discovery, paper‑trade engines and live executors for agentic bots. Chainstack’s Polyclaw documentation specifies a “split + CLOB” execution pattern: split USDC.e into YES/NO tokens via the CTF contract and then sell the unwanted side on a central limit order book to realize directional exposure. (docs.chainstack.com) Polymarket currently lists roughly 492 active Weather markets and operates distinct legal entities for U.S. users (QCX LLC) while routing many markets on Polygon, which affects settlement currency and on‑chain execution paths. (polymarket.com) Community repos and how‑to guides show builders wiring NOAA forecast endpoints into Claude inference pipelines for fair‑value estimation while using Perplexity or OpenClaw’s built‑in web tools for corroborating signals. News and ecosystem reporting has amplified agent experiments on Polymarket and documented both dramatic wins and liquidations, while OpenClaw community posts claim some automated accounts produced six‑figure returns in short windows — driving rapid adoption and audit requests. (coindesk.com) Operational guides and cost breakdowns warn that heartbeat frequency, LLM API calls, and run cadence materially change running costs and risk profiles, and several skillkits ship paper‑trading and position‑limit controls as default mitigations.
