September 2026 Summaries

Ted Nelson’s Project Xanadu envisioned a “docuverse” in which documents retained every version, quotations were referenced rather than copied, and durable links connected precise spans of content, but its ambitions were limited by expensive storage, missing naming systems, and the web’s preference for simpler, breakable links. The author argues that AI agents make this xanalogical model newly practical because they can inspect extensive histories, sources, discussions, and code context that humans cannot continuously retain. Delta and its underlying DeltaDB are presented as an implementation of related principles for software development, preserving conversations, code edits, provenance, and stable fragment-level anchors in a shared history rather than flattening work into isolated commits. The system draws on now-mature technologies including Lamport timestamps, Merkle trees, CRDTs, inexpensive storage, broadband replication, microVMs, and fast incremental parsing to support concurrent editing and permanent, connected records. Unlike Xanadu, Delta is designed to interoperate with existing Git repositories, branches, GitHub workflows, and agent tools, aiming to combine live collaboration with a durable and auditable record of how code and decisions evolved.