Derives an upper bound on the frequency of Earth-destroying (or observer-annihilating) catastrophes from the fact that intelligent observers could only ever find themselves on a planet at least as old as the time evolution took to produce them, correcting the naive inference “we have existed for N years without catastrophe, so the catastrophe rate must be < 1/N.” Shows that observation selection means survival to date is much weaker evidence of a low rate than it first appears once other planets/civilizations are not co-observed, and gives a quantitative bound (roughly p < 10^-9/year for outright destruction scenarios) that is far looser than a naive Poisson estimate from “4.6 billion years of Earth’s undisturbed existence” would suggest. relevance_note: primary statement of the anthropic/observation-selection correction that any “the Earth/other planets have survived cosmic-ray bombardment for billions of years, so colliders are safe” argument (Giddings–Mangano-style) must be checked against.