Personal AI memory needs portability and permission controls
AI summary
In an O’Reilly Radar interview, software engineer Duncan Davidson argues that AI services often retain conflicting versions of the same user, fragmenting personal context. He proposes shared mechanisms that let people carry context across services, grant access by scope, and approve requests before information is shared. The conversation also examined AI-assisted web development, developer training, and the strain AI-generated contributions place on open source review.
In-depth insight
The interview identifies a practical gap in current AI memory systems: user context is fragmented across services and lacks consistent, fine-grained permission controls.