Taste is triage
Taste is triage. I'll earn that line by the end, but here is the claim it stands for. Once making something costs almost nothing, the value moves from making to judgment: deciding what is worth building, and being able to say why.
This is a story about vibe coding, which is the practice of building software by feel. You let the machine generate, and you steer by taste instead of by hand. It is also the story of my career, because those two things turned out to be the same, about fifteen years apart.
I started as a graphic designer, and I learned to code the way you learn to swim, by getting thrown in. No computer science, no plan, no spec. I would open a file, change a value, refresh, look at it, and change it again, tweaking until it felt right. That was the entire method, and it worked well enough to get me a job in tech as a designer who could make the thing move.
A professor from those years trained the reflex I still run on. In a crit, if someone said they liked something, he would stop them cold. Why. Why do you like it. A cow can like something. Tell me why, or why not. Liking is free; the value is in the because. He was not teaching me taste so much as teaching me to interrogate it.
Then I got good at the job, and the job got serious: real clients, real budgets, real consequences. Somewhere in there, the part I loved went quiet. When you go professional you get a plan, a brief, a spec, a scope, and those exist for a good reason, to make you accountable to someone who is not you. But a plan also ends the work at somebody else's satisfaction, and it makes you justify every choice on function, which is exactly where feel goes to die. You stop saying "because it felt right," because that is not an answer you can defend in a status meeting. For years I thought I had simply grown up. What actually happened is that I traded the part of the work I loved for the part I could bill. Feeling your way costs a week nobody wants to pay for, so the plan won every time, and the plan was right to win. That is the part that stings.
Then the cost of a wrong guess collapsed. Code got cheap. A bad idea used to cost me a week; now it costs an evening. Cheap failure handed me something better than speed: permission to guess again, to feel my way, to leave the dice live and see what came out. I started tweaking again, the same refresh-look-tweak I did fifteen years ago in a job I was probably underqualified for.
Around then the industry handed me a word for it. Vibe coding. When Andrej Karpathy coined the term, he meant something close to surrender: forget the code exists, accept whatever the machine gives you, do not read the diffs, fine for a throwaway weekend project. By that definition, what I do all day is not vibe coding at all. I read everything, and I own the outcome.
The worry underneath the word is real, and I will not pretend otherwise. If you are shipping to production, someone has to understand the code, or you are just generating liabilities faster than before. But here is what the caution misses. The machine made code cheap and it made options cheap; the one thing it did not make cheap is human attention. You cannot read every diff anymore, because there are too many of them, arriving too fast. So you go on feeling. Feeling does not replace judgment; it is how you decide where to spend the judgment you have. In an age of speed, that is not indulgent. It is closer to mandatory. That is the because, and it is what the anchor line means: taste is triage, the discipline of deciding what deserves your attention when you cannot give it to everything.
The word grew up the same way I did, only faster. It went from surrender to craft. The industry made that trip in about eighteen months and put it on a conference slide; I needed fifteen years for the same distance, from art, to discipline, to clawing the art back without giving up the discipline. So when I watch people argue about whether any of this is real engineering or just confident slop, I recognize the argument. It is the one I had with myself for a decade. Am I a designer, or a guy who cannot stop fiddling. The answer was yes, and eventually that becomes a job title.
That title has AX in it now. Agent experience. For most of my career the discipline was UX, user experience, the craft of designing how people move through software. AX is the same instinct pointed at a new subject: designing how autonomous systems behave, and how people stay in control of them as the systems get smarter. (If you want that shift on its own, I wrote about it in When UX becomes AX.) What AX means day to day is that I get paid to feel my way through complex systems and be accountable for where they land.
One example, and only one. On Friday nights I started building a system I named Genie, to claw back my own creative time by handing an agent the routine parts of delivery work so my hours could go to judgment instead. It began as a personal experiment and got out of hand; my teams use it now. It grew an order of magnitude larger than where it started, and the size is not a sign of mess. Every dead end is kept as a note rather than thrown away, so what is there is everything I learned, still in the room. The only question I asked the whole time was whether it felt too complex, because the moment it did, I could not feel my way through it anymore. Early on I wired every part by hand and held the whole machine in my head. I cannot do that now, and that is the point: it became a layer of plain files and scripts, one system that holds the complexity so I do not have to. I do not wield it anymore; I feel the output and judge the result. Humans own the judgment, and the machine carries the volume. (The longer version of that story is in Building Genie changed me.)
People ask how you know when to stop, not when the ticket closes but when the thing is actually done. The fashion designer Coco Chanel is said to have kept a rule: before you leave the house, take one thing off. That is the whole answer. Requirements only ever add to a thing; taste is the only force that subtracts. Good vibes, in the end, are just taste telling you there is one thing too many on the page.
So. Taste is triage. You cannot read every line anymore, and neither can anyone else, so you feel your way and you judge the output, and whether that comes out as craft or as slop depends on one thing: whether you can say why. A cow can like something. You are not a cow. That is the job now.
This is the thinking I'm putting out in the open as I work through it. If it's the kind of argument you want to follow, come along.
