“Taste” gets invoked constantly in conversations about what AI can’t replace. But it’s often left undefined—a hand-wave toward something ineffable that separates good work from average work.

Yan Liu offers a working definition:

Product taste is the ability to quickly recognize whether something is high quality or not.

That’s useful because it frames taste as judgment, not aesthetics. Can you tell if a feature addresses a real problem? Can you sense what’s off about an AI-generated PRD even when it’s formatted correctly? Can you distinguish short-term growth tactics from long-term product health?

Liu cites Rick Rubin’s formula:

Great taste = Sensitivity × Standards

Sensitivity is how finely you perceive—noticing friction, asking why a screen exists, catching the moment something feels wrong. Standards are your internal reference system for what “good” actually looks like. Both can be trained.

This connects to something Dan Ramsden wrote in his piece on design’s value in product organizations: “taste without a rationale is just an opinion.” Liu’s framework gives taste a rationale. It’s not magic. It’s pattern recognition built through deliberate exposure and reflection.

The closing line is the one that sticks:

The real gap won’t be between those who use AI well and those who don’t. It will be between those who already know what “good” looks like before they ever open an AI tool.