@antirez
Post
Fast H3 implementation for Metal. Enjoy, modify, and so forth: https://github.com/antirez/h3.c Contains code from @liuliu which is welcomed in taking back whatever parts he likes for @drawthingsapp in case there are H3 plans there.
Explanation
What it says @antirez released a fast H3 implementation for Apple Metal as `h3.c`, inviting others to use and modify it. He notes that it incorporates code from @liuliu and explicitly welcomes liuliu reusing any relevant parts in Draw Things if that project adds H3 support.
Context The post is essentially a code-release announcement, not an explanation of H3 itself. From the saved material alone, it is unclear exactly which “H3” model/architecture is meant, what hardware was benchmarked, or how performance compares with existing implementations. The mention of Metal and Draw Things strongly situates the work around GPU-accelerated execution on Apple hardware, but the technical details are not included here.
Why it matters The interesting part is the implementation layer: a fast, relatively direct Metal implementation could make H3 substantially more practical on Macs/iOS devices and potentially be incorporated into applications such as Draw Things. It is worth returning to if you care about local model inference, Apple-GPU optimization, or minimal native implementations. The key things to inspect later are what H3 specifically refers to, benchmark results, supported architectures/models, memory requirements, and how much of the speed comes from Metal-specific optimization versus algorithmic changes.