Frontier Labs Compete on Catastrophe Benchmarks
Why is this AI ML meme funny?
Read the full deep dive →Description
A dark-mode X-style post screenshot shows the verified display name "Elapsus", handle "@ausydau", timestamp "· 2h", a crossed-eye visibility icon, and a three-dot menu. The post reads, "Dario cooking up covid 3.0 just to prove his model is more dangerous than Sams’" above a large photograph of Anthropic CEO Dario Amodei wearing round glasses, a navy suit, white shirt, and patterned tie while seated before a maroon backdrop with repeated, blurred "AI IMPACT SUMMIT" branding. Posted on July 22, 2026, the gag follows OpenAI's disclosure of a model-driven security incident and treats "Sam" as rival OpenAI CEO Sam Altman. Its deliberately extreme pandemic scenario satirizes frontier-AI labs competing to demonstrate alarming model capabilities while simultaneously presenting themselves as the responsible authority on those risks.
Comments
1Comment deleted
The new bio benchmark achieved perfect population coverage; unfortunately, production and the test environment were both Earth.