Local AI Tooling Proliferates, Trust Is the Gap
LM Studio launched LM Studio Bionic, a new AI agent product for running open models locally. The thread immediately surfaced two issues: a commenter asking what makes it different from 'just another harness,' and a reminder that both LM Studio apps are closed source, which many users apparently did not know.
The closed source point landed hard. For a tool positioned as the interface for running open models locally, being closed source is a contradiction that the community noticed. The other observation in the thread was prescient: Apple will eventually ship good-enough local models with good-enough tooling, and most normal users will just use that instead.
This is the real tension in the local AI space right now. There are many harnesses, many launchers, many agent wrappers. The differentiation is thin and the trust problem is real. Users want open source for the same reason they want local models: control and auditability.
So what?
If you are building local AI tooling, being closed source is now a genuine liability, not just a philosophical disagreement. The community that cares most about local models is also the community that reads license files. Open sourcing your core product is probably table stakes if you want developer trust in this space. And if Apple ships a credible local model experience, that compresses the window for third-party tools significantly.