AI Makes Execution and Average Answers Cheap—Taste Now Sets the Ceiling. How Do You Improve Your Taste?

When execution gets cheap, judgment gets valuable

The biggest change generative AI brings is shifting creative work from "can we build it" to "what should we build." Writing, visuals, interfaces, code, and product prototypes can all start from a single conversation. Ideas that used to sit forever in a notes app can become something running and discussable in hours.

This speed is enormous leverage for the individual, but it also creates a new temptation. The first generation of output already looks usable, so the creator stops before they've figured out what they actually wanted to make. The work functions fine. The visuals are passable. The language flows. And yet it's hard to feel that any particular person is behind it.

The reason isn't mysterious. Models are good at combining common references, popular compositions, and safe phrasing. If the creator only checks "does it work," "will people accept it," and "does it look mature," then AI will efficiently push every choice toward the average. The internet is filling up with content that is technically fine and utterly forgettable—this is how it gets made.

This is where two uses of AI diverge. One use reduces skepticism and research, skipping the uncomfortable thinking that happens while an original idea is still forming. The other use expands the search space: more directions to test, more threads to connect, more assumptions to challenge. The model doesn't decide which path you take. It amplifies however much curiosity, judgment, and care you bring to it.

"Good enough" is dangerous in creative work, but that doesn't mean everything should be polished forever. Products still need to ship, be observed, and iterate. The problem is mistaking "it runs" for "it meets the standard for this piece of work." Shipping can be a deliberate experiment. Stopping too early just means no standard was ever set.

The lowering of execution barriers also shifts responsibility. In the past, if an idea couldn't be realized, you could point to the team, budget, permissions, technical support, or timing. Now that those obstacles are partly removed, it's harder to blame the conditions when your work lacks a point of view. AI can turn thoughts in your head into testable shapes. But deciding what deserves to exist is still a human job.

Taste is trained discernment

When cars first appeared, consumers cared most about whether they could replace the horse carriage. Once cars became a mature product, silhouette, color, materials, handling, and overall experience began to separate one model from another. Software is going through the same shift. Working features are increasingly the baseline. Brand, interaction, intuitiveness, and overall experience now decide who gets remembered.

When tools are scarce, people value owning them. When tools are everywhere, people value who can use them to make better choices.

Taste is that trained instinct. It looks like instant judgment, but behind it lies a lot of input, analysis, making, and correction. Having "an eye" usually means you have seen enough good work, studied why it works, and personally felt the frustration of knowing what was right but not being able to build it yet.

The difference between generated output and real creation tends to hide in unglamorous details: rhythm, hierarchy, atmosphere, restraint, clarity, negative space. The lower the execution barrier, the bigger the product and technical leverage that this kind of eye can command.

This discernment usually has clear boundaries. A person may be razor sharp about type and interfaces, yet oblivious to clothing or food. Taste is more like resolution trained along a specific lane; it rarely covers all of life evenly. Admitting that unevenness actually helps clarify what you want to train: what kind of problems are you trying to see more clearly?

Training starts with high-quality references. Designers look at good design. Writers read good writing. You can also follow who the people you admire are following. The value isn't in how many things you collect. It's in keeping your brain in steady contact with structurally sound choices, gradually recognizing how rhythm, scale, transitions, hierarchy, and detail relate.

Next, translate feelings into reasons. "I like it," "it's elevated," "it has texture," only record the response; they don't guide the next round of making. More useful questions: why does this interaction feel natural? How does the information hierarchy guide the eye to what matters first? Which element, if removed, would weaken the piece? When reasons can be stated, taste starts to become transferable judgment.

Finally, you have to make things yourself and let criticism back into the loop. Looking at excellent work trains observation, but it won't train execution. Once you actually build, the gap between eye and hand gets exposed. Early work falling short of your own standards is a sign that you can now see the distance. Specific criticism from the right people turns blind spots into the next round of practice. Rework is what presses judgment into the work itself.

A small exercise can start this loop: pick one excellent work and one ordinary work that solves the same problem. Compare only three dimensions: goal, hierarchy, and cost of use. Write a causal reason for each dimension. Make a small variant of your own that changes only a key choice. Then seek feedback with one specific question in mind. Don't ask "what do you think?" Ask "is the first thing visible on the screen the right thing?"

But taste training can't only happen when you are about to make something. Appreciation itself is a form of taste. A person may never paint, yet have extremely high resolution for painting. They may never make films, yet can say exactly why a shot works. Creators with taste are usually deep appreciators first—not just people who have seen a lot, but people who will linger on one thing, try it, compare it, turn it over, and notice the details that only appear on close looking.

Taste is also not a badge of identity, something you either have or don't. It's a level of resolution that increases along a particular lane. Instead of forcing yourself to be discerning about everything, choose one question you are genuinely curious about and let a trusted recommender, a mentor, or that person who always catches what you missed take you deeper into the reference space. The seven actions below are drawn from Brie Wolfson's 23 notes—covering appreciation, attention, guides, craft, filtering, and patience—and assembled into a daily practice table. This grouping is ours, not a structure from the original text.

What's easiest to overlook here is that the plus-minus scale isn't only for revising your work. It's also for curating your input. Every image, article, product, song, and space you scroll past every day is training your judgment. When something genuinely moves you, put down a "+" and one sentence about what exactly moved you. When something feels hollow, imitative, or merely stimulating for a moment, put down a "-" and note where it fails. Both saving and deleting should come with reasons. That's how a noisy information feed slowly becomes your own input garden.

Another easy-to-miss move is returning to the materials. Talking abstractly about form, structure, and meaning cannot replace familiarity with materials, tools, and constraints. When you actually make something, note why a typeface breaks at this size, which paper takes the ink, which interface element makes users hesitate, which sentence reads stiff. Creating forces the appreciator to train both expression and craft at once. Your inner feelings can only become something uniquely yours when you have concrete technical means to bring them out.

These moves won't generate a stable style immediately. The world needs to be digested slowly. After a while, reopen your plus-minus log and check which preferences still hold and which were just platform and trend effects at the time. Taste has no shortcut proof, but it will grow its own logic through long, consistent attention, choice, and making.

Your eye is being shaped by your environment

Taste can be trained, but it doesn't form in a vacuum. The compositions, colors, and typefaces we think look "just right" are the accumulated residue of long debates, education, imitation, and cultural change. Many rules have been repeated long enough that they move from visible choices to unexamined intuition.

Once you accept that taste can be trained, the next question appears: are we training a purely private pair of eyes, or a judgment that can be discussed together? Historically, there are three paths. David Hume saw taste as an ability that must be refined through repeated exposure and education. Immanuel Kant focused on how aesthetic judgment, while personal and felt, is still guided by principles like harmony and balance. Pierre Bourdieu warned that what counts as "good taste" is also shaped by cultural capital and can even become a tool of social distinction.

These three paths put taste into training, principles, and social context. They don't replace the full theories of those thinkers, but they are enough to explain a key tension: personal feeling is real, and at the same time, a community is always shaping the language we use to judge work.

Community shaping taste does not mean wealth, price, or in-group status can prove taste. Wealth can increase access to good work, but it does not produce judgment. Snobbery is only one possible external posture of taste, not its essence. Treating "expensive," "niche," or "picky" as quality is exactly how identity replaces actual judgment.

The innermost layer is fundamentals. Composition, proportion, hierarchy, clarity, legibility, the emotion of color, and the tone a typeface sets before a word is even read—all of these explain why a piece works. They are not timeless style formulas; they are tools for organizing attention and reaction.

The next layer outward is history and community. Baroque drama, modernist restraint, postmodern irony—each is tied to the problems of its own era. Studying only the surface look of a work easily turns one period's answer into a false universal. Truly useful references cross time and discipline, and they include knowing what environment a piece was responding to.

The outermost layer is platforms and algorithms. Digital platforms amplify certain aesthetics, and algorithms mix them quickly. Yesterday's novelty can be a template in a few weeks. Input is richer than before—and also more easily homogenized. A person thinks they are browsing freely, when in fact they have been surrounded for a long time by one kind of look: easy to spread, easy to copy.

Training taste, then, also requires checking your input regularly: are all references coming from the same platform and the same professional circle? Are you seeing originals, or versions that have been imitated several times over? Can you say what time-bound problem or use context a choice serves? If the trending direction reversed, would your reason for keeping this judgment still hold?

For products, resisting trends also doesn't mean only chasing utility. Usability and personality must hold at the same time: the user shouldn't have to guess how to use it, and the work still shouldn't lose its own voice.

Good taste has to end up in the work's tradeoffs

High standards without the ability to build only produce more precise criticism. Strong execution without standards just reaches the average answer more efficiently. For taste to become productive, it has to change decisions at the site of creation: which problem to pick, which ornament to cut, which difficulty to accept, when to tear up an early plan.

Fundamentals, history, and platforms explain where taste comes from, but they don't yet answer the most concrete question of the making session: what exactly should be cut, changed, or torn up in the next version? Paul Graham—a programmer, writer, and investor—turned this into a more useful question in his essay "Taste for Makers": if taste really can improve, then do good work in math, software, architecture, painting, and writing share any common features? He ended up with fourteen. They are not a style formula. They read more like fourteen questions that force you to look at your work again.

To bring those fourteen items directly into the making session, we have grouped them here into four working areas: structure and problem, time and originality, craft and rework, and environment and courage. That grouping is our way of using the material, not Graham's own structure.

The first working area checks structure and problem. Simplicity means the work stands on a few truly strong structures; decoration can't hide an empty core. Solving the right problem matters more than making a beautiful answer to the wrong one. Leaving room for generation lets the reader or user complete the work. Symmetry, recursion, and learning from nature can reduce complexity—but mechanical repetition and surface-level biomimicry need to be watched.

The second working area checks time and originality. Enduring time means using the past and the future to escape the control of the current trend. Deliberate borrowing requires knowing where an idea comes from, and puts getting it right ahead of performing originality. Genuine strangeness can't be forced; aim for accuracy first, and the result may naturally end up a little odd. A touch of humor comes from confidence in the work—it doesn't need to take itself so solemnly.

The third working area checks craft and rework. Worthwhile difficulty forces you to strip away extra decoration and face the hard problem directly. Bad friction comes from capricious demands and unreliable conditions—it adds cost without improving the solution. Good work looks effortless usually because a lot of practice has moved what once needed conscious control into fluent ability. Experts also expect early plans to be discarded, and they choose tools and processes that make it easy to admit mistakes and change.

The fourth working area checks environment and courage. Great work often appears in clusters, because high-quality communities provide density of questions, intensity of criticism, and visible standards. Courage requires noticing the gap between consensus and what is true. Many improvements start with an understated complaint: this part is too sloppy; there has to be a better way. Disliking ugliness is not enough by itself—you need to understand the domain first to know where things are really worth fixing.

These standards cannot be turned into a machine-graded rubric. Simplicity doesn't mean everything should be minimal. Enduring doesn't mean ignoring history. Difficulty doesn't mean longer hours always equal higher quality. Borrowing is not a free pass for copying. Their value is in forcing creators to account for choices, to notice decoration that dodges the real problem, and to find direction for the next redo.

Keep the human filter

Taste ultimately shows up as a series of refusals: refusing to treat the first draft as the finish line, refusing to let popularity replace reasons, refusing to let decoration hide the wrong problem, and refusing to hold onto a design just because time has already been sunk into it.

Novelist George Saunders described revision as an internal "plus/minus meter": when weighing a word, a sentence, or an entire passage, you sense whether it moves the work in a positive or negative direction, then make a small adjustment on the spot. Taste takes shape through thousands of these micro-decisions. Intuition may send the first signal, but the creator must still test it through cuts and edits — not treat the feeling of "this feels right" as a license to stop explaining.

This kind of refusal isn't passive. It lets the creator keep what genuinely works, overturn the parts that look fine but have no justification, and keep shaping until rhythm, hierarchy, atmosphere, restraint, clarity, and open space all serve one intention. You might keep iterating through apps, articles, design directions, or small systems. All of these experiments practice the same thing: preserving your own filter while taking advantage of convenience.

These six questions can be applied directly to the next AI-generated article, design, product, or code. First, write down the problem the work is solving, then open the first draft. Change one high-impact choice at a time, and note both why it is better and what was sacrificed. When you can't find a reason, don't paper over it with "it just feels better"; and when nothing seems worth changing, be suspicious—you may simply be numbed by how fast it was produced.

AI makes average answers cheaper and makes real choices more visible. Volume alone no longer proves creativity. The work eventually exposes what you saw, what you refused, why you redid it, and whether you can drag your own standards into reality.

Takeaways: everyday taste training and a reference kit

The training kit below was put together for this piece by reworking actionable methods from five different articles; it is not a list that appears as-is in any of the originals. The seven-item table in the middle presents the daily methods from the fifth article; this section instead threads all five together into three rhythms: 15 minutes per day to train resolution, 45 minutes per week to build up references, and a verification pass at the end of each project. All the material is shown directly on the page, or you can copy the whole package or download it as Markdown.

This training does not demand one major piece each day. What it asks is that you leave behind comparable evidence of judgment, so that inputs, reasoning, craft, critique, and rework gradually connect. Real accumulated taste ultimately shows up through a trail of references, versions, and trade-offs.

The main thread of this piece draws on four articles by Juliette, Emil Kowalski, Raphael Salaja, and Paul Graham, while also incorporating Brie Wolfson's additional notes on everyday appreciation, attention, input filtering, craft building, and micro-decisions. Full links to all sources are at the end of the piece.