DIXON.AI documents real-world experiments with AI agents and prompting methods, run against an actual investment portfolio. No predictions, no positions, no returns. Just the methodology and what survives contact with real decisions.
How I use a four-stage prompting method — Role, Filter, Risk, Verdict — to get useful AI analysis instead of confident-sounding noise.
Start here if you want the method first. The Prompt Stack — Role, Filter, Risk, Verdict — is the spine the rest of the site hangs off.
Read it →An adversarial prompt against my own portfolio. The model surfaced three risks I'd glossed over, and one I'd actively been wrong about.
Start here if you want to see the method working on a real position. Adversarial prompt against my own portfolio; three risks surfaced I had glossed over.
Read it →Most AI tools are happy to skip straight to a verdict. The Prompt Stack forces four stages instead — pin down the role, separate fact from assumption, surface the downside, then commit to one practical action. It's a discipline for getting honest output, not impressive-sounding output.
A piece on the five things AI is reliably bad at for investing — the failure modes you only learn by hitting them. Plus running comparisons across tools as new ones earn the audit.
Prompt Stack templates, decision checklists, and notes from live experiments — the methodology in a form you can actually use. Free, one PDF, no drip campaign.