E-means claims 10x faster k-means clustering at equal loss

A project called E-means has been shared on Hacker News, asserting that its k-means clustering method runs about ten times faster than standard approaches while reaching the same loss. The post links to a dedicated project page describing the approach. The only details available are the headline, the project URL, and a small amount of community engagement, with three points and no comments at the time of posting. The article text provides no methodology, benchmarks, or dataset specifics to independently verify the speed claim. Readers would need to consult the project page for technical details on how the performance gain is achieved.