We started designing in code
Julio, is the design ready? The fix wasn't hiring a designer. It was letting engineers design in code, behind feature flags — because code got cheap enough to be the sketch.
Julio, is the design ready? Same pattern, different loop. After the PRD, features waited on my screens, my flows, my opinion on every corner of the UI.
I'd sketch, spec, annotate. The team would translate my mockups into code, and then we'd argue about the difference between the two. Every iteration crossed my desk twice.
The assumption that broke
The whole handoff — design first, code later — rests on one assumption: code is expensive, so you'd better get the drawing right before you build. That assumption died. With AI, generating a working version costs about as much as describing it.
When code is cheap, the mockup is overhead. The real thing is the sketch.
What we do instead
At Riqra, engineers now design in code. They take the PRD, build the feature behind a feature flag, and present the real, clickable thing — to me and to clients. Feedback happens on the product, not on a picture of it.
If something's wrong, we don't polish the drawing. We regenerate. Code became a commodity: cheap to produce, cheap to throw away. Feature flags make that safe — nothing reaches customers until it's ready, but everything is real from day one.
Two things made it work: flags as the default for anything new, and the discipline to treat generated code as disposable until it earns its place.
What changed
Design stopped being a queue with my name on it. Clients react to real screens earlier. And the engineers own the experience end to end — they're not implementing my drawing, they're designing the product.
Founder question: how many of your features are waiting on a mockup that a working version would answer faster?