The Craft Of Judging Well

I used to think taste is un-arguable.

You look at the work. You feel the thing. You know whether it has life in it. Some people have it. Some people don’t. Very elegant. Very mysterious. Very convenient, especially for people who like to believe they have taste.

The longer I build, write, lead, and try to engineer systems that help people think, the less I believe this. Taste is not just subjectivity wearing a better jacket.

Taste is trained judgment.

And you cannot train judgment by standing outside the work, pointing at inputs and approving outputs. Judgment is trained in the middle. In application. In applying yourself to the transformation until the work starts applying something back to you.

This became painfully obvious to me while trying to build verifiers and evals for work that does not live in the clean mathematical world of code. Code is almost luxurious in that sense. A software developer (or a coding agent) can write tests. They use fixtures, mock the world of the program and see if the machine behaves.

Strategy, transformation, creativity, any form of abstract cognitive work … all less polite when it comes to objectively verifying the quality of the work.

A brief can be clear and still useless. A recommendation can be complete and still cowardly written. A narrative can be coherent and yet not move anyone. And we can have agents setup plans that pass every structural test and still collapse the moment they meets an actual room full of actual humans with actual incentives, fears, politics, fatigues, ambitions, and … taste.

Reality is resistant to mockups. Reality is not an experiment. It just is.

That is where craft begins.

A craftsperson doesn’t simply look at a piece of wood and imagines a table. The craft lives in the pressure between hand, tool, grain, mistake, correction, patience, force and restraint. The wood pushes back. The tool reveals. The material teaches. The craftsperson is merely executing a plan, they are being educated by the transformation itself.

The same is true in strategy. The material is messier, but the craft is no less real.

The input might be research, culture, business pressure, a leadership concern, or a technological shift, a strange little thought that refuses to leave you alone. The output might be a decision, a sharper reality, or simply someone having a better day because your interaction made them feel seen, oriented and capable.

But the work is what happens between.

How does meaning come together? Where does the fog lift? What needs to happen before a team begins to share reality? Where does a sentence become a decision? Where does a possibility become worth protecting? What is the point where force transitions from being helpful, actually breaking things?

That middle is not administrative.

It is the craft.

As we are entering the age of abundant output – AI can generate more concepts, drafts, strategies, notes and briefs, plans and deck-shaped objects than any sane human should ever agree to review – we realize, this abundance is not free, of course. Somewhere behind the magic trick sit energy, chips, capital expenditure, infrastructure, water, heat, and the economic theater of pretending tokens fall gently from the sky like digital snow. Or, depending on your perspective, they are pumped like oil.

But from the user side, output feels cheap.

The old fantasy was that AI would make everyone more creative by giving everyone more options. Maybe. But more options mostly create a new burden: the burden of judgment. What is worth asking for? What is worth making? What is worth refining, and what can be safely ignored? What is worth the compute, the attention, the meeting, the edit, the next hour of your finite human life?

You cannot chase every possible output just because the machine can produce it. Or maybe someone can. I am just not convinced I want to donate my life to proving it.

So the question becomes: what is the new craft?

I think it has three fields.

Aperture. Aim. Application.

Aperture is the judgment of what you allow in. What you expose yourself to. What you permit yourself to notice. What excites you, yes, but also what disturbs you, contradicts you, embarrasses your prior thesis, or reveals that your taste has become a little too pleased with itself.

Aim is the judgment of what deserves to exist. Not every idea wants to become a thing. Not every spark deserves a bonfire. Aim asks what change would actually matter, for whom, at what cost, with what consequence.

Application is the judgment of how fully you enter the transformation. How much pressure does this deserve? How much resolution? How much time, detail, care, and patience? Where do you stay loose? Where do you become exact? Where do you let the material teach you something you did not know when you began?

Application is not busyness. It is not effort cosplay. It is also not heroic suffering performed for the approval of productivity priests.

Application is contact.

It is the quality of your participation in the making.

Aperture, aim and application are not steps. They are fields, acting on us continuously: What we notice changes what we want. What we want changes what we notice. What happens through application changes both (it’s almost uncanny, how closely the three mirror Heisenberg’s uncertainty principle).

This is why outsourcing transformation, cognitive processes like strategy, creativity or innovation is dangerous for an organization. And this includes any form of outsourcing. Whether there is an AI agent on the other end of the process or some external entity: you may retain taste for a while, you may still know what looks good, sounds smart, feels right and passes the vibe check. But if you never wrestle with the mess yourself anymore, your taste slowly loses calibration.

Because here’s the thing: craft does not merely execute judgment, craft produces judgment.

The writer learns taste by writing the sentence and hating it enough to write a better one. The builder learns taste by shipping the feature and discovering where reality does not care about the elegance of the architecture. The strategist learns taste by watching the room misunderstand the beautifully phrased thought and realizing the work was not done. The leader learns taste by noticing that the right decision, delivered with the wrong human texture, will damage trust.

Relationships are craft.

Leadership is craft.

Selfhood is craft.

We should treat ourselves with the same meticulous diligence a craftsperson brings to their material. Not with cruelty or perfectionist panic. Stones are shaped by discipline and heart. The artists sees the from and the to.

From reactivity to agency. From noise to attention. From cleverness to usefulness. From performance to contribution. From wanting to be right to becoming more capable of serving the work.

This is not a call to reject AI. I am building with AI every day. I am interested in what agents can do, how they can reason, how they can help us carry more complexity than a human mind comfortably holds alone. But I do not believe the future belongs to people who merely orchestrate agents while remaining untouched by the work.

No matter how many agents you run, the puck still stops with you. You are still responsible for the aperture. You are still responsible for the aim. You are still responsible for the application.

And yes, I find it funny that some of the more aggressively enlightened AI people are now trying to bottle taste into recipes, scoring systems and agentic evaluation rubrics, as if taste were just a function waiting to be parameterized.

Maybe they are right. Maybe taste can be modeled better than we think. Maybe creativity can be captured in the mathematical bottle after all. I have spent so much time around machines that I have seen time and time again how human mystery often becomes less mysterious when we build the right instrument to see.

But even if taste can be modeled, that does not mean it can be abdicated.

A map of taste is not the same as having taste. A verifier of quality is not the same as becoming the kind of person who can make quality more likely. Because that is the form apprenticeship we need now: Not just studying computer science, but apprenticing yourself to building systems. Not just studying strategy, but apprenticing yourself to better decision making. Not just studying leadership, but apprenticing yourself to be more consequential.

Taste is the craft of judging well.