On July 2, 1881, at the Baltimore and Potomac Railroad Station in Washington DC, President James A. Garfield was shot twice by an assassin named Charles Guiteau.
Garfield did not die from the bullet.
He died eleven weeks later, on September 19, 1881, of massive infection caused by the repeated probing of his wound by physicians who refused to wash their hands, sterilize their instruments, or adopt the antiseptic techniques that Joseph Lister had established as best medical practice more than a decade earlier.
The physician primarily responsible for Garfield’s care was Dr. D. Willard Bliss.
Bliss was aware of Lister’s antiseptic techniques. He had explicitly rejected them as unnecessary. He inserted unwashed fingers directly into Garfield’s wound. He probed with unsterilized instruments. He created a system of infection so severe that when Alexander Graham Bell brought a metal detector he had invented specifically to locate the bullet in Garfield’s body, Bliss controlled the examination and directed Bell to search the wrong side of the President’s chest — where Bliss had already convinced himself the bullet was.
The bullet was on the other side. Bell’s device could have found it. Bliss’s certainty prevented the discovery.
When Garfield died, Guiteau said at his trial: “The doctors killed him. I just shot him.”
Guiteau was hanged. He was also, in a specific technical sense, correct.
The Bliss failure mode
The pattern Bliss demonstrated has a specific structure that recurs across every domain, every decade, and every industry. The pattern:
A senior authority figure is confronted with new evidence, new techniques, or new information that contradicts their existing beliefs. The evidence is well-established. The techniques are demonstrated. The information is clear.
The authority figure refuses to update. Not because they are stupid. Because updating would require acknowledging that their prior confidence was misplaced. And the social cost of the acknowledgment is higher, to them, than the operational cost of persisting in error.
The persistence in error is not a private failure. It is exercised through their authority. Other people who could have helped — Bell with his metal detector, the nurses who were arguing for handwashing — are overruled by the senior authority who insists that his existing beliefs are correct.
The result: catastrophic outcomes that would have been preventable if the authority figure had exercised the discipline of updating.
The 2026 application
The Bliss failure mode is the single most dangerous pattern in enterprise AI leadership right now.
Senior executives who formed their beliefs about AI in 2022 and 2023, when the technology was less capable, are now applying those beliefs to decisions in 2026, when the technology has advanced substantially. The beliefs are out of date. The senior executives, in many cases, know this at some level.
They also know that updating would require acknowledging that decisions they made two years ago on obsolete beliefs were miscalibrated. The social cost of the acknowledgment is high. So the updates do not happen.
Their organizations, meanwhile, are being run on beliefs that were true two years ago and are no longer true. The persistence in error is being exercised through executive authority. Junior employees who understand the current state of the technology are being overruled by senior executives who insist that their two-year-old models are still correct.
The outcomes are being produced accordingly. Not by hostile actors. Not by inadequate technology. By senior leaders’ refusal to update beliefs that need updating.
Three practical questions
One: what specific beliefs about AI did you form in 2022-2023 that have not been updated since? List them. For each, ask when you last tested the belief against current evidence. If you cannot remember, the belief is likely obsolete.
Two: who in your organization is currently trying to tell you that your existing beliefs need updating? The junior employees running unauthorized AI experiments. The engineers who have been using frontier tools daily. The consultants pushing back on your assumptions. Are you listening, or are you Bliss?
Three: what is the specific evidence that would cause you to update your current AI investment thesis? If nothing would cause you to update, you have a Bliss belief. Bliss beliefs, when held by senior leaders, produce Bliss outcomes.
The closing thought
Dr. D. Willard Bliss was a competent physician by the standards of his time. He had been medically trained. He had years of surgical experience. He was chosen by the White House to be President Garfield’s primary physician.
He also killed the President of the United States by refusing to update his beliefs about antiseptic practice. The evidence was available. The techniques were demonstrated. Other qualified physicians, including some at Garfield’s bedside, were arguing for the update. Bliss refused.
Guiteau went to the gallows saying the doctors killed the President. He was legally guilty of assassination. He was also technically correct about the specific mechanism.
In 2026, senior executives running AI transformations on 2022 beliefs are producing Bliss outcomes at organizational scale. The organizations that will define the next decade will be led by executives who exercise the discipline of updating. The organizations that will be diminished will be led by executives who defend their prior confidence at the cost of updating.
The world has changed. The leaders who notice will be the ones the next decade is built around.
