Five registers that move taste from one head to many: perception, decision, craft, system, transmission. Each one a floor. Each one buildable.
The floor everything rests on. Before you can choose, you must notice; before you can reject, you must see. Accumulated exposure builds categories, and categories let you see. The more you have, the more the world differentiates.
Where preference acquires spine. Choosing against other good options, under time pressure, without complete data. AI multiplied the options; without a decision layer teams drown in possibility. With one, they ship.
Where decision becomes artifact. Every material has a grain. Large language models included, with sharp edges where they fail. Craft means building with that grain instead of against it, and putting guard rails where it breaks.
Individual craft evaporates without system. Principles that get cited during tradeoffs, patterns that spread without an enforcer in the room, governance that protects character at scale. The least-photographable register, and the most consequential.
The ceiling. The test: do people on your team make decisions you would have made, without you in the room? Transmission is teaching through proximity. Crits run as decision meetings, demos that elevate the team's work over the leader's.
In 1995, an HCI researcher named Noam Tractinsky published a study that became one of the most-cited findings in interaction design: users form aesthetic judgments about interfaces in roughly 50 milliseconds. Half a tenth of a second. Before they read a word.
This is the physiological floor of the perception register. We are judging constantly, pre-verbally, beneath conscious access. What changes with cultivated taste is what we perceive, not whether we do. A builder with developed taste sees things the untrained viewer does not yet have the categories to see. Accumulated exposure creates categories, and categories let you see.
The capacity beneath this, and the one most often missed, is stillness. The ability to pause long enough to actually see what is in front of you. Most builders, under pressure, default to reaction. They squint at the output, feel the first thing they feel, and call it judgment. Real perception waits longer. It watches what is actually happening before it names it.
Perception is not opinion. It is resolution. The more you have, the more the world differentiates.
The practical consequence for any leader: perception is largely invisible. You cannot put it on a roadmap. You cannot photograph it. But it is the foundation everything else rests on. A team whose perception is dull will produce work that is technically fine and emotionally flat. A team whose perception is sharp will surprise you.
When every team can generate twelve variants before lunch, the scarce capacity is not production. It is reflection that resolves. The ability to sit with the options long enough to see which one is right, then close the loop and ship it. That closing is the hard part.
This is why decision-layer taste is where teams most often fail. Not because they don't see the right direction. Because they cannot close the loop under pressure. They hedge. They A/B test what shouldn't be A/B tested. They defer to consensus.
Taste that cannot commit is just opinion. Opinions are valuable in a discussion. They are useless in a launch.
The move at this register is to protect decision capacity. Kill the meetings that dilute. Name a single decision owner. Make decision rights explicit. A small number of people with real authority beats a large number of people with partial opinions, every time.
Every material has a grain. Push it wrong and it tears. Push it right and it opens for you.
Large language models have grain too. They are confident where they should not be, vague where specificity would help, and prone to producing plausible-sounding noise at the exact moments that demand precision. Craft, at this register, means designing with that grain instead of against it. Knowing where the material will fail and putting guard rails there. Writing response patterns that force structure onto the output, because left alone the model will ramble.
Taste without craft is an op-ed about how things should be. Craft is the op-ed made ship-able.
Craft is the micro-attention layer. The polish layer. The place where taste either shows up in the product or gets lost in handoff. It is, frankly, the part of the work that does not photograph well. And so it is the part that gets cut first under schedule pressure. Resist that. The edge that separates memorable products from competent ones almost always lives here.
Here is where the existing literature on taste quietly ends. Most essays on taste, and most of them are excellent, stop at craft. They describe a tasteful person. They describe a tasteful product. They do not describe a tasteful organization, and the reason is that an individual can be tasteful by exercise but an organization is tasteful only by design.
The System register is infrastructure. It is the set of artifacts, rituals, and constraints that make it possible for taste to survive its originator's departure. A principles document that actually gets cited during tradeoffs. A pattern library that is genuinely used, not just maintained. A governance surface that protects the product's character against quarterly pressure to dilute it.
"We prefer clarity over delight" is a real principle. "We love great design" is wall art.
This is also where AI builders stop being generalists and become institutions. The System question is not just what features should we ship? It is which problems deserve durable infrastructure? Which patterns are worth pre-deciding so that your next ten decisions can happen fast.
The final register, the ceiling of the architecture, is transmission. It is the capacity to cultivate taste in other people. Not through lectures. Through proximity. Through ritual. Through the accumulated weight of watching someone with strong taste decide, out loud, in the room with you.
A leader at this register designs the conditions under which taste compounds automatically. They turn design crits into decision meetings. They turn engineering reviews into architecture conversations. They turn product planning into tradeoff-naming sessions. They put their strongest specialists into rooms they have never been in before. A staff engineer sitting with a product marketer. A PM watching a design critique run live, a writer joining an engineering postmortem. Because exposure is how taste calibrates.
Taste spreads by proximity. It moves through rooms, not decks.
This is ultimately what taste-led leadership looks like. Not the person with the best eye in the building, but the leader who makes the building taste better. Through the structures they build, the decisions they protect, and the reps they create for everyone around them.
Build the conditions. The taste follows.
Taste begins in perception, gets pressure-tested at decision, ships at craft, scales at system, and replicates at transmission. Most teams stop at the first three. The real work is the last two. That's where one person's good eye becomes a whole team's instinct.