AI-augmented knowledge environments
An exploration of local and cloud applications that help organisations capture knowledge, connect experience, and work with what they know in AI-augmented environments.
Why this interests me now
Organisational knowledge often sits in documents, conversations, systems, and individual memory without an easy way to see how the pieces relate. I’m interested in applications that help people capture that knowledge, give it useful structure, and return to it as their understanding develops.
The aim is not to make AI an unquestioned source of answers. I want to explore environments in which people can retrieve, connect, challenge, and reinterpret what an organisation knows while remaining responsible for meaning and decisions.
Local and cloud explorations
MySotongkoo explores the local side of this interest: a focused AI interface for working with information and learning. Cloudflare-first solutions gives me a cloud environment for prototyping and publishing applications. Together they let me investigate where knowledge should live, how people encounter it, and what different technical boundaries make possible.
Content Studio
The Content Studio is a local authoring environment for this knowledge system. It is built with Vue 3 and Vite in the interface, Fastify and TypeScript on the server, and the existing Markdown, YAML, Liquid, Jekyll, and Git workflow underneath. CodeMirror and a simple visual editor support writing; the rendered site remains the preview and the files remain the source of truth.
Its central idea is to put deterministic tools before model inference. Search, filtering, relationship validation, graph checks, diffs, atomicity diagnostics, and link ranking are inspectable operations. Codex or Hermes through local Ollama can then be asked manually to review a shortlist or explain a possible split, merge, or link. The model adds interpretation to a bounded result; it does not silently edit, publish, or replace the evidence.
The atomic content studio records how this authoring model is evolving around focused Cards, structural Pages, explainable links, and deliberately triggered AI assistance.
From a shared brain to this website
The inquiry grows from Making a shared brain with my GenAI buddy, where I explored small, linked notes with a generative-AI collaborator. This website develops that idea into a visible system: its connectome turns connected notes about my practice into a map that people—and AI tools—can navigate.
Questions I’m exploring
- What makes captured knowledge useful in everyday organisational work rather than merely stored?
- How can people see where knowledge came from, question it, and add context as situations change?
- Which parts of an AI-augmented knowledge environment belong locally, and which benefit from cloud access?
- How can an interface reveal relationships without pretending that the map is complete?
Related cards
- Making a shared brain with my GenAI buddy — The shared-brain experiment is the originating idea for organising knowledge as small, connected notes with an AI collaborator.
- The connectome — This website’s connectome demonstrates how connected knowledge can become navigable.
- MySotongkoo: an AI infobutler — MySotongkoo explores a local-AI application for retrieving, questioning, and working with information.
- Cloudflare-first solutions — Cloudflare provides a cloud environment in which I can prototype and publish related applications.
- Organisations and systems — The inquiry connects knowledge tools with how information and learning move through organisations.
- Connected notes — Review the lexical and graph proximity between AI-augmented knowledge environments and Connected notes.