Practical notes on local-first knowledge systems, AI agents, and team intelligence.
Nov 1, 2026
When AI drafts in seconds, the pile moves to the desk of the person who has to sign. That pile is a new cost, and it can be calculated.
Read more: The bottleneck is not writing. It is approval.Nov 1, 2026
Running 2,000 documents a month through AI costs under a dollar in the cloud, about the same on EU hosting, and one €3,000 machine locally. Price does not decide where a document goes; the shelf it belongs on does.
Read more: What may go to the cloud: decide per document type, not per companyOct 31, 2026
Three quarters of small companies that already use AI use it for one isolated task, and the savings you are looking for sit one step further down the process.
Read more: AI that gets stuck on one taskJul 6, 2026
What actually happens in the first week of turning scattered team knowledge into a working, AI-augmented knowledge base.
Read more: A company AI brain in 7 days — a realistic walkthroughJul 6, 2026
Over 10,000 organizations run on Obsidian, yet almost nobody will install and operate it for a team. That gap is the whole opportunity.
Read more: The gap in the knowledge marketJul 6, 2026
How scattered conversations, documents, and decisions become a linked, queryable asset an AI agent can safely work with — in five stages.
Read more: The Knowledge Refinery, explainedJul 6, 2026
The bottleneck between an impressive AI demo and a system your team actually relies on is not the model. It is architecture, governance, and maintenance.
Read more: Pilot purgatory: why most AI agents never reach productionJul 6, 2026
A two-plane architecture that gives teams collective intelligence without giving up data sovereignty or GDPR compliance.
Read more: Public vs. private: how teams use AI without leaking data
