- 02 of 03
Design ops
A process only survives if the team has somewhere to put things. Six pieces that make the loop repeatable by someone who is not me.
A component library that lives in code
Every component, pattern, and interaction state lives in the same repo the engineers pull from. There is no translation layer. When someone implements a design, they are looking at the same thing the design team is.
A design system that grows out of the work
A model can draft a screen in minutes, so the first pass of any feature is deliberately loose. New ideas get their own one-off components and nobody is asked to fit the system yet. That keeps early exploration as fast as it was when the product was a start-up with nothing to protect.
Then the pattern earns its place. A component is promoted into the shared library the second time a screen needs it, and from then on the model drafts against it. The system is never designed up front. It solidifies as the product does, one repeat at a time.
One use is exploration. Two is a pattern.
A research base a model can query
Session notes go into an inbox. An AI pass turns them into findings. A small server lets Claude query the whole base from any session, so patterns surface across studies without anyone tagging them.
Review where the work is
Design review happens on the pull request, not in a separate meeting with a separate file. The AI first pass catches heuristics. I catch judgement.
Decisions written down
When the design system evolves in pull requests it is easy to move fast and forget why you moved. So the process is written down, and the why goes where the change is, in the pull request.
It is the part that slips first, which is why it gets a rule of its own.
Onboarding into the loop
Designers and engineers learn the same three things: how to brief a model, how to review what it produces, how to know when to stop.
Productive in weeks rather than months, because the process is written down and the tools are shared.