Circle of Competence
The size of the circle does not matter. Knowing exactly where its edge runs is the whole discipline.
The Model
There is a domain where your judgment beats the average, and outside it your accuracy collapses while your confidence stays exactly where it was.
That second half is the problem. Competence is domain-specific and confidence is not.
Confidence travels with you across the boundary and gives no signal that anything has changed.
Buffett's framing in 1996 was that what counts is not the breadth of the circle but the precision with which you know its perimeter. A tiny circle, honestly mapped, outperforms a wide one with a soft edge, because the tiny one never bets outside itself by accident.
Note what the model does not say. It does not tell you to stay small, and it is not a licence to refuse anything unfamiliar.
It tells you to know which side of the line you are standing on before you commit capital, time or reputation. Inside, trust your read. Outside, buy someone else's.
Why operators get this wrong
The common failure is category confusion. The words are the same, so the game is assumed to be the same.
A founder builds a four-million services business over nine years and then moves into consumer product, because both are business, both have customers, both have margins. He runs a sales-led operating rhythm — pipeline reviews, quota, a weekly forecast — in a category actually decided by shelf placement, packaging and repeat purchase rate.
Two years and most of the profit from the first business go into learning that the vocabulary transferred and nothing else did.
The second failure is evaluating a discipline from outside it, where everything looks simple. A technical founder refuses to hire a salesperson because sales is just talking to people. From outside the circle, that is exactly what it looks like.
The third is subtler and hits experienced operators hardest. You do not lose the circle when you enter a new field. You lose it when your old field changes underneath you and you keep trading on a map that expired.
Applied
Body
Your reading age is not your training age.
Someone who has read every hypertrophy paper published in the last decade and has nine months under the bar will write themselves a periodised programme, and it will be defensible on paper and wrong in the room.
The gap is not information. It is never having felt what week six of accumulated fatigue does to bar speed, or how a warm-up feels on the morning before a joint gives you trouble. That data does not exist in text.
Coaching is the purchase of somebody else's circle for the years you have not lived yet. That is the honest description of what you are buying, and it is cheap at the price.
Rehabilitation is the sharpest case. It sits outside almost everyone's circle, and it arrives at the exact moment your confidence peaks, because the pain has stopped. The pain stopping is the least reliable signal in the entire process.
Business
Draw the line by outcome, never by title. A title is a claim; the circle is a record.
The working definition: you are inside it where you have made the same class of decision at least five times and can explain why each went the way it did. Five, because three is a coincidence and one is a story.
The practice that maps the perimeter is a prediction log. Before each decision of consequence, write two lines — what you expect to happen, and the mechanism you think will cause it.
Date it. Do not revise it.
After a quarter, sort the outcomes. The domains where your stated mechanism turned out to be the real cause are inside the circle. The ones where you got the outcome right for a reason that was not the reason are not inside it, and those are the dangerous ones, because a win reinforces the map regardless of whether the map was correct.
Most operators discover their circle is narrower than their org chart and wider than their job description.
AI Leverage
A language model performs most fluently in exactly the domains you are least equipped to audit.
Inside your circle, its output is useful and you catch the errors immediately, because you know what wrong looks like. Outside, you cannot distinguish a correct answer from a confident one — and confidence is the only property that is guaranteed.
That makes the boundary a hard operating rule rather than a preference. Use models to accelerate work you could grade yourself. Outside the circle, use them to arrive at the expert conversation better prepared, not to replace it.
The specific things not to generate and ship unaudited: a contract clause, a tax position, a medical or health claim, a compliance statement. Each one produces output that is textually indistinguishable from the version that would have been correct, and the cost of the difference lands months later on somebody else's desk.
There is an inverted use worth more than any of this. Ask the model what a specialist in this field would ask me that I have not thought to ask. That question maps your perimeter faster than any amount of self-assessment, because it is answered from outside.
The Drill
One page, two columns. Left: I have done this at least five times. Right: I have read about this.
Force six entries into each. The forcing is the point — the fifth and sixth items in the left column are usually the ones you had stopped counting as skills, and the right column is where the flattery lives.
Now take the largest decision currently on your desk and file it into one of the two columns. Be strict. Adjacent does not count.
If it lands on the right, your entire job this week is to find one person for whom it lands on the left, and buy an hour of their time. Not a course, not a book, not four hours of reading. One hour with someone who has made that specific decision more than five times.
Come with your prediction already written. You will learn more from where they disagree than from anything they volunteer.
Stoic parallel
The Stoics claimed Socrates as their forerunner, and the claim rests largely on one move he made in the Apology.
Told he was the wisest man in Athens, he went looking for a counterexample and found that the craftsmen genuinely did know their crafts — and that on the strength of that knowledge, each one believed he also understood the largest questions. Their real competence had licensed a false one.
Socrates concluded he was wiser only in this: he knew the boundary of what he knew.
Epictetus put the operating instruction on top of it. Do not hold forth on principles you have not digested, he told students who wanted to sound like philosophers before they had become them. Eat first, then show the result in how you live — do not vomit the book back at people.
That is the circle stated as a rule of conduct rather than a rule of investing. Same edge, same discipline in staying behind it.
One model per week.
Applied to training, business, and AI leverage. No fluff.
Related models
- Arguing from First PrinciplesReduce the problem to what is irreducibly true, then rebuild without borrowing anyone else's conclusions.
- Opportunity CostsThe real price of anything is the best thing you gave up to get it, and it never appears on an invoice.
- LeverageOutput per unit of input. Change the multiplier, not the effort behind it.
Origin: Warren Buffett and Charlie Munger, set out in the 1996 Berkshire Hathaway shareholder letter