The Blind Spot of Mastery
And Why AI Matters
The Universal Problem of Mastery
I recently listened to a podcast with Atul Gawande, the surgeon and writer, who described a strange and unsettling idea:
In some fields, even the best performers have coaches. In other fields, even the worst performers don’t.
Roger Federer, at the height of his power, perhaps the most naturally gifted tennis player ever, still had someone watching him from the outside. Not because they could hit a forehand better than he could, but because no one, not even Federer, could fully see themselves from the inside.
Itzhak Perlman once said that his wife shaped him as a violinist because she sat in the audience and told him, without flattery, what she heard and what could be fixed. A great performer cannot watch their own performance.
And yet, in many fields, there is a strange silence. You can train for decades, accumulate thousands of hours of experience, and at a certain point you are simply… alone. No one watches you anymore.
Invisible Plateaus
Gawande himself had performed thousands of thyroid surgeries on cancer patients when, out of curiosity, he asked a senior colleague to observe him. He was confident there was nothing meaningful left to improve. He felt he had reached mastery.
The feedback came back and he was shocked. There was a long list of things he could do better. Things he had simply never noticed. He hadn’t stopped improving because he’d reached mastery. He had plateaued invisibly.
The unsettling implication is that what feels like mastery may simply be a stable equilibrium. One in which no one is correcting you anymore.
Many of us who have learned Torah seriously for years recognize what Gawande felt after discovering this plateau.
You learn consistently and understand more than you once did.
Your learning is confident and competent. Even solid.
And yet something has stalled.
Because we rarely name this phenomenon, we quietly accept it.
Chazal identified the condition that makes these plateaus possible: learning without exposure to correction. A careful reading of the Gemara in Makos 10a and Taanis 7a is striking. The Gemara says that one who learns alone becomes foolish, and it brings a couple of different Pesukim to prove it. What is surprising is not the idea itself. We all know the value of learning with others. But the framing is curious. Why does the Gemara attack learning alone instead of the opposite, which would be promoting learning with others?
The Maharsha offers a penetrating explanation:
כי הטעות מצוי בלמוד יחידי, משא”כ בלמוד של רבים, שאם יטעה אחד מהם יש שני להודיעו טעותו
Error, he explains, is common when learning alone. When others are present, mistakes are exposed.
Learning without a person next to you isn’t inferior because of motivation or loneliness or because you’ll be missing out on the Geshmak of a chavrusa. It’s inferior because there is no mirror. No external surface against which your thinking is forced to take shape and be examined.
Without that mirror, ideas feel coherent simply because they are unchallenged. Assumptions pass for proofs. Confusion masquerades as depth.
This is why the Gemara says you’ll become foolish. Without a learning partner you are treading dangerously. You’re creating the conditions for a slow decay of thinking when there is no corrective feedback. You don’t remain static. You drift. Your errors harden into habits. The Gemara is describing the challenge of reliable self-correction.
The Limits of the Chavrusa System
Of course, the Beis Midrash does not ignore this danger. Its default format of learning is with a chavrusa.
The chavrusa concept is rightly celebrated as one of the great educational innovations in Jewish history. It sharpens ideas, forces clarity, exposes weak proofs, and keeps learning alive.
But it does not solve everything. It addresses the danger Chazal identified. But rarely at the level that enables perpetual individual growth.
Your chavrusa is inside the Sugya with you. They are not standing outside your learning, observing your habits. A chavrusa is Federer’s hitting partner, not his coach.
It was built to test ideas in real time, not to surface long-term patterns in how a person thinks.
It rarely addresses patterns such as:
recurring misreadings
overreliance on familiar Lomdishe frameworks
weak summarisation of the Shakla V’tarya
persistent blind spots across areas of Shas
These are not Sugya-level problems. They are learner-level problems.
This is what coaching looks like.
Chazal didn’t merely observe this as a psychological risk. There is Halacha around this idea.
The very first se’if in Yoreh Deah (1:1) addresses someone who knows the laws of shechita and considers himself fully competent. The Mechaber rules that such confidence is insufficient: a shochet may not rely on his own assessment. Only after being observed, tested, and certified by another may he act independently. Knowledge, experience, and familiarity with the halachos do not qualify a person to judge himself. Competence, halacha insists, must be confirmed from the outside.
Chazal identify the risk of uncorrected learning. The chavrusa system addresses it partially. But no institution ever emerged to provide systematic, habit-level feedback to adult learners. Almost no one is taught how to identify their weakest cognitive patterns, let alone how best to fix them.
The Gemara gives us a fascinating illustration of this in a well-known story in Bava Metzia 84a.
After the death of Reish Lakish, R’ Yochanan takes on a new chavrusa. But the replacement is not what he needs. Each statement R’ Yochanan makes is met with agreement and support rather than challenge.
What devastates R’ Yochanan is not the absence of a chavrusa. He finds one. He follows Chazal’s advice. And yet the learning collapses.
Because a chavrusa is only part of the solution. Without someone who exposes weaknesses, forces clarification, and refuses to let ideas settle too easily - the form remains, but the function is gone.
Even the greatest minds did not merely need a chavrusa. They needed the right kind of feedback.
What all of these cases share is not failure, but a feature of mastery itself. As competence increases, obvious errors fall away first, leaving behind subtler habits of thinking that no longer announce themselves as problems. When those patterns stop being surfaced from the outside, mastery doesn’t break, but it quietly plateaus.
AI - A New Kind of Observer
And now we come to AI. AI cannot replace a chavrusa. It cannot replace a rebbe. It cannot replace the fire of face-to-face Torah.
But it can, if we choose to build it that way — do something unprecedented.
For the first time in history, it is realistically possible to create tools that can objectively monitor how a person learns over time. Tools that trace patterns, compare hundreds of mistakes side by side, and surface habits that remain invisible in the moment.
Not because AI understands Torah better than people do, but because it can observe processes across time and make meaning of patterns in a way no human ever could.
Some may argue that in practice we aren’t quite there yet. Some of this exists only in fragments. And the tools we have today aren’t always reliable. Maybe this is true. But it is coming and it is no longer science fiction.
Such tools could map recurring errors, detect where thinking consistently collapses, identify gaps in knowledge of certain sugyos, and highlight habits that quietly limit growth.
And then say:
These are the patterns holding you back. Fixing them will change how you learn.
The question, then, is not whether AI will enter the Beis Midrash. It already has. The real question is what kind of AI we choose to build.
This is not Torah replacement. It is Torah enhancement, rooted in a deeply Jewish insight articulated by Chazal themselves:
A prisoner cannot free himself from prison (Brachos 5b). You cannot correct your own blind spots. Without external correction, thinking deteriorates.
The Beis Midrash has always given us partners to argue with. What we now have, if we choose it, is the opportunity to build tools that help even the most serious learners continue to grow, rather than plateau unseen.