How we source

This wiki covers a contested subject, so it has to be checkable. Every factual sentence links to its source, and every source sits in one bibliography with a DOI or a stable link.

Three kinds of claim

Label	Means	Example
Evidence	Measured: official data, a randomised trial, a peer-reviewed study	Alaska’s dividend did not reduce employment
Forecast	A model or scenario about the future, however careful	AI will add at most 0.66% to productivity over ten years
Proposal	What someone argues we should do	Pay every citizen an automation dividend

Forecasts and proposals can be good arguments. They just aren’t facts yet, so they never share a label with evidence.

Which sources count

1. Official data: statistical agencies, central banks, Our World in Data.
2. Peer-reviewed research and preregistered trials.
3. Working papers from established series (NBER and similar), labelled as such.
4. Books and essays, cited for ideas and arguments, not for numbers they don’t source.
5. Commentary, including David Shapiro’s Substack, cited for what its author argues.

AI-generated research reports, including the deep-research PDFs in Shapiro’s public repository, are used only to find primary sources; they are never cited.

Images

Every image is public domain, CC0, CC BY or drawn by us, and the credits page lists each one’s source and licence. Memes are credited like any other source.

Corrections

Found an error or a better source? Open an issue or use “Edit page” at the bottom of any page. Corrections are made in the open, in the repository’s history.

How this site is made

The research, structure and editorial rules are Javier Ponz’s; AI coding agents write the code and help draft text, and every claim is checked against the source it cites before it ships. Writing and original graphics are CC BY 4.0.