With tools like AI, it has never been easier to turn an idea into something that runs. That made me ask a simpler question: if execution is cheap, what still matters? I keep coming back to taste.
How I learned to see
Photography made this clearer for me. A lot of a picture is decided by what you leave out of the frame. After looking at enough photographs, including my own bad crops, I can often feel when something is off before I can explain why. Learning to see is less about collecting references and more about noticing which details belong and which are just noise.
I learned the same thing late in software. Early on, I reached for the most general abstractions I could find. They looked elegant on their own, and they often made the system harder to understand. Mistakes taught me when an abstraction helps and when it is just noise.
More than intuition
Taste can feel like intuition. However, seeing patterns is not enough. You also have to know which pattern fits, which to ignore, and when the familiar solution is the wrong one.
The people around you shape this too: what they value, and whose taste counts. Having your own point of view helps. But it can also make you dismiss something you have not learned to see yet.
Craft, care, taste
In product and design circles, taste, craft, and care get mixed together and thrown around a lot.
The distinction I find useful is this:
Craft is how well you make the thing, care is how seriously you take the thing, and taste is knowing what the thing ought to be.
Taste is not a palette. It is judgment: what deserves attention, what belongs, and what should go.
Emily Segal talks about good taste as discernment, pattern recognition, and idiosyncrasy.1 She also says taste is social and depends on context. I agree. Experience shapes judgment, and experience is never neutral.
The work you cannot see
One way to use taste is to leave things out. When adding something gets easier, the hard part is deciding whether it belongs. Taste is not less complexity. It is the right amount.
Some of the best decisions leave no mark in the product. A user never sees the feature a team chose not to build. That makes taste hard to credit and easy to confuse with minimalism.
Those decisions are also hard to defend when shipping more is what gets rewarded.
Taste in software
Users rarely see the code, but they feel where you put the complexity. A technical choice can change how a product behaves, how fast it feels, and how quickly a team can change it later. An API can make a team faster or slower depending on how it was designed.
A good abstraction makes the common case obvious and the unusual case possible. Two APIs can do the same job: one feels inevitable, the other leaks the system underneath.
Clever design shows what you can do. Tasteful design shows what you refused to make everyone else understand.
Can a company keep this?
I have often heard people say a product used to be better. Inside companies, I have heard the same feeling from the other side: things feel slower, or quality feels lower, than at the start.
As companies grow, more decisions go through shared systems and processes. Those systems matter. Without them, quality gets uneven. The risk is using them instead of judgment.
A design system can keep good decisions about type, spacing, and interaction. It cannot tell you whether a new feature belongs in the product. Old decisions are a useful base. They do not answer the next question for you.
Keeping taste at scale is hard. But I do not think it is impossible.
Systems can support taste. They cannot keep it alive on their own. What keeps it alive is room to disagree: people who can say a feature does not belong, even when it follows every rule, and leaders who treat that disagreement as part of the work, not as friction to remove.
The arrogance of it
Taste often asks you to decide before you have all the evidence. Conviction is not the same as certainty.
“This feels wrong” is not the same as “this query took two seconds.” One is a judgment. The other is a measurement. Whether two seconds is too slow still depends on what the product needs.
The danger is not having strong taste. It is treating a strong preference as if it were strong evidence. Metrics can be arrogant in the same way when we ask them to settle a question they cannot answer.
You cannot always measure your way to a good product, and you cannot rely on taste alone either. Using taste does not mean ignoring data. When the numbers answer the question, use them. Sometimes there is no history to look at, the work is new, or what matters is hard to quantify. That is when taste helps most.
Success is not proof of taste. A tasteful product can still fail. A successful one can still feel incoherent. Caring about the work means holding both of those possibilities without treating your preferences or the numbers as the whole answer.
What still makes sense later
Work made with taste can still make sense after the fashion that produced it has gone. Style answers what looks right now. Taste answers what makes sense here. A twenty-year-old product can look old and still feel well thought-out.
Tools do not give you taste. Someone still has to decide what fits, what matters, and what to keep.
Taste alone was never enough. Taste chooses the direction. Craft decides whether you can get there. Care decides how far you are willing to go. When almost anything can be made, deciding what deserves to be made is the work that remains.
References
-
Emily Segal, “TASTESLOP”, Nemesis Memos, May 5, 2026. ↩