ai · llm · civic-tech · open-source
xCO Learning System
LLM-based organisational learning infrastructure with agent orchestration — a federated wiki with natural language as the interface for authoring and retrieval.
Most organisational knowledge dies in the format built to preserve it — a wiki nobody updates, a doc nobody re-reads, a Slack thread that was the actual decision. xCO is an attempt to stop treating institutional memory as an archive and start treating it as something that grows: a federated wiki where the pages write themselves, if you’re willing to talk to it.
The problem
A non-technical principal, running an organisation with more happening than any one person can track, needed a way to capture what the organisation was learning without becoming its own full-time documentation job. Structured tools ask people to translate their thinking into fields and folders before it counts as knowledge. Most of what matters gets lost in that translation, or never survives it.
What I built
As lead developer and design technologist, I built xCO as LLM-based organisational learning infrastructure with agent orchestration underneath a federated wiki structure — using natural language as the interface itself, rather than a layer bolted on top of one. Authoring and retrieval both happen conversationally: an agent orchestration layer routes what’s said into the right part of the federation, links it to what’s already there, and surfaces it again when it’s needed, without asking the principal to think in schemas.
Where it stands
xCO is live and in daily use by multiple people inside the organisation — infrastructure that keeps running because using it and maintaining it are, finally, the same act.