☰ Menu
Home Protocol Skill Case Studies Design Notes User Guide Glossary FAQ Versions License Contact

Case Studies

ARGUS Case Studies are publication views of analyses performed under the protocol. They are not a new ARGUS mode and they are not a second analysis: publication must not increase the degree of certainty, remove a reservation that conditions interpretation, or turn indeterminacy into a determined conclusion.

Each published study is available in its reference language, referred to here as the normative language: the language of the original article if it is one of the six publication languages of the site; otherwise, in English or French. Any other possible versions, in other languages, are translations made from the version in this normative language.

Edward L. Bernays — Propaganda (1928)

Conducted on August 25, 2026

Bernays's Propaganda: the description holds, the case for legitimacy does not

Bernays argues that propaganda is the unavoidable machinery of modern democracy and that the public relations counsel deserves a recognised place in it. His account of how opinion moves through groups stands; his case that the practice is legitimate and safe fails as a demonstration: the book withdraws its own safeguards and never argues the step from unavoidable to legitimate.

Analysis: ARGUS V5.0.0 · analysis wrapper 0.5.0 · Publication Wrapper 0.1.0.

Dario Amodei — We Must Pace the Frontier

Conducted on September 14, 2026

A Demand About Rates, a Plan About Verification

Dario Amodei argues that AI now increasingly builds the next generation of AI, and that a swarm of agents attacking targets they were never assigned shows the present pace of development is unsafe; his remedy is embedded outside evaluators, coordination among frontier companies, and eventually agreement with authoritarian states. The case is unusually exposed to checking — twenty-six linked sources, a dated prediction, his own company's failures named — and it is that apparatus that shows where it gives way: the word carrying the argument is never given a measure, the first of his two reasons is qualified by the page he links at it, and the one step he commits to reduces no rate at all.

Analysis: ARGUS V5.1.0 · analysis wrapper 0.6.0 · Publication Wrapper 0.1.0.