Why “open weights” still aren’t the same thing as open source

aronchick · x · 2026-07-28

The post argues that calling model weights “open” is not the same thing as calling software open source.
It traces the term “open source” back to 1998, when it was coined to distinguish source code access from merely free-to-download binaries.
The essay’s core point is that open weights may be useful, but they still need a clearer category: what does a download actually give you, and what rights or abilities come with it?
It frames the current debate as a category problem, not just a pricing problem, and references recent arguments that open-weight AI may be entering its own Kubernetes-like moment.

More from Companies & People