You walk into a room and something is off. The furniture is fine. The colors are fine. Everything works. But the room has no soul.
You cannot point to the broken thing. There is no broken thing. It just does not feel right. Someone assembled this from a catalogue without understanding how people actually live in a space.
That is what happens when you build without taste.
Taste Is Not Subjective
People treat taste like personal preference. Chocolate or vanilla. Blue or green. But that is not what taste is.
Taste is a navigation system. It tells you what fits. What clashes. What the context demands. What to add and—more importantly—what to leave out.
A chef tasting a sauce is not asking “do I like this?” They are asking: does this work for this dish, for this guest, at this moment? Is it balanced? Does it need acid? Would it overwhelm what comes next? The Germans call it Abschmecken—tasting to adjust, not tasting to enjoy.
An architect choosing materials for a hallway is not picking favorites. They are reading the building: the light, the traffic, the purpose, the budget. A marble floor in a startup office is bad taste. Not because marble is bad—because it is wrong here.
A designer laying out a dashboard is not making things pretty. They are deciding what the user needs to see first, what can wait, what should not be there at all. Good taste in design means the user never notices the design—they just find what they need.
Taste operates across every creative domain. Cooking, interiors, typography, music, fashion, architecture, software. The medium changes. The skill is the same: judgment under context.
Why Taste Matters Now
AI can write code. Any code. Fast. The marginal cost of implementation is approaching zero. You can spin up a prototype in hours. You can rewrite a module overnight. The mechanical part of building software is essentially free.
And yet most AI-generated products are bland.
They work. They run. They meet the spec. But they do not land. They are technically correct and experientially empty—like a room furnished from a catalogue by someone who has never lived in one.
What is missing is not engineering. What is missing is taste.
When production is cheap, the bottleneck moves upstream. The bottleneck is now the person who decides what gets built, for whom, and why. The person who looks at the output and knows—before any user test, before any metric—whether this is right or wrong.
What Taste Looks Like in Product
Taste is not one skill. It is the integration of many:
- Knowing what to build. Not everything that is technically possible. The one thing that matters right now.
- Knowing for whom. Not an abstract persona. A real person with a real problem at a real time. If you would not eat your own cooking, do not serve it to others.
- Knowing what to leave out. The feature you did not build is often the best product decision you made.
- Knowing when good enough is good enough. And when it absolutely is not.
- Knowing how to present it. A great feature buried in a confusing interface is a beautiful meal on a paper plate.
- Knowing the context. Is this a quick weeknight dinner or a holiday feast? A startup MVP or enterprise software for a regulated industry? The answer changes everything.
A technical founder has this. Not because they read a book about product sense—because they have shipped enough products, watched enough people use them, and adjusted enough times. They have been wrong often enough that their compass recalibrated.
The New Role Is Not “Developer”
The industry is still hiring “AI developers” and listing frameworks that will be legacy in twelve months. Job descriptions read like shopping lists: LangChain, Semantic Kernel, CrewAI, DORA metrics.
The role that actually matters in 2026 looks different. It looks like a technical founder who happens to lead with AI. Someone who walks into a customer conversation, understands the problem, goes back to the workshop, directs the AI agents, evaluates the output, adjusts, and ships something that people want to use.
Someone who can furnish the room—not just assemble the furniture.
The market is starting to call this an “AI Product Engineer.” I think that is close. But what it really is: someone with taste.
They are rare. Because taste requires repetition. You cannot prompt your way to judgment. You build it by doing the work—shipping, watching, adjusting, shipping again. By getting it wrong and learning what right feels like. Across domains, across contexts, across years.
The Test
There is a moment—every craftsperson knows it—where you step back and look at what you made. Not to admire it. To evaluate it. Does this work? Does it fit? Is it right for the situation?
In cooking, that moment is the spoon before serving. In design, it is the prototype in someone’s hands. In product, it is the feature in front of a real user.
That moment cannot be automated. It cannot be captured in a prompt. It lives in the person who has developed taste through years of doing the work.
I am building a team at ecovium that works this way—AI as the engine, human taste as the compass. Currently hiring a Senior AI Product Engineer who does not just build what the spec says, but knows what the product needs.
If that resonates, reach out.
And I am genuinely curious: What is the last product decision where you trusted your taste over the data—and turned out to be right?