Exocortex R&D: a lab where evidence decides about AI
A public lab in which a decision to deploy AI rests on a result measured against a criterion set in advance, instead of an impressive demo.
A system that pre-synthesises threads from different sources around whatever I am working on: a client, a project, an industry or a field of knowledge.
Exocortex is the foundation of the way I work. Several dozen documents a day in total: a dozen or more meeting transcripts, emails saved to read later, mostly about AI, and public articles. The system connects them into threads before I start thinking about them myself.
I have ADHD, and Exocortex is the foundation of how I function. For neurodivergent people a system like this can be very strong support, and that is one of the reasons I talk about it openly.
Exocortex runs on my own hardware, with locally run models. You can ask it a question and get an answer with links to the sources. Every night an automated test checks the whole path, from loading notes to the answer. The first such test found a real bug, which has been fixed. Known gaps are written down openly.
The lab is a complete, separate project that uses only the Exocortex engine. It runs in public and tests hypotheses on publicly available knowledge, so it is completely separated from the data of clients, employers and my own private data. One engine, two installations with opposite rules: a private one for daily work, a public one for experiments.
Questions about: Exocortex — a knowledge system