Chapter Ten
Taste is
the Product
AI generates options. You choose
the right one. That's the skill.
AI can generate a website in thirty seconds. Most of them look the same: rounded corners, gradient backgrounds, generic hero sections, placeholder stock phrases. This is AI slop — the default output when nobody exercises taste. It's technically functional. It's aesthetically forgettable. And it's what you get when you accept whatever the AI gives you without pushing for something better.
The antidote isn't more prompting. It's having standards. Knowing what good looks like. And showing the AI what you want instead of hoping it guesses right.
References Over Descriptions
"Make it look modern and clean" tells the AI almost nothing. "Make it look like stripe.com — minimal, lots of whitespace, one accent color, big bold headings" tells it everything. References beat descriptions every time. A screenshot is worth a thousand words of prompting. Collect examples of things you love — websites, apps, designs — and feed them to the agent. "I want this layout but with these colors." "This typography but applied to a dashboard."
Taste isn't innate. It's exposure plus judgment. You develop it by looking at great things and understanding why they're great. The more references you collect, the sharper your eye becomes. And the sharper your eye, the better your instructions to the agent.
The three evaluation lenses
Judge what the agent builds through three lenses. Visual: does it look like my references? Functional: does it do what I specified? User: would I actually want to use this? If the answer to any of these is no, iterate. You don't need to read the code to apply these lenses.
The Human Edge
AI generates options. You choose the right one. That judgment — taste, empathy, vision — is your edge. AI can produce ten logo variants in a minute. But only you know which one feels right for your brand. AI can draft ten versions of a landing page. But only you know which one will resonate with your audience. The scarce skill isn't generation. It's curation.
What AI fundamentally can't do: choose what to build. Decide who it's for. Determine why it matters. Set the creative vision. Read the room. Know when something is "off" even though you can't articulate why. These aren't bugs in AI — they're features of being human. And in a world where generation is free, they're more valuable than ever.
Rank these options from best to worst. Then see how the experts scored them — and why.
There's one chapter left. No more theory. It's time to combine everything you've learned and build something real — something only you would think to make.