# File-system access changes AI knowledge work > [!quote] > "You and your AI share the same file system, the same workspace… cross-folder synthesis [lets you] interrelate and interconnect or even merge and synthesize whole different collections of documents." The productivity jump comes from reach and memory (reading/synthesizing across whole collections), not from the model being smarter — the long-horizon bottleneck is scaffolding. ## Links - [[3. Resources/Permanent Notes/Long-Horizon AI Shifts the Bottleneck From Model Capability to Memory and Effort Scaffolding]] — shared file access is the concrete memory scaffold behind the Permanent's long-horizon claim - [[Lit · Fable 5 prompting — effort and memory]] — Anthropic's guidance names the same lever (externalized memory + effort) that shared-filesystem reach exploits - [[Personal Knowledge Management MOC]] — shared files turn the vault itself into the boundary between capture, synthesis, and delegated knowledge work ## Source - [[3. Resources/Readwise/Podcasts/Give Claude Your Files and Watch What Happens|Give Claude Your Files and Watch What Happens]]