Product taste, the most important skill in modern product companies?

A friend at a Series A asked me last month what I look for when I hire a product manager in 2026. I told him the truth — product taste, more than anything else.

Five years ago that answer would have sounded squishy. Today it's the most concrete thing I can name.

Here's why. The cost of building a feature has fallen off a cliff. We talk about this in the abstract — AI-assisted development, prototype-to-production in a weekend. But the actual lived experience is: a problem you would have scoped as a quarter-long engineering project in 2022, you can prototype on a Thursday afternoon now. The functional output has compressed by an order of magnitude.

What hasn't compressed is the question of what to build.

The decision cost stayed constant

Pick the wrong feature and the time saved is irrelevant — you've still built the wrong thing. Pick the right feature for the wrong customer segment, same outcome. Pick the right thing for the right segment but at the wrong moment in their workflow, and you've built something people will use once and never again.

These decisions used to be diluted across a long build cycle. The cost of deciding was small compared to the cost of executing. When execution gets fast, every bad decision is exposed faster too. The bar moves up.

Product taste is the pattern-matching machine that lets you make those decisions quickly and well. It's accumulated by shipping a lot of things and watching what happens — what users actually do, what they say they'll do, what your engineers grumble about, what makes a sales call go quiet. You can't buy it, fake it, or replace it with a methodology. You build it the way mechanics build car-sense: by working under enough hoods.

What I think happens next

The PMs who get hired into the most ambitious roles five years from now will not be the most process-disciplined, the most technical, or the most data-literate. They'll be the ones who can hear three product ideas in a meeting and quickly tell you which of them has legs — and why — in a way that sounds obviously right after they say it.

Taste is also exactly the thing AI can't substitute for, because taste is built on context. Your specific market, your specific team, your specific moment. The models are general; your product can't be.

If you're starting out in product: ship as much as you can, hold opinions as openly as you can, and accept that some of your calls are going to look stupid in hindsight. That's the gym. There's no shortcut.