Redfin Company Growth, Stability & Outlook

Rocket’s engineering team is building systems designed to stay consistent and reliable as the company’s technology grows. By turning architectural knowledge into reusable, version-controlled rules that can work across teams and programming languages, engineers can preserve lessons learned, reduce repeated work and maintain standards even as individual employees and teams change over time.

“Today we maintain dozens of architectural rules comprising thousands of lines of deterministic enforcement, running in the editor, in pull requests, and in CI. Since adopting this approach, the volume of convention-based review comments has materially decreased. The question that now comes up naturally in design reviews and retrospectives: should we write a lint rule for that?

We're thinking about what this looks like at larger scale: rule sets that enforce standards across languages, architecture-as-code libraries, shared rule registries where teams build on each other's enforcement work rather than starting from scratch.

The harder and more interesting problem is organizational memory: the hard-won knowledge of what goes wrong and why, encoded in rules that outlast any individual engineer.

AI accelerates rule authoring, but enforcement remains deterministic, version-controlled, peer-reviewed, and auditable. The combination of human judgment and machine discipline is what makes this more than a productivity improvement. It's a more honest and more scalable way to build software.”