For decades, the highest-paid creative professionals were the ones who could see what good looks like. In the AI era, that's table stakes. The new edge is being able to articulate that taste in enough detail for an LLM to systematize it.
I've been thinking about the gap between a 70% AI output and a 95% AI output. It's never one or two big things. It's a hundred small judgments. The em-dash you'd never use. The line break you'd insist on. The opening word you'd cut because it sounds AI-written. The slide layout you'd reject because the eye doesn't know where to land.
Most professionals make these calls instinctively. That's what we've always called taste. What's changed is that taste alone is no longer the differentiator. The new differentiator is two skills running in a loop:
The second one is brutal. Every founder I know who has tried to automate themselves has hit the same wall. You sit down to write the prompt and realize you've been making decisions for years that you've never had to put into words. Externalizing those decisions is the new craft.
It's the same reason elite athletes often make mediocre coaches. Doing it well and explaining how you do it well are two genuinely different skills. AI is the first technology that collapses them into the same job description and prices the second one accordingly.
The people building standout products right now don't have the best taste. They have taste they've learned to compress into instructions, iterate on those instructions, and tighten the loop every cycle.