About

Signal & Noise is a one-person blog about AI, code, and data - and what happens when you test the hype against real numbers. I take a question I can't put down, run a careful experiment, and publish what I find: walk-forward splits, bootstrap confidence intervals, a pre-registered hypothesis so I can't move the goalposts, and the full code and data so anyone can reproduce or falsify it. Sometimes the answer is exciting. Usually it's a clean no - and I publish that too.

The first deep series asks whether you can beat a prediction market (the spoiler is in the data, not from me). More to come on AI, coding, and whatever else is interesting enough to keep me up too late.

No tips, no hype, no "I made $X" screenshots, no real money traded - paper research only, and nothing here is financial advice. Just careful work, flat opinions, and dead-ends left in.