On May 29, 1944, eight days before D-Day, the man in charge of the Allied air forces told General Dwight Eisenhower that one part of the invasion was unwise.
Air Chief Marshal Sir Trafford Leigh-Mallory had been warning about the American airborne drop since April. Two divisions, the 82d and the 101st, were to land behind Utah Beach in the dark. The Germans had just moved their 91st Division into the area. According to the U.S. Army’s official history, Leigh-Mallory now told Eisenhower that casualties were likely to run over 50 percent, and that the results would be too small to count on. In his memoir, Eisenhower recalled the phrase his air commander used: the “futile slaughter” of two fine divisions.
Eisenhower called him “earnestly sincere,” a man “merely giving me, as was his duty, his frank convictions.” He asked Leigh-Mallory to put it in a letter. Then he went to his tent alone to think. Canceling the drop, he concluded, meant canceling the Utah landing too, or sending it in with worse odds. He sent back his answer. The risks were great, he agreed, but the airborne landings were essential, and if the invasion went in, those risks had to be accepted.
The planes that would carry the American paratroopers belonged to Leigh-Mallory’s command.
The argument comes first
At the final weather conference, at 4:30 in the morning of June 5, the forecast was still below the planned minimum from the air point of view. Leigh-Mallory later wrote in his official report that he gave his opinion that the air forces could provide a reasonable measure of protection “and that airborne operations would be practicable.” Shortly after midnight, aircraft of the IXth Troop Carrier Command, part of his force, dropped the American paratroopers.
Argue hard before the decision. Once it’s made, carry it out as if it were your own.
The first half mattered as much as the second. A month earlier, Leigh-Mallory had objected so strongly that the airborne plan was substantially revised, and most glider landings moved to dusk. “His objections were sustained and the changes made,” the official history says. In May he was heard again, in person and in writing, and he lost.
The drop was messy: clouds and flak scattered some planes, and one battalion of the 101st spent most of D-Day assembling. But in his report Leigh-Mallory called the airborne air operations successful, “with a remarkably low casualty rate,” counting losses as a share of sorties flown. That outcome doesn’t prove his warning wrong; good decisions can end badly and bad ones well, a trap called resulting. By Eisenhower’s account, when the attack succeeded, Leigh-Mallory was the first to call him, to share his delight and to apologize for adding to his burdens.
Bezos’s phrase, Grove’s habit
Jeff Bezos put a name on it in his 2016 letter to Amazon shareholders. “If you have conviction on a particular direction even though there’s no consensus, it’s helpful to say, ‘Look, I know we disagree on this but will you gamble with me on it? Disagree and commit?’” The boss, he added, should do it too. When a team wanted to make an Amazon Studios original he doubted, he wrote back: “I disagree and commit and hope it becomes the most watched thing we’ve ever made.”
The idea is older. Richard Tedlow, a biographer of Andy Grove, told Business Insider in 2025 that disagreeing and committing “was the essence of how he felt you should comport yourself at Intel,” though Tedlow wasn’t sure Grove coined the phrase.
It can also be bent. The same article reported Amazon’s chief executive, Andy Jassy, telling staff in 2023 that it was time to “disagree and commit” on returning to the office, and that it was “probably not going to work out” for those who didn’t. Used as an ultimatum, the phrase skips the part Bezos had described as “a genuine disagreement of opinion, a candid expression of my view, a chance for the team to weigh my view.”
The deadline you think is wrong
Imagine your team is choosing a launch date, and you think March 1 is three weeks too early. You say so in the planning meeting, with the testing numbers. Your manager listens and picks March 1.
Half-committing looks reasonable. You tell sales the date “might move.” You hold back the extra testing hours, since the date will slip anyway. If March 1 fails, you were right, and nobody can tell whether the date was the problem or the slack.
Committing looks different. You send your manager two lines: the risk you see, and the test result that would make you raise it again. Then you plan to hit March 1 as if you had picked it.
When the AI disagrees with you
OpenAI’s Model Spec, which in the company’s words “outlines the intended behavior” of its models, uses the same pattern. In its August 2026 version, it compares the assistant to “a conscientious employee” who shouldn’t just say yes to everything: it “may politely push back,” while “remaining respectful of the user’s final decisions.” When a request seems to conflict with the user’s longer-term goals, the assistant should “briefly and respectfully note this discrepancy. Once the user understands the concern, the assistant should respect the user’s decision.”
That is a company’s statement of intent, not proof that any model does it. And with a model, commitment can’t be read from its tone. It shows up in the work.
Imagine you ask a coding agent to keep your app’s old date format. It replies that a standard format would be safer. You say keep the old one. Check the result: search the changed files for the format it preferred, and read the diff before you merge.
If the old format is in every file, the agent disagreed and committed. If its own format turned up in two of them, it only said it would.