For most people, the frontend is the software. It is how they apply for jobs, book medical appointments, send money, and access public services. When it is slow, confusing, or inaccessible, a real person gets stuck.
As frontend developers, we are responsible for that experience. We care about design, performance, accessibility, interactions, and the hundreds of details that decide whether software works for the people who need it.
Coding agents are starting to build the web with us, but today we mostly judge their work with unit tests. Tests can tell us whether code compiles or a function returns the expected value. They cannot tell us whether an interface makes sense, matches its design, works across devices, or excludes someone. An agent can pass every test and still produce work none of us would be proud to ship.
Models learn the average, and frontend developers know the average is not good enough. If we do nothing, models will repeat the web’s existing problems at a scale no team can review.
We build open-source tools like React Doctor and React Scan to help developers ship better frontends, and ReactBench to bring those standards to coding agents.
Our technical bet is that models will only improve at frontend by taking on harder, more complete work:
- Building an accessible production interface from a design across screen sizes
- Debugging visual, interaction, and performance failures in an unfamiliar codebase
- Migrating a product to a new framework or design system without regressions
- Taking a feature from a product brief through deployment and verifying it in production
We are looking for frontend engineers, designers, accessibility experts, and performance experts to help us teach coding agents how to build better frontends. If you think people deserve better software, we want to work with you.