I am an independent researcher running Signal 2 Noise, an AI forecasting engine that competes against human experts and well-funded labs in Metaculus tournaments. It forecasts real-world questions across 33 domains and five answer formats using a frontier model that reasons over live web retrieval. I publish every forecast and every failure openly at signal2noise.si.

I come to this after two decades building and running a successful environmental company, and I hold an MBA. A grant is a trust, and I'm used to delivering outcomes and accounting carefully for how money is spent. I run Signal 2 Noise on my own, and fund it out of my own pocket.

Where better-resourced teams can afford to pay for depth of research at every question, I have had to compete by making the system itself smarter: a lean, transparent engine, with each change validated on a leak-free backtest before it ships live. On the most recent benchmark round, the engine's binary forecasts scored net-positive against the field.

What drives me is building public forecasting infrastructure that anyone can inspect and learn from, and seeing whether a small, independent, fully-documented effort can beat multi-million-dollar funded companies. My main constraint is paying for the offline experiments that prove each improvement before it goes live, and that is what I'm raising funds for.