The last lesson left you with an uncomfortable fact: the edge of your circle is invisible from the inside. The band just past your boundary is exactly where a subject still feels familiar, so your sense of competence stays high while your actual competence has already quit. That’s the cruel design of it — the feeling lies most convincingly precisely where you most need the truth.
So mapping your circle can’t rely on how you feel. If the feeling is the broken instrument, you need instruments that don’t run on feeling — external checks that report back what your gut can’t see. This lesson hands you four of them: keeping score, the “what would I need to know?” test, the teach test, and a guard against outcome bias. Together they let you draw the boundary from the outside in.
Before you read — take a guess
Before we start — which of these is the most reliable signal that something is genuinely inside your circle of competence?
Keep score: track your forecasts
Start with the single most powerful instrument, because it’s the one your ego can’t argue with: keep a written record of your predictions and check them later. Think of it like a pilot’s logbook. A pilot doesn’t decide they’re a good flyer by how confident they feel in the cockpit; they have hours logged, approaches flown, outcomes recorded. You can build the same thing for your judgment.
The method is humble: when you make a call, write it down, attach a probability, date it, and revisit it. “I think this launch hits its Q3 target — 80% confident.” “This hire works out — 70%.” Then, weeks or months later, score yourself. Over enough calls, your hit rate per domain quietly draws your circle for you. Where you’re reliably right, your maps are good. Where you’re not, they aren’t — no matter how it felt at the time.
Here’s the math made concrete. Suppose you log ten confident predictions in Domain A — your home turf — and you’re right on nine of them. In Domain B, you log ten equally confident calls and land four. Both felt the same from the inside; the confidence was identical. But 9/10 versus 4/10 isn’t an opinion, it’s a measurement. Domain B is outside your circle, full stop — and the only reason you can see that is the scorecard. Without it, both domains would still feel like home.
This is also the bridge to a model you’ll meet later: calibration — being right about how right you are. A well-calibrated person who says “70% confident” is correct about 70% of the time across all their 70% calls. You can’t get there by introspection; you get there by tracking forecasts and comparing your stated confidence to your real hit rate. Scorekeeping is calibration’s raw material.
Keep a one-line decision journal
You don’t need software or a system. A single running note works: one line per real decision or prediction — what you expect, how confident (a percentage), and the date. Add a column for the outcome and leave it blank until you know. Two months in, you’ll have something no amount of self-reflection can give you: evidence. The point isn’t to feel good about the wins; it’s to catch the domains where your confidence and your accuracy quietly disagree.
When to use it
Start the log now, before you need it — calibration data only exists in hindsight, so a scorecard you begin today is worth a fortune in six months and nothing in six minutes. Lean on it hardest for recurring decisions in domains you suspect you overrate: the ones where you feel sure but keep being surprised. That gap between felt-confidence and measured-accuracy is the exact shape of your edge.
The “what would I need to know?” test
The second instrument works before you have any track record, which makes it the fast one. Ask yourself: “What would I need to know to have a real opinion here — and do I actually know those things?”
The tell is whether you can even generate the list. If you can name the handful of things that genuinely drive the answer — the key variables, the failure modes, the questions a real expert would ask first — and you know them, you’re probably standing inside the circle. If you can’t even produce the list, that’s the signal you’re outside it. You don’t merely lack the answers; you don’t know what the questions are. That’s the deepest kind of outside: you don’t know what you don’t know.
Worked example. Someone asks you whether a particular city apartment is a good buy. Inside your circle, you’d rattle off the list without strain: the price per square meter versus comparable recent sales, the building’s reserve fund and pending special assessments, the neighborhood’s supply pipeline, the rental yield if you let it, the financing rate, the resale liquidity. You can name what matters and fill in most of the values. Now ask the same person whether a biotech startup’s lead drug candidate is worth backing. If the honest internal response is a blank — “I’m… not even sure what determines that” — you’ve found the edge. The inability to write the list is the answer.
You're asked to judge whether a new manufacturing supplier is trustworthy. You immediately think: 'I'd need their defect rate, their on-time delivery history, their financial stability, and a reference from a current customer' — and you know how to get and read all four. What does the 'what would I need to know?' test say?
The teach test (explain it simply)
The third instrument is the famous one, often credited to physicist Richard Feynman: if you can’t explain it from first principles, in plain language, without leaning on jargon or authority, you don’t understand it. You’re merely familiar with it — and that’s lesson 1’s distinction made operational. Familiarity lets you nod along and reach for the right words; understanding lets you build the idea up from the bottom for someone who’s never heard of it.
The trick is that jargon and name-dropping are camouflage. “It works because of the synergistic optimization of the throughput layer” sounds like understanding and contains none. “It works because as you add more requests, the queue here fills up first, and once it’s full new requests wait, which is why response time spikes” — that’s a mechanism a child could follow, and you can only produce it if you actually have the model. The teach test strips the camouflage: explain it to a smart twelve-year-old, from first principles, and watch where you stall. The place you stall is the edge of your real understanding.
This connects straight back to lesson 1: understanding versus familiarity isn’t a vibe, it’s testable. The teach test is the test. If you find yourself saying “well, the experts say…” or “it’s complicated, but trust me,” you’ve located a spot where you have the labels but not the machine.
Sort each signal: does it put you inside the circle or outside it?
Pick the right option for each blank, then check.
'I can derive why it works from the basics, step by step' is the circle. 'I can name the experts and the buzzwords but not the actual mechanism' is the circle. 'I can predict how it will behave in a brand-new situation' is the circle.
Beware outcome bias
Now the instrument that protects the other three from corruption. A good result does not prove competence. You can be right for entirely the wrong reasons — luck wearing the mask of skill — and if you score yourself by outcomes alone, luck will quietly redraw your circle in the wrong place. This is outcome bias: judging a decision by how it turned out rather than by whether the reasoning was sound.
The fix is to judge the process, not just the result. Ask: given what I knew at the time, was that a well-reasoned call? A great decision can have a bad outcome (you played the odds and the unlikely thing happened) and a terrible decision can have a great outcome (you got lucky). When you keep score in Domain A, you’re scoring a process you can repeat. When you score a one-off lucky win in an unfamiliar field, you’re scoring noise.
And here’s why that one lucky win is the most dangerous data point you can collect: it falsely widens your perceived circle. Worked example. You know nothing about commodities, but on a whim — a tip from a stranger, a hunch — you put money on cocoa futures, and they happen to triple. You walk away $5,000 richer and with a brand-new, completely false belief: I have a feel for commodities. The win taught you the wrong lesson. Your circle didn’t grow an inch; your perceived circle just annexed a region where your real hit rate is a coin flip. The next bet, made with that inflated confidence, is the one that hurts.
One win is a sample size of one, and a sample of one can’t distinguish skill from luck — that’s the whole problem. If a thousand people each guess a coin flip, roughly five hundred are “right,” and every one of them can tell themselves a story about why. The winners are the only ones who feel entitled to conclude anything, which is survivorship bias stacked on top of outcome bias: you only weigh the time it worked, and you only remember the outcome, not the flimsy reasoning behind it. The honest question isn’t “did it work?” It’s “if I made this exact call a hundred times, knowing only what I knew then, how often would it work?” A single right answer is the most expensive teacher when it convinces you a coin flip was a skill.
The pitfall
The seductive trap is treating your highlight reel as your track record. Your memory volunteers the wins and buries the near-misses and the lucky escapes, so a fair-minded review of “how have I done?” comes back rosier than reality every time. That’s exactly why the written scorecard from the first section matters: it logs the calls before you know how they turn out, so luck can’t quietly edit the record afterward.
A worked map: the hiring manager
Let’s draw one real person’s boundary with all four instruments at once. Maya is an engineering manager. For backend candidates, she’s deep inside her circle: she can pose a problem and probe how a candidate reasons through it, spot the difference between memorized answers and genuine understanding, predict fairly accurately how they’ll perform on the job — and her track record of backend hires backs that up (the teach test and the scorecard both pass). She knows exactly what she’d need to know about a candidate, and she knows how to find it out.
Now she’s asked to evaluate design hires. Same office, same hiring process, same confident feeling of “I know good work when I see it.” But run the instruments: she can’t reliably tell a genuinely strong portfolio from a lucky-looking one, she can’t probe a designer’s reasoning the way she can a coder’s, and she has no track record to lean on. She can name a few buzzwords but not the mechanism of good design. By every external check, design hiring is outside her circle — even though it sits right next to a domain she dominates, which is exactly what makes it the danger zone.
The competent, honest move is not to bluff a rubric and hope. It’s to bring a designer into the loop — someone whose circle covers this — and weight their read heavily. This is the part people miss: competence includes knowing whom to defer to. Recognizing the edge and routing the decision to someone inside their circle isn’t an admission of weakness; it’s the most expert thing Maya does all week. A manager who insists on judging design hires solo, on the strength of “I’ll know it when I see it,” is the one quietly making expensive mistakes just past the rim.
The deference move
Mapping your circle has a payoff that isn’t just defensive. Once you can see your own edge clearly, you can route each decision to whoever’s circle actually covers it — yourself when you’re inside, the right specialist when you’re not. Knowing the boundary is what makes good delegation possible. You can’t hand off what you can’t admit you don’t know.
Recap
- The feeling of competence is the broken instrument, so map your circle with external checks instead of introspection.
- Keep score. Write down predictions with probabilities and dates, then check them. Your hit rate per domain (9/10 vs 4/10) draws the boundary that confidence can’t — and it’s the raw material for calibration.
- Run the cheap tests. “What would I need to know?” — if you can’t even list it, you’re outside. The teach test — if you can’t explain it from first principles in plain words, you’re familiar, not competent.
- Beware outcome bias. Judge the process, not the result. A lucky win in an unfamiliar field is the most dangerous data point you can collect, because it falsely widens your perceived circle.
- Competence includes knowing whom to defer to. Recognizing your edge and routing the call to someone inside theirs is an expert move, not a weakness.
Quick check — mapping your circle
Across ten confident predictions in Domain A you were right nine times; across ten equally confident predictions in Domain B you were right four times. Both domains felt identical from the inside. What's the honest read?
Check your answer to continue.
Where this goes next
You now have a toolkit for finding your real boundary from the outside in — scorekeeping, the “what would I need to know?” test, the teach test, and a guard against the lucky win that lies to you. You can locate your edge.
The natural next question is how to move it. The next lesson, “Widening Without Faking,” is about growing your circle outward honestly — how genuine expansion actually happens, and the crucial skill of telling real new competence from the convincing impression of range. Because the same illusion that hides your edge will happily fake the feeling of having crossed it.