Beginner’s mind gets sold as a virtue, and that’s exactly what makes it useless. Stay curious, approach every problem fresh. Nobody can act on that, because acting on it fully would mean re-deriving your whole stack every morning and shipping nothing. The version worth having is narrower and more expensive. Beginner’s mind is a budget you spend deliberately, and the best return comes from spending it on the things you’re most certain about.
I’ve been learning this on a mat. I’m a few weeks into Shaolin-Do kung fu, which qualifies me to say nothing about kung fu and quite a lot about being a beginner. Twice a week I go be visibly bad at something in a room full of people, then go back to a job where I’m the one people ask. Being a genuine beginner stops happening somewhere in your thirties unless you arrange for it, and the contrast is sharp enough to be uncomfortable.
The same form, for decades#
What actually reorganized my thinking wasn’t the white-belt part. It was watching the other end of the room.
In tai chi, the form a beginner is learning is the form the advanced students are doing. Same movements, same sequence, for decades. What changes is the quality of attention someone brings to it, and the corrections stop being about where your foot goes and start being about things you didn’t know were happening. Nobody graduates out of the basic form. They just keep doing it with more attention.
That’s the inverse of how expertise works in my day job. The Zen framing, usually attributed to Shunryu Suzuki, is that the beginner sees many possibilities where the expert sees few. Read as inspiration, it’s a nice line for a slide. Read as a bug report about your own head, it’s more useful. Expertise works by caching conclusions so you stop paying to re-derive them, and caches go stale without ever telling you they have.
The belief nobody re-tested#
The shape is always the same, and you’ve probably got one right now. A team decides the integration suite is too slow for every pull request, which is correct the day it’s decided. It goes into the runbook, then into the culture, and stops being a decision at all. Then a dependency upgrade quietly fixes the slow part, and nobody re-times it, because nobody re-times settled things.
Pick your own version. The library that doesn’t support the feature it shipped two releases ago. The vendor rate limit everyone still designs around, raised in a changelog nobody reads. The manual deploy step that survives a permissions fix by about three years.
These never announce themselves. No outage, no incident review, just a slightly worse workflow that everyone adapted to and stopped seeing.
None of that requires anyone to be careless, and that’s the part worth sitting with. The rule survives because it’s settled, and settled things don’t get audited. You re-check what you’re unsure about, which is rational triage and also means your confidence map and your audit map are inverses of each other. Whatever you’re most sure of gets examined least. Certainty is the thing protecting a stale belief from review.
Expertise doesn’t come with a timestamp#
Senior engineers are the worst at this, and it isn’t a character flaw, it’s the mechanism working as designed.
Being senior means carrying a large library of “I’ve seen this go badly before,” and that library is most of what you’re paid for. The pattern-matching is the job. But every pattern in there was learned against a particular version of the world: a specific team size, a specific tool version, a specific set of constraints that felt permanent at the time. The pattern doesn’t store any of that context. It gets filed as knowledge, indistinguishable from the parts that are still true.
The tell is a four-word sentence: “We tried that already.” It arrives with enough authority to close the topic, and it’s missing two of the three things that would make it load-bearing. When did we try it. What specifically failed. What’s changed since. If you can’t answer all three, you don’t have a finding, you have the memory of a feeling, and the team just made a decision on it.
This is a different problem from the one in “I don’t know yet” is a senior move. That’s about a gap you can see. This one is harder, because the gap is filed in your head as an answer, and answers don’t raise their hand.
The mechanism the mat has#
Here’s the structural thing the mat has that the office doesn’t.
Martial arts institutionalizes the return to fundamentals: in the curriculum, on a schedule, whether you feel like it or not. A senior student still stands in the beginner’s stance and still gets corrected in it, and the correction lands harder at that rank precisely because they’ve done it ten thousand times and built everything else on top of it. The system assumes your foundation drifts, and is designed around that assumption.
Software has nothing like this. There is no belt test where a principal engineer has to re-derive why the deploy window is Tuesday. The incentives run the other way, because re-examining a settled thing looks like wasted motion right up until the moment it doesn’t, and it never produces a ticket you can close.
So you build the mechanism by hand, small enough that you’ll actually do it.
The free audit most orgs destroy#
The first one is free, and it expires.
For roughly the first ninety days, a new hire has beginner’s mind without cultivating it. It’s just a fact of where they’re standing. They ask why the deploy window is Tuesday, why the runbook says restart the service before checking the logs. They aren’t being difficult. They’re reporting that the reason isn’t visible from where they are, which is real information about whether the reason still exists.
That’s an audit of every undocumented decision you own, and most teams accelerate its ending without meaning to. The reflex answers do it: “that’s just how the pipeline works,” “historical reasons,” a small laugh and a change of subject. None of those are hostile, and each one teaches the person that this question doesn’t produce answers here. After three or four they stop, and you’ve converted an auditor into one more person who knows the workaround.
The fix takes ten seconds and is slightly uncomfortable: answer with the real reason, or say out loud that there isn’t one anymore. “The deploy window is Tuesday because in 2022 the vendor ran maintenance on Mondays. I don’t know if that’s still true. Want to go find out?” That turns a naive question into a work item and tells everyone in earshot what happens to questions here.
Then write them down. A running doc of the why-do-we-do-it-this-way questions, each with its answer or a conspicuous blank where the answer should be, costs nothing to keep. The blanks are a backlog you didn’t have to invent.
The second mechanism is yours: pick one settled belief a quarter and re-derive it. Good candidates score on load, meaning how many downstream decisions rest on it, and on age, meaning how long since anyone last touched it. Those two overlap more often than you’d like. One a quarter is plenty, and it’s roughly one more than most teams manage.
Where it’s genuinely hard#
There’s a real failure mode on the other side, and I’ve been on teams with it. The engineer who treats every settled decision as permanently open, who reopens the ORM question quarterly, who calls it first-principles thinking. That isn’t beginner’s mind. That’s an organization with no memory, and it costs more than the stale belief ever did, because now nothing stays decided long enough to build anything on.
The difference is a stop condition. Real beginner’s mind asks a question that can be answered and then accepts the answer. Time the suite, read the vendor’s changelog, run the thing once and look. It ends in a fact, and the fact settles it for a while. The other version asks questions that can only be argued about, and it never resolves, because resolving was never the point.
The budget applies here too: you can’t re-check everything, so pick on purpose and let the rest stay decided.
When did you last check?#
Beginner’s mind isn’t pretending you don’t know things. You know things, and that’s the entire reason anyone put you in the room. It’s remembering that everything you know was learned on a particular day, under conditions that have quietly moved since, and that your strongest beliefs got the least scrutiny on the way in because they arrived looking obvious.
The advanced students are still doing the beginner’s form. They’re just doing it with their eyes open, checking the same stance they’ve been standing in for years. The question was never whether you’re an expert. It’s when you last checked.
