On Friday, March 21, 1986, Ray Cox checked himself into the East Texas Cancer Center, in Tyler, Texas. He was about to undergo the last of a series of radiation treatments to remove the few remnants of a shoulder tumor surgeons had operated on a few weeks before.

In a familiar routine, technicians placed Cox on a table beneath a huge radiation machine, a Therac-25. The massive machine, then one of only eleven in North America, was state of the art. Barely two years old, it gave a much wider and much more flexible range of radiation treatments than the old cobalt radiation machines it replaced. The computer-controlled machine could deliver high-intensity beams to destroy big tumors deep in the body, or it could use low-intensity beams to destroy tiny tumors near the skin's surface. Which intensity it used depended on the settings technicians typed into the computer controlling it.

That day, Cox was supposed to receive a short, low-intensity burst, but there was an unnoticed problem in the computer program controlling the machine. Whenever a technician set the machine for a heavy radiation dose, then quickly changed the setting in a certain way, the computer program lost the correction and retracted the machine's safety interlocks. No one knew of the flaw, even though it had resulted in deep radiation wounds to a patient in Georgia the year before.

In the small lead-lined treatment room in Texas, Cox was lying face down on the table beneath the machine, waiting. In the next room, two technicians were setting up the computer, telling it what to do. When the machine powered up, Cox felt an electric shock pass through his shoulder. He saw a bright flash of light and heard something frying. Seconds later, a second burst struck him in the neck and a spasm shot through his body. Alone in the sealed radiation room, he jumped from the table and yelled for help, pounding on the heavy door. The next day, he began spitting up blood. His eyelids drooped, his pupils dilated, and he lost the use of his left arm and most of his sweat function. Doctors had no idea what had happened but could tell that he had suffered irreparable nerve damage. He spent the next five months in a hospital bed, then died.

Cox wasn't the first to die at the hands of the machine. A month after his burn, another patient got a lethal dose in the same bed and the same room. Instead of burning his shoulder and neck, the beam went deep into his brain. He died less than a month later. It happened again in January 1987. The errors of a computer programmer and a medical technician, coupled with a poorly designed safety-interlock system, had claimed three human lives. It took another year for the problem to be finally tracked down.