David Marquet knew submarine physics. What he didn’t know was his own boat. He had spent a year preparing for one boat when another’s captain quit early and the Navy sent him to the USS Santa Fe, a class he had never served on. “I knew the physics and the theory,” he said in a 2025 interview, “but not the ship-specific systems.”
The style he was good at, in his words, was “making the right decision and then getting the team to execute it.” On the Santa Fe, the captain was not the person most likely to be right.
One day an officer brought him a plan to fix a torpedo tube. Marquet thought it would mess up other parts of the operation. Instead of arguing, he offered his chair. “Here, sit in my chair while I grab a coffee. When I come back, tell me what you would do if you were me.” Thirty seconds later the officer had a different plan, and a good one. “I intend to” tells the rest of that exchange.
What could thirty seconds show? In a 2015 Q&A, Marquet had already described the idea. “Most leaders think that first people demonstrate competence and clarity and then we’ll give them a modicum of control. I found that the opposite is more effective.” Give a bit of control, he said, and let people “tussle with the problem for a bit,” even for 30 seconds. “That’s when you’ll learn what the gaps in competence and clarity are.”
Hand people the decision, and with it the know-how and the why.
One stool, three legs
Marquet’s name for what he built, in Turn the Ship Around! (2012), is the leader-leader model. Its opposite is the leader-follower model he was trained in, which he traces to the industrial age: “one group of people making decisions, and a different group carrying them out.”
In leader-leader, he told InfoQ, “each person is thought of, and thinks of himself or herself, as a leader. That means they make decisions and take responsibility for those decisions.” The captain gets no exemption: “the boss is accountable for following the organizational principles just as much as everyone else.”
One former Santa Fe crewman, quoted on the publisher’s page, says he preaches “the three legs (control, competence, clarity) of Leader-Leader” and moves “the decisions to where the information lives.” Competence is knowing the work well enough to decide it. Clarity is knowing what the organization is after, which Marquet’s site describes as leaders conveying “what they are trying to achieve.”
It can go wrong in both directions. Years before the Santa Fe, as engineer officer of the USS Will Rogers, Marquet gave his team intent instead of orders; his website says they made poor decisions that led to errors, and he went back to giving orders. Read against the three legs, the team had control and not much else. Then, early in his Santa Fe command, he called for a speed the backup motor didn’t have. His most experienced officer of the deck knew it and passed the order on anyway: “Because you told me to.” He had the knowledge; the decision wasn’t his. Marquet vowed then to stop giving orders.
Driving into port
His examples of building the other two legs are small. One is pulling into port, where a submarine is frequently evaluated. A captain who cares only about today’s grade, he said, will use his best officer every time. “But eventually, he’s going to leave.” Rotate the officers, take a small hit, and “now everyone learns how to drive the ship.”
On the clarity side, he asked his officers to write their own evaluation for the following year. “Now they’re the ones framing their goals,” he said.
By Marquet’s account, retention went from one in ten to 100%, and ten of his officers were later selected to command submarines.
The jammed conveyor
Say you’ve just taken over a warehouse night shift and have never run its sorting system. A conveyor jams, and the shift lead asks what to do. Deciding yourself means guessing, and “your call” can fail in ways you won’t see.
The chair version: “If you were running this shift, what would you do?” He has a reroute for the conveyor within a minute. To clear the backlog, he proposes holding one large order until morning. He can’t know that this customer is this week’s priority.
One answer showed competence, the other a gap in clarity, and they call for different moves. The reroute can be his call from now on. The priority list is something you owe him, posted where the shift can read it.
A plan before the edit
Coding agents get their own thirty seconds. In Claude Code’s plan mode, Claude reads files and runs commands to explore, then writes a plan, “but does not edit your source.” In most sessions, edits stay blocked until you approve it.
Anthropic’s best-practices guide explains why: “Letting Claude jump straight to coding can produce code that solves the wrong problem.” A plan shows a clarity gap, such as a wrong reading of the request, while it is still cheap.
Planning adds overhead, the guide says. It is most useful “when you’re uncertain about the approach, when the change modifies multiple files, or when you’re unfamiliar with the code being modified. If you could describe the diff in one sentence, skip the plan.”
Unfamiliar code is Marquet’s situation on the Santa Fe: the one approving knows the territory least.
The difference is in what grows. In a 2015 keynote, Marquet said control should move in small steps, up what he calls a ladder of leadership. A year after his vow, an officer could come to him with “I intend to submerge the ship” and a full list of checks already done, and get two words back: “Very well.” The guide’s test for skipping a plan is about the size of the change, not a track record. On the Santa Fe, the chair went to a person. In plan mode it is a prompt, and one of its answers is “No, keep planning.”