Peter Gabriel Neumann died on May 17, 2026, in a hospital in Santa Clara, California. He was ninety-three years old. The people who loved him, and the newspapers, and the Association for Computing Machinery, all said the same thing: complications of a fall.

That was the whole report.

He had spent forty-one years teaching the rest of us what a report is for. A headline is where the file begins. The file is how a stranger stays alive. He had a mailing list for that. It was called the Forum on Risks to the Public in Computers and Related Systems. People called it comp.risks. He moderated it from 1985 until April of 2026, a few weeks before the floor got him.

The last issue went out on a Sunday, the twelfth of April. Volume 34, Number 91. Then he fell. Then he died. The digest did not carry the item.

Listen: he was a serious man. He spent a life trying to keep the new machines from killing the public by accident, which they are always on the verge of doing. He had breakfast with Albert Einstein on November 8, 1952, and what stuck was the line about making things as simple as possible, but no simpler. He helped build Multics. He spent decades at SRI trying to make operating systems a person could trust. He wrote a book called Computer-Related Risks, which is still true. The industry has given it plenty of new material. He said security was an end-to-end, total-system, emergent property. He said we wanted strength in depth and had weakness in depth. He was right.

He was also a mammal. Mammals fall down. After a certain age, falling down is how a great many of them die. In the United States, in 2024, more than forty-three thousand people over sixty-five died of preventable falls. One in four people that age fall every year. The death rate is climbing. Among people over eighty-five it is a slaughter, quiet, domestic, and almost never on the news.

There was no computer in the loop. So it was news of a kind he left uncollected.

* * *

He was born in New York on September 21, 1932. He died a computer scientist. In between he did the rare thing of looking at disasters on purpose. Airplane software. Radiation machines that gave ten times the dose because of a race condition. Rockets that converted a number from sixty-four bits to sixteen and then destroyed themselves. Hospitals, elections, telephone switches, robots that killed mechanics.

His method was simple, and he used it for forty-one years. He gave a risk a name, so people could see it. He studied the wreckage after a miss, so a miss became a lesson. He advised a practice, so the next person could catch the obvious in time. He cared most when the obvious could kill, and when a little ordinary care would have spared the body.

That is how grown civilizations treat wreckage. When an airliner comes down, people in orange vests pick up the boxes. They publish the chain. The next thousand flights get a procedure, or a widget, or a law. Tenerife gave us crew resource management. The Comets that came apart gave us hulls built another way. Needless repetition is supposed to get expensive.

A fatal fall in old age is, by body count, a bigger annual catastrophe than commercial aviation. What it usually produces is a death certificate, an ICD code, and a polite sentence in the paper. Sometimes the medical examiner stops there. “Accident.” Closed. If it happens in a hospital there may be a form. If it happens at home, which is where old people live, the story often ends at the polite sentence.

Peter Neumann died in a hospital after a fall, and the people who knew how to write an incident report (his people, the RISKS people, the ones who had been reconstructing other folks’ bad days since the Reagan administration) wrote obituaries instead. They said he was warm and kind and generous, which was true. Kind words left out where he fell, and on what, and whether it was night, and whether he lay on the floor, and what he had on his feet, and whether the house had a rail, and how long until help. His daughter Helen told The New York Times it was complications of a recent fall. That is a label. A label is a cause waiting to be written.

A family may keep a bathroom private. Grief has its rights. The larger habit is the trouble. We have decided that when an old person hits the floor, the correct genre is eulogy. We call it complications. We file it under expected. Aviation files a crash under a chain of events, even when the airplane is old.

A hidden risk is a gift to the next body. Silence is a rug over a hole in the floor. You walk. Then the hole finds a foot. Speak of the dead as they died, and the living stay awake. Open the file. That is the whole method. Comfort is a pillow. Truth is the only house a lesson can live in.

Therac-25 is in the body of knowledge because somebody reconstructed the race, the six-bit counter, the missing hardware interlock, the operators. “A machine killed people” would have taught a shrug. “A ninety-three-year-old died after a fall” teaches the same shrug. Nothing you can bolt to a stair.

So the disaster sat outside the ledger. The man who kept the ledger became a statistic of the kind the ledger was built to refuse.

* * *

Here is a fact that would have interested him if it had happened to a program.

Physical harm was in scope for his forum when a computer could be blamed. A Kawasaki robot killed a mechanic: item. An automated cart killed a man at ALCOA: item. Elevator interlocks failed in Ottawa and a person stepped into a shaft: item, with a long argument about door-latch switches and people who bypass safeties. A Stanford robot veered off course and fell down stairs: item. That is the only “fell down stairs” in the great catalogue. The mobile agent was a machine.

When the organic did the harm, the file stayed shut.

In 1994 he headlined a little story: Aging software ages suddenly! It was a date overflow in a program that analyzed aging. Aging meant a crystal, or a clock, or a listing. Legacy systems, bit rot, the year 2000, old hardware on life support. The forum is stuffed with the aging of machines. The aging of the operator sat offstage.

In 2005 he co-wrote a column called “Disability-Related Risks.” Aging appears as short-term memory and bad websites. Impaired mobility appears as a reason you might want to log in from home. The aging body is a user of software. It is waiting, in that column, to fail on a step.

In 1996 he abstracted a lawsuit about keyboards that wrecked women’s hands, and then he told on himself. His own keyboard had hurt the little finger of his left hand, the emacs finger. He installed foot pedals. The pain went away. He wrote that a person should leave the keyboard for a while. Then he wrote: Unfortunately, I don’t take my own advice enough.

That is as close as the written record gets to a man watching knowledge sit unused. Even then the body counted as a peripheral of the machine. The injury came from sitting. The work that keeps people his age alive is the other kind: getting up, and keeping the getting-up in working order, and making the path to the toilet a path a mammal can finish.

In 2020 he published “A Holistic View of Future Risks” and said almost everything is interrelated, which is true, and that a holistic approach had to anticipate human physical and mental health, which was a sentence about a virus and infrastructure. Health entered as a world-system. The stair waited.

His wife Elizabeth died that same year. He lived six more years and kept working. “Still working” is a lovely line in a eulogy. It is a thin safety case. The man who is still fine leaves the house as it was. Overconfidence is a classic RISKS theme. He had filed it under other people’s names for decades.

* * *

Would he have wondered, as a case study, what reasonable thing would have kept a mammal on his feet, or made a fall something a person could walk away from? That is a fair question. It is also the question he spent forty-one years asking of machines. Name the risk. Study the miss. Change the practice. Other people already do this work. They put rails into studs. They light the path to the toilet. They practice standing up from a chair. They move the bed downstairs. They keep a way to call for help within reach of the linoleum. Ordinary engineering. The same work he taught us to do with computers: see the hazard, change the system, spend the time. A useful thought. A realistic one.

That is the analysis that went unused. The house and the body are one system. Gravity is a reliable process. A first fall is already incident response. Strength is a control that needs years. Clutter is leftover state. Stairs are an unguarded operation. A throw rug is a debug flag somebody shipped. He had language for all of this. He used it on kernels. The hallway waited its turn.

What went quietly accepted was a charm: that remaining intellectually sharp is remaining physically competent. That a life of seeing risks is a talisman against ordinary ones. That a house may stay as it was at fifty. That forty-three thousand deaths a year are weather, while ten deaths in a bad airplane are a scandal.

He had already written the counter-spell. The operator is in the system. Remove the hazard. The hazard you leave unmodeled is the one that gets you. Apply those lines one inch outside the designed world and you have a rail, a clear floor, a lit path, a remaining squat, a watch that can call a daughter. Weakness in depth, he called the other arrangement. Two neglects stacked: the house unedited, the body untrained.

* * *

People will say it was his job to watch computers, and they will be right about the job, and they will have missed the tribute. A seismologist may drown in a flood and still have been a fine seismologist. A man whose subject is risk to the public, who says everything is interrelated, who says safety is a property of the whole, and who then draws the whole around the part with a processor in it: that man has made a philosophical choice. He has decided which kind of knowing counts.

The useful philosophy is the homely one. An idea is worth what it can do on a Tuesday, in a body, in a hallway. Computer risk, done from a chair as a spectator sport, is a kind of chess against imagined machines. It is a good game. It has saved lives. It is also a temptation: to keep the ledger of other people’s systems and leave the linoleum to luck. Abstractions are clean. They leave the rug where it is. They leave the gym for later. They leave the statistically busy killer unnamed, because it is old, and dull, and has no vendor.

We are complete animals. A complete person walks the method into the house. Total-system includes the creature who has to urinate at two in the morning. Simplicity is a one-floor plan. Defense in depth is a rail plus light plus a clear path plus a way to call from the floor. A note on the fridge is theater.

The computer focus made a whole profession brilliant at the new danger. The modal danger of late life is old. Novelty and lethality are different animals. The floor has no patch Tuesday. The organic is the system that takes time, and rails, and an honest file. The dark is your business if you mean the word risk.

Kind words about the clarifier, with the method left in the drawer: that is the poor tribute. He named risks so we could see them. He studied what follows a miss so a miss could teach. He advised practices so other people could catch the obvious, especially the obvious that is fatal and largely avoidable. Using that three-step on the accident that took him: that is the tribute he earned. The case study is still waiting. The particulars went with him. The class of accident is in the street, and in the bathroom, and on the stair, tonight.

Maybe it was a wet bathroom. Maybe a curb. Maybe a new pill and a stand-up too fast, which is a medical accident we call a fall. Maybe he did the ordinary engineering and ninety-three was still enough, which is allowed, which is how organisms end even in a well-edited house. A well-run life can still lose to a tile. What we have for certain is the shrug. We learned a eulogy. We left the lesson unwritten.

If someone you love dies of a fall, and you can stand it, write down the chain. De-identify it. Give it to the rest of us. Aviation buries the crew and publishes the wreckage. Publish the wreckage. That is respect of the kind he taught.

Do the boring work while walking is still dull. Edit the house as you would edit a dangerous program. Put the organism in the total system at fifty, when it looks like vanity.

* * *

Peter Neumann was a good man who did useful work. The floor had no opinion of Communications of the ACM. He kept score for the rest of us so we would pay the same tuition once. Then he paid a tuition his forum was built to see in every domain but this one, and we issued kind words, and we skipped the receipt.

Expertise travels when you walk it somewhere. Knowledge that stays in its genre is furniture. The digital game of assessing other people’s systems is a fine game. A person still has to stand up. The principles were in his own book. The creature who wrote it was the last system in the room.

The risk that went quietly accepted was the organic one: a human being is a decaying animal in a house, and the house and the animal are one system, and this system fails in the most ordinary way, and a culture of brilliant machine-watchers will call that ordinary, and go back to the interesting problems, and be surprised.

He taught that the hazard you leave unmodeled is the one that gets you.

He modeled computers.

He was ninety-three, and still at it, and then he fell, and then he died, and the digest did not carry the item.

The rest of us are still upstairs. There is time. That is the only nice part.