Updated
Updated · O'Reilly Media · Aug 11
Author Prototypes 1 User-Controlled AI Context App in Apple Ecosystem as Agent Lock-In Grows
Updated
Updated · O'Reilly Media · Aug 11

Author Prototypes 1 User-Controlled AI Context App in Apple Ecosystem as Agent Lock-In Grows

3 articles · Updated · O'Reilly Media · Aug 11

Summary

  • A SwiftUI prototype is testing whether a user-controlled personal context repository can live inside Apple’s ecosystem, with iCloud syncing data and MCP exposing selected context to authenticated AI agents.
  • The design aims to stop users from rebuilding preferences and history across multiple agents by keeping a canonical, inspectable record under user control rather than inside each vendor’s silo.
  • After trying Markdown files on a laptop and a web service behind a URL, the author concluded a phone-centered ecosystem offers the most practical trust boundary for identity, storage, sync and native integration.
  • The prototype still leaves the hardest problem unsolved: negotiation over who can read, propose, change or delete parts of context, while preserving provenance, scope and time-based history.
  • The broader thesis is that personal context should be portable and exportable as plain text, so users can switch agents without losing the durable record those agents help build.

Insights

Will managing your own AI context give you ultimate freedom, or become a complex security nightmare you cannot escape?
Why are tech giants quietly allowing your AI assistants to trap your digital identity inside their walled gardens?
If a malicious agent poisons your centralized AI memory, could it silently manipulate every digital assistant you use?