Karri Saarinen, the CEO of Linear, gave a talk at Figma Config that I’ve been thinking about since. His argument: technology makes it easier to build, but harder to care about the craft. We replaced purpose with metrics. We delegated judgment to A/B testing.
He’s right about the observation. I think he got the cause wrong.
Technology has always been deflationary. More with less. That’s what it does: the deflation is the feature, not the bug. The problem isn’t that better tools exist. The problem is what happens in a hyper-competitive market when the barrier to building drops toward zero.
VCs demand volume, scale, and speed. Companies race to ship before competitors do. AI makes the race faster. And in a fast race, the things that take time get squeezed out first: judgment, iteration, care, refinement.
Most teams end up with functional interfaces. Competent output. Nothing you’d remember.
Quality becomes differentiation precisely because craft gets scarce when production gets cheap. Which makes the teams that hold onto it under pressure the interesting ones to study. Karri’s own company is the clearest example I know.
Competent is no longer the craft
For most of design’s history, producing something that looked professional required years of tool mastery. That scarcity was the craft. Getting to competent was the barrier, and the barrier created the value.
The floor has moved. The number of people who can produce decent-looking design — not exceptional, not memorable, but good enough to ship — is nearly vertical now. An afternoon with v0, Cursor, or any current generation tool will get you there. What used to require months of Figma fluency is now a prompt away.
This changes what exceptional actually means.
The traditional T-shaped designer, deep in UI craft and broad in adjacent skills, is getting much broader. Not because anyone decided this was strategically wise. Because the work demands it. When I write a PRD, break it into GitHub issues, build the front end with Claude Code, run the branch locally for user testing, and orchestrate QA through agentic tests and CodeRabbit, I’m not doing a developer’s job. I’m doing a designer’s job in 2026, where owning the outcome means owning the full loop.
The “stay in your lane” instinct made sense when lanes were hard to cross. Switching required months of ramp-up, and discipline required constraint. But most of those crossings are cheaper now. You can operate in parts of the stack that were previously someone else’s territory, not because the work got easier, but because the tools handle the parts that required rote execution rather than judgment.
What you still can’t delegate is knowing what to do. An AI will generate a hundred dashboard variations. It won’t tell you that your users don’t actually want a dashboard. They want relevant data pushed to them before they know to ask.
That’s not a prompt. That’s judgment built from time spent watching real people use real things under pressure.
That’s what taste protects. And it’s the only part of this job AI hasn’t touched.
Also in May: About Craft, Taste, and Speed · What Linear Did with Taste · Restraint Is a Taste Decision
Daniel Arevalo is Product Experience Lead at Omny Security. He builds AI-native tools and writes about design, product leadership, and what it means to direct systems that execute at a speed humans never had before.