Why Is Everything Starting to Feel the Same?

I am obsessed with the gap between technological capability and human needs. I have spent my career building emerging technologies at companies like Apple, Microsoft, Dolby, and Sony, thinking about how to close that gap.

Some of my most meaningful work centered around inventing new ways for devices to understand and respond to human context, work that ultimately became part of features experienced by millions of people across different devices. One of the hardest challenges was figuring out how a system can understand what matters to someone at a particular moment. A device might know that a notification has arrived, but whether it should interrupt you, and how it should reach you, depends on context. The technology can be extraordinarily capable, but if it does not understand the user's context, and deliver at the right moment, it can still feel remarkably dumb.

More recently, that question has led me somewhere even more personal: can technology begin to understand the qualities that make us distinctly ourselves? Our preferences, sensibilities, quirks, and tastes. The things that give us personality and differentiate one person from another.

That question has become something of an obsession for me.

What is taste?

Even the word feels inadequate. Ask ten people what makes something beautiful, stylish, elegant, cool, cute, or simply 'right', and you may get ten different answers. Taste is personal, contextual, and often frustratingly difficult to articulate.

Yet we recognize it instantly when we see it.

We know when a room feels like us. We know when an outfit doesn't. We can walk into a cafe and immediately love its atmosphere without being able to explain whether it was the lighting, the architecture, the materials, the colors, or simply the feeling of the place. We can be drawn to the shape of a car, the typography on a book cover, or the packaging of a perfume before knowing anything about what is inside.

Taste is not one attribute. It is the relationship between many of them, filtered through our experiences, memories, culture, mood, and identity.

For decades, technology has largely ignored this part of human decision-making.

Technology Has Been Built Around What We Could Measure

Most digital systems were designed around attributes that could be easily quantified and categorized. Shopping platforms understand price, size, brand, color, and category. Travel sites understand destination, dates, star ratings, and budget. Real estate platforms understand square footage, bedrooms, location, and price. These are all useful signals, but they don't capture the full picture of how people actually make decisions.

Historically, subjective qualities such as mood, aesthetic identity, and taste were extraordinarily difficult for computers to represent or capture. So we built systems around the things machines could understand.

But difficult to measure does not mean insignificant.

Taste Directly Shapes Value

Taste may be subjective, but its economic impact is anything but.

Aesthetic appearance has a measurable impact on consumer decisions across industries, from real estate, where distinctive design can drive up to 20% more engagement, to fashion, where 40% cite design as an important factor, and consumer electronics, where that number rises to 56%.

Perhaps most strikingly, aesthetics can even change our perception of the exact same product. Airbnb famously doubled its revenue after improving the photography of its listings, without changing the homes themselves.

With this much influence on human behavior, the absence of taste from the way our digital systems understand us is a major blind spot and it demands attention from the tech community.

The Strange Sameness of Generative AI

Artificial intelligence is advancing at an extraordinary pace. Models can reason, write, code, generate images, interpret video, and increasingly interact with the world on our behalf.

And yet, alongside that progress, something strange is happening; a lot of AI-generated content is starting to feel the same.

Emails have the same polished cadence. AI-generated interiors converge on recognizable aesthetics. Images can be technically impressive and somehow still feel generic. Generative AI learns patterns at enormous scale. Given a prompt, a model predicts what should plausibly come next. That ability is incredibly powerful. But probability naturally creates a gravitational pull toward what is recognizable, conventional, and statistically likely. Taste often works differently.

The things that make us interesting are not always the things most people would choose. It might be the strange pair of glasses you love, the brutalist hotel you choose over conventional five-star luxury, or the slightly imperfect joke you write that an AI assistant wants to "correct."

From the perspective of optimization, these can look like outliers. From the perspective of identity, they may be the most important signals of all.

Human Identity Lives in the Tail

For years, recommendation systems have been built around patterns across populations: people who bought this also bought that; people who watched this enjoyed that. That approach has created enormous economic value, but personalization still falls short. This becomes more important as generative AI moves from a tool we occasionally use to an interface through which we make decisions.

Imagine asking an AI agent to find you a hotel, furnish your apartment, select clothes, recommend a restaurant, design an invitation, or plan a vacation. Knowing that you want a hotel in Rome for under $500 a night is useful context. Knowing what kind of place makes you feel something is a different level of understanding.

This is the question that ultimately led me to build meCore. We are exploring how taste can become a form of context that people can express and control, and that AI systems can use to understand us better.

As Intelligence Becomes Abundant, Context Becomes the Differentiator

Applications can increasingly draw upon models capable of similar classes of reasoning, generation, and multimodal understanding. What differentiates the experience is context: the same highly capable model can produce dramatically different results depending on what it knows about you, your world, and your taste.

Today, much of that context is factual or behavioral: your location, previous conversations, purchase history, calendar, budget, clicks, or browsing activity. But there is another category of context we have barely begun to capture: personal sensibility.

What do you find attractive? What feels boring to you? What kinds of environments energize you? What do you actively dislike, even when everyone else seems to love it?

These preferences shape decisions across nearly every part of our lives, from what we buy and wear to where we travel, whom we date, the events we attend, the food we eat, and the entertainment we choose. Yet very little of this rich, deeply personal context can be easily communicated to AI systems in a form they can act on today. This information is messy. It changes. It can be contradictory and remarkably hard to articulate.

And for the first time, multimodal AI may give us the tools to begin understanding it. Modern models can interpret images and language together, recognize subtle visual relationships, describe moods and aesthetics, and connect concepts that previously lived in completely separate databases.

At meCore, we believe people should be part of teaching AI their taste. But rather than expecting them to find the perfect prompt or keywords, or asking them to answer 30 onboarding questions, we bring them into a visual, playful journey that lets them actively shape how a system understands them. Our approach is based on a simple principle: when it comes to taste, you often know it when you see it.

The Next Generation of AI Should Make Us More Ourselves

There is an understandable temptation in AI to optimize everything: make the email more professional, the room more beautiful, the outfit more fashionable, the sentence clearer, the recommendation more likely to appeal to the average person. But optimization always raises another question: optimized toward whose definition of better?

If every system learns from similar data, uses similar models, and optimizes toward similar measures of quality, we risk building an extraordinarily intelligent machine for producing the average of everything. AI systems shouldn't tell us what the average person would choose.

They should understand why you wouldn't.

The opportunity ahead is much more interesting. We can give people greater control over how machines perceive them, rather than quietly inferring an identity from their purchasing history. The next generation of machine intelligence should help us discover, articulate, and express the parts of ourselves that live in the tails.

As AI becomes ubiquitous, intelligence will increasingly be abundant. Individuality will not.