I ported Kubernetes to the browser
Points and comments are a snapshot, not live.
A partial TypeScript port of Kubernetes runs clusters entirely in the browser for educational content.
The project, webernetes, is ~140KiB gzipped and runs in the browser without compiling Kubernetes to WebAssembly. It includes ports of the kubelet, pod scheduler, kube-proxy, deployment controller, and a browser-based container runtime and container network interface. Images are defined via a TypeScript API rather than pulled from registries. Almost all ~100,000 lines of code were authored by LLMs, but the author manually reviewed every line and created 204 integration tests that run identically against a k3s cluster and webernetes, plus 1,855 unit tests ported from the Go codebase. The project uses 2 months of effort and is intended for interactive Kubernetes educational content, not production use.
What commenters are saying
The thread is overwhelmingly positive, with comments calling the project "really cool" and "great." Several commenters highlight the development workflow described in the article-reviewing every line, writing integration tests against a real cluster-as the correct way to use LLM-assisted engineering. One commenter notes the parallel trend of using AI to rewrite existing systems in new languages, citing projects like Bun, Flow, and the GP's own pgrust. A minor discussion considers whether costs for closed models may rise, but others counter that open-weight models and local inference keep prices competitive. One comment mentions the author's earlier load-balancing and queueing explainers as related prior art.