The real AI moat is taste

GPT-5.4, Claude, Gemini 3. Pick your favorite. They can all write code, draft essays, summarize research, and generate images. The APIs are nearly identical. The tutorials are the same. The pricing is converging. If everyone has access to the same intelligence, what actually separates the winners from the noise? Not technical skill. Not access. Not speed. Taste.

The convergence is real

A few years ago, choosing an AI model felt like choosing a programming language. Each had real tradeoffs, distinct capabilities, and meaningful gaps. Today, frontier models are roughly interchangeable for most tasks. OpenAI's own docs recommend GPT-5.4 as a default, but Anthropic and Google offer comparable alternatives at similar price points. As one Forbes analysis put it, "the AI tech stack is rapidly becoming a commodity. Intelligence is turning into something like electricity, easily accessible, with the option to switch to a better or cheaper provider effortlessly." This isn't speculation anymore. DeepSeek proved that a team with a fraction of the compute budget could train models rivaling those from the best-funded labs. Microsoft built internal models that compete with OpenAI's, its own strategic partner. The moat around model performance is eroding fast. When the underlying capability is commoditized, the differentiator moves up the stack. It moves to the human layer. It moves to taste.

What taste actually means here

Taste, in this context, isn't about aesthetics or personal preference. It's the ability to make good decisions about what to build, what to cut, and what to prioritize. It's judgment applied to creation. David Okuniev, co-founder of Typeform, described it well in a recent interview: "In a world where it's just so easy to put things together and the language models are doing a lot of the heavy lifting, taste and design and being able to direct it is going to be the big differentiator." He built a full Swift app over a holiday break using AI-assisted coding. The coding was the fast part. The days were spent iterating on design, getting the feel right, deciding what belonged and what didn't. That's taste. Not the ability to generate output, but the judgment to shape it.

Why some AI output feels hollow

You've seen it. A LinkedIn post that's technically correct but says nothing. A landing page with perfect copy that somehow feels empty. A blog post that covers all the right points but reads like it was assembled from a parts bin. The issue is almost never the model. It's the person directing it. A shallow prompt produces shallow output. A vague brief produces vague results. AI reflects the quality of the input it receives, and the quality of the input depends entirely on the taste of the person providing it. A Reddit user captured this perfectly: "A tool doesn't magically erase good taste, or create it." People who write hollow content with AI would have written hollow content without it. The tool amplifies whatever judgment you bring to the table. The difference between AI-generated content that resonates and content that feels like slop is not a model parameter. It's the human who decided what was worth saying, what angle to take, and when to stop.

Taste is the builder's edge

This applies beyond writing. In product development, the apps that win aren't the most technically sophisticated. They're the ones that made better tradeoffs. Typeform didn't invent forms. It reimagined them. The one-question-at-a-time format wasn't a technical breakthrough. It was a taste decision, a choice to prioritize how something felt over how much it could do. That single design choice turned a commodity category into a beloved product. The same pattern shows up everywhere. The best products tend to have simpler scope, clearer value propositions, and fewer features done well. That's not a technical achievement. It's a taste achievement. Someone decided what to leave out, and that decision mattered more than anything they put in. With AI making it trivially easy to build functional software, this dynamic only intensifies. Okuniev's vision for a modern startup team is just three people: a design-driven founder, an engineer, and a marketer. The bottleneck isn't building anymore. It's knowing what's worth building.

Topic selection beats prose quality

For anyone creating content regularly, this principle becomes obvious fast. Writing daily (or even weekly) means the hardest part isn't crafting sentences. It's choosing what to write about. Pick a compelling topic with a clear angle, and even rough prose will land. Pick a boring topic, and no amount of polish will save it. AI can help you write faster, but it can't tell you which ideas are worth pursuing. That requires pattern recognition, cultural awareness, and a sense for what your audience actually cares about. All of which are taste. The same logic applies to product roadmaps, marketing campaigns, and hiring decisions. The upstream choice, the "what," matters more than the downstream execution, the "how." AI is rapidly commoditizing the "how." The "what" remains stubbornly human.

Taste is upstream of distribution

There's a popular startup maxim that distribution beats product. But taste is upstream of both. You can't distribute something people don't want, and knowing what people want is fundamentally a taste problem. The founders who build viral products aren't just good at marketing. They have an instinct for what will resonate. They choose categories where design improvements are visible and shareable. They build things that make people ask, "What tool is that?" That instinct isn't random. It's taste, developed through experience, feedback, and paying close attention to what works. In a world where AI can generate anything, distribution strategy matters. But the thing being distributed still has to be worth someone's attention. Taste is what ensures it is.

Taste isn't innate

Here's the encouraging part: taste is a skill, not a gift. It's developed through volume, feedback, and reflection. You build it by making things, seeing what works, noticing patterns, and adjusting. Okuniev's approach is instructive. He doesn't validate obsessively before building. He builds three or four things in a month or two and sees what sticks. That rapid iteration isn't reckless. It's how you train your judgment. Each attempt sharpens your sense for what resonates and what falls flat. The same principle applies to writing, design, product thinking, and every other creative discipline. The people with the best taste are usually the ones who've produced the most volume. They've made enough bad decisions to recognize good ones.

The human job description, rewritten

AI can generate anything. Text, code, images, music, strategies, analyses. The output is abundant and getting cheaper by the day. What AI cannot do is decide what's worth generating. It can't feel why one version of an idea lands emotionally while another falls flat. It lacks the lived experience and cultural intuition to know what's just right. As one Atlantic essay argued, AI "cannot originate style with intentionality." That's the new job description for humans in the AI era. Not to produce, but to curate. Not to execute, but to decide. Not to write the code, but to know what code is worth writing. That's taste. And for now, it's not automatable.

References