Ask five people in this industry to define "mixed reality" and you'll get five different diagrams, most of them borrowing Milgram's 1994 reality-virtuality continuum and drawing the boundary lines in a different place. I've stopped trying to find the "correct" taxonomy. I don't think one exists — and I've started to think that's fine.
The convenient fiction
The tidy version everyone learned goes: VR replaces your vision entirely, AR overlays digital content on the real world, MR sits in between where digital objects interact with real ones. It's a useful diagram for a slide. It's a much worse description of what's actually shipping. A Vive user boundary-walking around their real furniture while fully in a virtual scene is doing something that "VR" doesn't quite capture. A HoloLens user whose hologram gets correctly occluded by a real wall is doing something "AR" undersells.
Why the boundary is dissolving
The technical trend line explains why the categories won't hold. Passthrough cameras are creeping into VR headsets. Spatial mapping and occlusion are creeping into AR headsets. Every "pure VR" or "pure AR" device is quietly absorbing capabilities from the other side of Milgram's continuum, because those capabilities are just useful — regardless of which side of an academic diagram they originally lived on.
Why I think that's good news
A fuzzy boundary between VR and AR means hardware makers are optimizing for what actually helps the user rather than for which quadrant of a diagram they're supposed to occupy. As a working definition, I've landed on: mixed reality is any experience where the system understands your real environment well enough to react to it, whether or not you can see that environment directly. Under that definition, most of what's shipping today — VR included — is already mixed reality. The label fights are a distraction from the more interesting question, which is what all of this hardware, whatever we call it, actually lets us build.