TAOS Brain
Notes, decisions, open work and daily records live in one repository; the AI layer on top loads the context itself at the start of every session and picks the work up where it was left.
In most companies knowledge lives in people's heads, in chat threads and in a folder on someone's desktop. When a person leaves, the knowledge leaves with them; a new hire asks the same questions for three months; the same proposal is written from scratch for the tenth time; and nobody remembers why a decision was made six months ago. TAOS Brain pulls that knowledge into a single repository, points at one source of truth per topic, and puts an AI layer on top of it. We ran this product in our own company before we started selling it, and we still use it every day; the company brain we install for businesses is the productised version of it. The numbers below come from our own daily use.
Problem
The same fact often sits in four different files, worded four different ways, and nothing says which one is current. A person or an AI agent can read the outdated one and decide wrongly. Even when nothing is missing, it stays unclear which information still holds.
What changed
Measured on 16 September 2026, on our own brain. The archive is 287 files and roughly 709,000 tokens: it does not fit into a model's context, and reading all of it for every question would be wasteful even if it did. Answering a question used to mean finding the right file and reading the whole of it: across four real questions the files opened averaged 23,000 tokens, the largest 56,000. Now the lines that match those same questions come back at an average of 730 tokens, with the source file path, in a tenth of a second. Context shrank about 32 times, and the source of each answer became visible. The system has been used every day for 22 days since 26 August 2026, and these numbers come from that use.
How it works
Every topic has exactly one source of truth, and a topic map says which file that is; when another file touches the same topic it links instead of repeating. Open decisions live in their own list, and meeting notes count as frozen records. The AI layer on top loads the calendar, the work that has come due and the latest state at the start of each session, and writes down what happened when the session ends. Changes go into version control, so every entry keeps its date and its reason.
Screens
Highlights
One source of truth
A fact lives in one place and is never copied; other files link to it. When the fact is updated, no stale copy is left behind.
Session memory
The system remembers where the conversation stopped. Open topics, decisions taken and pending work enter the context at the start of every session, so nothing has to be re-explained.
Dated reminders
Work that has come due is listed at the start of every session. A parked decision, an unpaid invoice or an untested procedure does not drop out of sight.
Dated measurements
Anything that changes over time is written down with the date it was measured; answers come with their source and date.
Open to AI agents
The knowledge base is exposed to agents over MCP, and every search result comes back with its source path, so each answer can be audited.
Repeated work becomes a routine
Anything done a second time turns into a defined flow: daily routines, reports and scans run as those flows rather than by hand.
Let's talk
Let's start with a free 30-minute call; just pick a time that suits you from the calendar.