memory.wiki: A Personal Knowledge OS

memory.wiki is a platform designed to transform fragmented AI-generated insights into a persistent, AI-readable knowledge graph, positioning itself as a "Personal Knowledge OS" rather than a traditional memory database. By utilizing URL-based markdown, it enables users to maintain ownership of their intellectual output across disparate AI tools and platforms.

Key claims

  • [EXTRACTED] memory.wiki serves as an ownership layer for AI-produced knowledge, distinguishing itself from "AI developer infra" (like Mem0 or Zep) and closed-ecosystem features (like ChatGPT Memory) [doc-1].
  • [EXTRACTED] The core problem is not "AI memory" but "delivery"—the fact that valuable insights generated by AI are lost when browser tabs are closed or trapped within specific, siloed containers [doc-2].
  • [EXTRACTED] memory.wiki functions as an API-accessible URL; any LLM can fetch, read, and interpret the markdown content, bundles, or hubs without requiring custom SDKs or installations [doc-2].
  • [INFERRED] By shifting from document-centric storage to an entity-relationship model, memory.wiki enables a "Self-Model" that tracks a user's core motivations, strengths, and evolving narrative over time [doc-3].
  • [AMBIGUOUS] While memory.wiki is positioned as a "Personal Knowledge OS," its utility relies on the user's active curation of knowledge, which may face competition from established productivity tools like Notion or Obsidian if those platforms improve their AI-native accessibility [doc-1, doc-4].

Cross-references

  • AI Memory vs. Knowledge OS: [doc-1] and [doc-4] both contrast memory.wiki with "AI developer infra" (Mem0, Zep, Memori Labs). While the latter focuses on database layers for agents, memory.wiki focuses on the human-centric organization of thoughts and relationships.
  • The "Delivery" Problem: [doc-1] and [doc-2] agree that the current state of AI usage is fragmented; [doc-1] proposes specific surfaces (Chrome, iOS, VS Code) to solve this, while [doc-2] identifies the root cause as the lack of an AI-readable, platform-agnostic URL format.

Open questions / gaps

  • Interoperability: How does memory.wiki handle potential conflicts when multiple AI agents attempt to update the same knowledge graph simultaneously?
  • Privacy/Security: As a URL-based system, what mechanisms ensure that sensitive personal knowledge graphs remain accessible only to the user's authorized AI agents?
  • Scalability: How will the system maintain performance and relevance as the "Personal Knowledge Graph" grows from a few projects to years of accumulated data?

Provenance

  • [doc-1]: Defines the positioning of memory.wiki across three primary user surfaces: web, mobile, and developer environments.
  • [doc-2]: Explains the philosophical motivation behind the project, identifying "delivery" rather than "memory" as the primary failure point in current AI workflows.
  • [doc-3]: Outlines the structural requirements for a "Personal Knowledge Graph," emphasizing entity-based relationships and self-modeling.
  • [doc-4]: Compares memory.wiki to Memori Labs, clarifying the distinction between infrastructure-focused tools and knowledge-focused OS.

CONFIDENCE TAGS:

  • [EXTRACTED]
  • [INFERRED]
  • [AMBIGUOUS]