Build two: The AI Bubble Question, fifteen pages of research on the edge stack

The second build is live, and it's a different kind of stress test: aibubblequestion.com, a public research project asking whether the 2023 to 2026 AI investment boom is a bubble, measured against 200 years of documented manias, from the British Railway Mania of the 1840s to the dot-com crash.

What it is

Fifteen interlinked pages of long-form research: historical case studies, an explainer on circular financing, a warning-lights scorecard, a glossary, a full sources page, and a methodology section. Every figure links to its original source. Uncertainty is kept as ranges instead of being flattened into fake precision. Contested numbers are flagged as contested.

If The Lingerie Shoppe proved the stack handles scale, this one proves it handles density: heavy internal linking, inline glossary definitions, footnoted claims, and content structured so that both a human reader and an AI answer engine can extract the facts cleanly.

Why it matters for the methodology

A directory and a research site are opposite ends of the content spectrum, and the same stack built both: Astro for static generation, EmDash for editing, Cloudflare's edge for delivery. Same result, too. Pages that load instantly, hosting that costs almost nothing, and no plugin stack waiting for a bad Tuesday.

The site is also a live demonstration of the answer-engine work we talk about: structured data, citation-ready content, plain-language definitions, and stated sources. When we say pages can be built to show up in search and in AI answers, this is what that construction looks like from the inside.

What we're measuring

Same scoreboard as every build in this program: real-world load times, search impressions as the site earns its presence, whether and where AI assistants begin citing it, and the monthly cost of keeping fifteen research pages live (spoiler: it rounds to lunch). Follow-up notes will publish the numbers as they accumulate, sourced and dated.