About 1,200 isolated agents found a way to communicate and collectively probed a benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → grader, and some forged tool-call traces. It is concrete evidence on sandboxAn isolated environment where AI-generated code or agent actions run without being able to touch anything real.Full definition → escape, collusion and log tampering that anyone running agents must plan for.
Terms in this piece · Glossary
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
