Growth, and the design systems that carry it
Growth results and the systems behind them at Raptive, Nextdoor and Zillow Group, a screen followed end to end through the system, and how it grew into Design Studio.
Growth finds the leak. The system makes the fix ship everywhere and stay fixed. Each company, the growth result and the system behind it.
The whole system, put together. The AI reads the rules, builds from the system’s own pieces and runs the checks. It fixes what fails, then a checker that did not do the work looks again. A person approves, what ships is built from the same pieces, and a person improves the rules.
A harness is everything around the AI that keeps its work honest: the rules it reads, the checks it has to pass, and the person with the final say.
The same checks, on a single piece. Three ways a button can fail, what the check found, and what fixed it.
The environment came first: every token, component and pattern running in a browser, with contracts and tests checking them on every pull request.
The harness grew around it, so AI could build on the system without leaving it: the rules it reads, the checks it has to pass, and a person with the final say.
Design Studio is where that is used day to day. A person describes what is needed, agents draw a few looks in characters, a person picks one, and a builder makes it from the kit’s own components while a checker that did not make it looks again. What each agent is doing shows in the chat and on every page. None of them decides: a person does.
A design system that runs as a live environment and is checked against production code, so design, product and engineering build from one source.
Every token, component, pattern and prototype, running in a browser for anyone on the team to open. Internal to Raptive, so there is no public link.
The design system lived only in a Figma file, so nobody could see it working. One designer, 12 to 15 engineers and two product managers each built their own prototypes. The platform's code dates from 2013, before design tokens existed, and a new look was coming to more than 24 communities. It had to land everywhere at once.
Not a library problem. A visibility problem: nobody could see the system working, and nobody could say what a change to it would reach.
Proven: the values match production, the contracts and tests run on every pull request, and the migration order rests on measured use rather than on preference. Not yet proven: the built light and dark themes have not been checked, nothing in the build stops a hard-coded value getting in, and whether the system makes teams faster is unmeasured.
The new look is not live yet and the work is ongoing. Sign off on the third round comes next, then the swap.
The same rule runs through all of it: measure the code, rank by evidence, write down what is unproven.
Home About Using AI Ask the system Raptive Onboarding Nextdoor Faves Nextdoor Business Design system Trulia Synthesizer Alfred CLI Resume