The OODA Loop’s Founding Story Doesn’t Hold Up (and Boyd Knew It)

Ten to one. That is the number every OODA loop explainer reaches for: ten MiG-15s shot down over Korea for every F-86 Sabre lost, proof that the pilot who cycles faster through Observe, Orient, Decide, Act beats the one who cycles slower.

The number has been shrinking for a while. Doug Dildy and Warren Thompson went back through the records in 2012 and got 5.8 to 1 overall, and 1.4 to 1 against the Soviet units that flew the MiGs best. Thomas McKelvey Cleaver’s MiG Alley (2019) put the whole war between 1.1 and 1.5 to 1 and called the ten-to-one figure propaganda. Dick Becker, a Sabre pilot, said of the early fighting that they fought the Russians to a draw.

If you searched “ooda loop” and landed here, you wanted the four steps and an example. You will get both. But the founding story matters, because the lesson people take from it, faster wins, is the one Boyd spent his last decade correcting. And it matters more now, because AI tools have just made two of the four phases nearly free. What happens between them is the whole game, and the speed story taught everyone to skip it.

What the OODA loop is

The OODA loop is a four-step decision cycle (Observe, Orient, Decide, Act) developed by John Boyd.

John Boyd was a U.S. Air Force fighter pilot and instructor who never wrote a book. His ideas survive as briefings he delivered from the 1970s until his death in 1997, and in Robert Coram’s biography.

The loop has four phases.

Observe. Take in what is happening: the other aircraft’s position, the instruments, the sky.

Orient. Make sense of it. Boyd’s own definition is the one to hold onto: orientation is the set of “images, views, or impressions of the world shaped by genetic heritage, cultural tradition, previous experiences, and unfolding circumstances.” Everything you bring to the observation before you have consciously thought about it.

Decide. Pick a course of action from the ones your orientation makes visible.

Act. Do it, which changes the situation, which produces new observations. The loop restarts.

The popular reading is that the loop is a race: cycle faster than your opponent and they are always responding to a situation that no longer exists. Boyd did say this. In his Korea story, the Sabre’s bubble canopy let pilots observe more, its hydraulic controls let them act faster, and the MiG pilots ended up disoriented and paralyzed.

Where the speed reading breaks

The evidence for the speed reading was the kill ratio. The ratio now sits between a draw and an ordinary advantage, and the historians who deflated it credit other causes: a design flaw that made the MiG-15 snap into uncommanded rolls at high speed, Soviet pilots rotating home before they got good, and by 1953 an American advantage in numbers.

To be clear about what that revision does: it does not prove orientation won Korea. It removes the only evidence that speed did. The case for orientation has to come from somewhere else.

Here is the part the explainers skip. Boyd’s mature work does not rest on the story either. In Organic Design for Command and Control, a 1987 briefing, he wrote: “Orientation is the Schwerpunkt. It shapes the way we interact with the environment,” and so it “shapes the way we observe, the way we decide, the way we act.”

Schwerpunkt is a German military term for the point where effort concentrates. Boyd was saying the middle of the loop is not one phase among four. It determines what the other three can do. A pilot with poor orientation observes the wrong things, decides among the wrong options, and acts on a picture of the world that was never accurate. Cycling faster gets him to the wrong place sooner.

Replace the pilot with someone pasting a model’s first draft into a document and you have the same failure at desk speed. The observation was fast, the action was instant, and nothing in between was theirs.

Boyd never retracted the number that I can find. But his final sketch of the loop, from 1995, has orientation as the largest box, feeding into everything else.

An OODA loop example that is not a dogfight

A product team looks at a dashboard on Monday morning. Conversion dropped four points over the weekend.

Observe is the dashboard. Orient is where the team’s history takes over: last time it was a checkout bug, the person who built checkout is on vacation, marketing ran a campaign Friday that nobody told engineering about. Decide is whether to page someone, roll back, or wait. Act is whichever they choose.

Now run the same morning with a team that has been through eight of these drops. They see, with little discussion, that the campaign traffic is lower intent and the drop is expected. No page, no rollback. They look like they skipped orientation. Boyd borrowed a German word for this: Fingerspitzengefühl, fingertip feeling. It is orientation compressed by repetition until it runs at the speed of instinct.

The first team is not slower. Their orientation is thinner. Give them a faster dashboard and they will page the wrong person faster.

From the outside, compressed orientation and skipped orientation look the same: both are fast. The test is a question. Ask the experienced team what they would have done if the campaign had not run, and they will tell you. Ask someone who pasted a first draft what they rejected, and there is nothing to say. Speed you can explain is expertise. Speed you cannot is a shortcut.

Why the misreading matters now

Speed does still win sometimes. When two sides are oriented about equally well, the faster one has the edge, and where the situation changes faster than anyone can think, cycle time decides. That is the honest version of Boyd’s Korea claim, and for fifty years it was harmless, because observing and acting were expensive.

That constraint is gone. Both ends of the loop have been accelerated for everyone at once. Everyone observes and acts at the same pace now, so the one condition under which speed decided, roughly equal orientation, no longer holds.

Orientation is the only variable left, and the middle is where most people are not doing anything. Output comes back from the model and goes straight into the document. Observe to Act, with Orient and Decide removed. We have a name for what that produces: slop. Orientation is the one phase no tool accelerates for you. It is built from previous experience, which means reps.

How to practice orientation with AI output

Artists get their reps in critique, where five people read the same piece five ways and you decide what to do with all of it. You can build the same thing around a model, if you treat what comes back as a critique rather than an answer. Three exercises, in order.

Ask for five, keep one, say why. Ask the model for five versions instead of one. Pick, then write one sentence about what the four rejects have in common. At first it will be vague (“too generic”). By the tenth rep it will be specific (“all four open by restating my question”), and specific rejections are the ones you can act on.

Argue with the good one. Ask the model to make the strongest case against the version you kept, then decide whether it is right. Most people skip this because the draft already looks fine, which is when orientation atrophies.

Write down what you always reject. After a week you will notice the same rejections recurring. Write them down as rules, with the reason. “No paragraph ends on a takeaway, because it hides what the paragraph was for.” That list is your orientation, externalized: previous experience turned into something the model can read and you can revise.

The third exercise is where this stops being a metaphor. A file of your standing rejections is a working instruction set for an AI tool, and building one from your own drafts is what I teach on Maven.

Boyd’s real lesson

The founding story of the OODA loop was that faster wins. The story is wrong. Boyd moved past it. The loop is not a race. It is a description of how much of what you do depends on the sense you have already made.

That reading holds up, and it matters most now that the phases on either side of it have been handed to machines. Orientation is learnable. Why the critique model transfers best to AI feedback, and how to build the practice, is what I wrote about in The Orientation Practice on Substack. The exercises above are the free version.

Keep the four verbs. Drop the stopwatch.

Sources

• John R. Boyd, Organic Design for Command and Control (briefing, 1987)
• John R. Boyd, The Essence of Winning and Losing (briefing, 1995)
• Robert Coram, Boyd: The Fighter Pilot Who Changed the Art of War (2002)
• Doug Dildy and Warren Thompson, F-86 Sabre vs MiG-15: Korea 1950–53 (Osprey, 2012)
• Thomas McKelvey Cleaver, MiG Alley: The US Air Force in Korea, 1950–53 (Osprey, 2019)
• From Balloons to Drones, “The Korean War and the OODA Loop: What Happened to the Kill Ratio?” (2025)