Answers
I keep a second brain for myself. What is the second brain for the AIs my team uses?
It is a shared record the AIs read before they act, holding what your team has actually decided. A personal second brain is written to be re-read by its author: you wrote the note, so on opening it you silently supply everything it left out — that this one was a maybe, that it stopped being true in March, that the tone rule is for client work and not for internal drafts. An AI has nobody sitting beside it to supply any of that. So the second brain for your AIs has to say out loud three things a personal one never needs to: what is a decision and what is still somebody's proposal, when each thing stops being true, and who put it there. That is what makes it safe for a tool to read the record and act on it without checking with you first.
Last updated September 7, 2026
Your own notes work because you are the one reading them
Every personal system leans on the author being present at read time. A line that says "go with the shorter onboarding" is complete for you, because you remember it was the decision from the meeting in June, that it was about the trial flow and not the paid one, and that it is still what you would do today. None of that is written down, and none of it needs to be — until the reader is something that cannot ask. Point an assistant at those notes and it will read that line with the same confidence as a rule that was ratified, and it will apply it to the paid flow, and it will keep applying it a year after you changed your mind.
What the AIs' second brain has to say out loud
Three fields carry the judgement you would otherwise supply in your head. Standing means a person has ratified this, so an assistant may state it as the team's position; anything the AI itself proposed stays visibly a proposal until somebody sanctions it. Validity means each entry carries what would end it — a date, a cycle, a condition — so a rule that expired stops being served instead of aging in place. Provenance means the record says who put it there and when, so the reader can weigh a partner's call differently from a note somebody dropped in during a busy week. Those three are the whole difference between a pile of context and something that can be obeyed.
It gets read at the start of every action, not when somebody remembers
A personal second brain is a place you go to. This one is the opposite: the assistant reads it before doing the work, on every run, without being reminded — that is the only arrangement in which nobody has to remember. And the writing side matches it, because a record that only ever gets read goes stale: the assistants write what they established as they finish, so the log of what has been happening builds itself. What stays with the people is the veto — deciding what becomes the team's position, and what quietly gets retired.
What it looks like in practice
A founder keeps a personal file of product calls and pastes the relevant part into whichever assistant they are using that day. It works, in the sense that the pasted part is right. What it costs shows up elsewhere: the designer's assistant has never seen any of it, so it proposes a flow that was ruled out in June; and two weeks after the pricing call changes, the founder is still pasting the old paragraph, because the file was updated and the habit was not. Moved into a shared record, the same decisions are read by the designer's assistant, the founder's, and the one running the weekly report — each one before it acts. The pricing change is written once and every assistant is on the new version the next time it opens the project. The file on the founder's laptop does not have to go anywhere. It just stops being the only copy, and stops being carried by hand.
Questions people ask about this
- Can't I just point the AI at the notes I already have?
- You can, and for one person and a small pile it is a reasonable start. What it does not survive is scale and time: your notes have no way to say which lines were ratified, which have expired, and which were a thought you had on a Tuesday — so the assistant treats all of it as equally true and cites the expired ones with a straight face. Sorting that out by rewriting the notes is the work either way; doing it in a record built to hold standing and validity is the same work with the answer kept.
- Who writes it, then — me or the AI?
- The assistants write most of it, because they are the ones in the work when something gets decided, and people sanction. That is the split: an AI can propose a rule and record what it found, but only a person turns a proposal into the team's standing position, and only a person retires one. Nothing gets promoted by volume, by confidence or by repetition.
- Does this replace my own second brain?
- No, and it is not trying to. Yours holds what you know and is optimised for you re-reading it; this one holds what the team decided and is optimised for something that acts on it unattended. Most people who set it up keep both, and the honest test of whether it is working is whether they have stopped copying anything from one to the other.
Where this is verifiable
Product documentation on this site (How it works, Install) and the sanctioned product spec in the repository, for how standing, validity and provenance are stored and served. The failure described here — a note read as a ratified rule by an assistant that cannot ask — comes from our own use of the tool and from the intent people state in public when they describe pasting a personal list into every AI they open. Everything described here is behaviour the tools apply today, not roadmap.
https://www.arroway.app/en/answers/the-second-brain-for-your-ais