Peter Gabriel Neumann, the SRI computer scientist who warned for seventy years that institutional choices, not user error, drive most computer-systems failures, died May 17 of complications from a traumatic fall. He was ninety-three. [1]

The credentials would have made a normal career. Neumann was an architect of Multics, the 1960s operating system that taught Unix everything it knows about security. He founded the ACM Risks Forum in 1985 and edited every issue until last week. His 1995 book Computer-Related Risks read like a hospital morbidity-and-mortality conference for software — failure modes, named, dated, root-caused. [2]

The argument was older than the field. In a Multics-era memo, Neumann wrote that the question was never whether systems would fail, but whether the institutions building them had chosen to learn. He spent the next fifty years collecting the receipts: Therac-25, the 1990 AT&T crash, Heartbleed, Boeing 737 MAX, the OPM breach, the Vercel OAuth fix the paper has been counting for thirty days. [2]

He outlasted three generations of vendor optimism. His tone never changed. Patience, in his hands, was a form of accusation. [1]

The paper's Bundibugyo lead — three weeks of cartridge silence because a field lab tested for the wrong Ebola — is a Neumann problem with a virology label. He would have filed it under R for Risks. [3]

-- KENJI NAKAMURA, Tokyo