25 March 2026
A first-principles approach to design in the age of AI
Most teams are trying to fit AI into their existing design process, keeping the same roles, handoffs, and structure, just asking AI to make each step faster. But AI fundamentally changes the economics of making software, and the real bottleneck has shifted.
Main point: Most teams are bolting AI onto their old design process. But AI makes code cheap, so the real bottleneck is now coordination and handoffs, and the process should change to match.
The mistake
When working code gets cheap, the expensive part is moving intent between tools and specialist roles.
A first-principles workflow
Design teams want to explore quickly, validate the hard parts early, and ship without losing intent. Today's tools are good at exploring but weak at validating: a static file can't tell you whether an interaction breaks with real data or permissions. So treat working code as a design material, run safely against real constraints with shared components and guardrails. The aim isn't to make designers engineers. It's to let them be right about the hard things before engineering commits.
What changes about teams
Less translation work, fewer bridging roles, and less engineering time spent transferring intent. Engineering becomes more valuable, focused on security, performance, reliability, deployment, data contracts and observability.
The enabler
This needs a safe place to build. That's Unplanned Studio: a sandbox where designers build real React apps on shared infrastructure, deploy them, and run quick usability sessions before anything enters the backlog.
Where this goes
Smaller teams, fewer silos, less coordination tax, more time on decisions that matter.