Discussion about this post

User's avatar
Steven Vincent's avatar

"There's a world of difference between a company giving you more control within its platform, and you having control independently of that platform." That's the line I'd underline in this whole series.

Sentinel is genuinely good engineering. Separating the intelligence that decides from the permission that allows, and keeping approvals out of the agent's own conversation, is exactly right. But a permission layer that lives inside Meta's cloud is still Meta's permission layer.

In a recent piece on Microsoft's shift beyond the model wars, I put a few simple tests to any "personal" AI. Can you change models without losing your accumulated knowledge? Can you move to another application? Can you decide what leaves your device? Your portable-permissions question belongs right alongside those. As I put it there, the point is that "the person should not disappear inside the infrastructure."

That's the heart of what I call Human-Machine Intelligence: the agent extends your agency, and the rules travel with you, not with the vendor. Looking forward to Part 3 on delegation vs. impersonation.

https://www.thesingularityproject.ai/p/microsoft-moves-beyond-the-model

Ready for more?