Earlier this week I linked to Gale Robins’ argument that AI makes execution cheap but doesn’t help you decide what to build. Christina Wodtke is making the same case from the research side.

Christina Wodtke opens with a designer who spent two weeks vibe-coding a gratitude journaling app. Beautiful interface, confetti animations, gentle notifications. Then she showed it to users. “I don’t really journal,” said the first one. “Gratitude journaling felt performative,” said the second. Two weeks building the wrong thing. Wodtke’s diagnosis:

That satisfaction is a trap. You’re accumulating artifacts that may have nothing to do with what anyone needs.

Wodtke draws a line between need-finding and validation that I think a lot of teams blur. Skipping the first and jumping to the second means you’re testing your guess, not understanding the problem:

Need-finding happens before you have a solution. You’re listening to people describe their lives, their frustrations, their workarounds. You’re hunting for problems that actually exist—problems people care enough about that they’re already trying to solve them with spreadsheets and sticky notes and whatever else they’ve cobbled together.

Wodtke’s version of fast looks different from what you’d expect:

The actual fast path is unsexy: sit down with five to ten people. Ask them about their lives. Shut up and listen. Use those three magic words—“tell me more”—every time something interesting surfaces. Don’t show them anything. Don’t pitch. Just listen.

“You’ll build less. It’ll be the right thing.” When building is cheap, the bottleneck moves upstream to judgment, knowing what to build. That judgment comes from listening, not prompting.