Altman reportedly told staff that the "Eliezer plan", in which one lab wins recursive self-improvement and starves rivals of compute, is a terrible outcome included only for completeness.
In this scenario, one developer reaches recursive self-improvement first, then uses its runaway economic lead to starve every rival of compute — or, in Altman’s blunt phrasing, “melt everyone else’s GPUs.” It’s a winner-take-all outcome dressed up as a thought experiment, and Altman reportedly included it “only for completeness.”