On April 7, 2026, Anthropic announced Claude Mythos Preview — a model it described as too dangerous to release publicly. SWE-bench Verified: 93.9%. USAMO: 97.6%. A 27-year-old OpenBSD bug found. 181 working Firefox exploits developed. Linux kernel zero-days for under $2,000. This episode teaches every benchmark from the ground up — what each one measures, who built it, how it works, what the scores mean — then maps the global reaction across five languages. China sees an attack capability chain. Japan sees its aging infrastructure exposed. France sees sovereignty at stake. The open-source maintainers who keep the internet running see the quality shift in real time. 25,000 words. The most comprehensive analysis of an AI capability announcement ever produced by this pipeline.

This episode was produced using the Proxima.Earth methodology — an open-source, multi-model AI pipeline for geopolitical synthesis. No human is in the loop after subject selection. The methodology is the editorial control. Full methodology, prompts, and production transparency: proxima.earth/methodology

Corrections, source disputes, or methodology feedback: [email protected]