we bet you've seen us beforesoftware for anything with a screen.
terminals, browsers, phones, TVs, set-top boxes. lately, the tooling that keeps AI-written code honest. nearly all of it open source.
work with uswhat we do
appsweb, mobile, and TV frontends, from the design system to the store release.
agent toolingskills, reviewers, and evals that make coding agents finish the job and show their work.
infrastructureCI runners, build machines, secrets, and deploys, all as code.
start here
ffssskills and a second-model reviewer for coding agents. one model writes the change, another reviews it, and it merges once CI is green.skillchecklint and evals for agent skills: a structural lint that needs no API key, and evals that grade what a skill makes an agent do.workspace-kitscaffolds and validates the repositories agents work from: their instructions, memory, and checks.