Your AI should already know this.
Brain Co. keeps the settled facts about your work, updates itself from where you work, and hands them to Claude, ChatGPT and your agents.
Some background first. I run partnerships at a software company. Acme is one of our three biggest clients. Last quarter we promised them faster onboarding and it slipped. Their CFO makes the real decision, even though we talk to someone else every week. We already tried a discount and it didn’t land.
Help me write the email for Thursday’s renewal call.
Geniuses with goldfish memories
Claude and ChatGPT can write, analyse and explain almost anything. They just don’t know your situation. So you explain it every time, or you get an answer written for anybody.
Is this $499 claim under our expense limit?
Is this $499 claim under our expense limit?
Claude remembers you. So does ChatGPT. Separately.
Built-in memory keeps getting better, but each app keeps its own, often split again by project, and none of them sat in on your meetings. A brain sits outside all of them. It hears what matters, keeps it current, and briefs whichever AI you open next.
Wherever you ask
Each one reads from your brain and can send it what it learns.
- The Claude appreads · adds
- Coworkreads · adds
- Claude Codereads · adds
- ChatGPTreads · adds
- Notion agentsreads · adds
Where you work
It keeps itself current from your meeting notes, and soon Slack and email, on a schedule you choose.
- Granola
- Notion meeting notes
- Slacksoon
- Emailsoon
Connected isn’t the same as knowing
Connect Claude to Slack, Drive and Notion and it can search everything you’ve ever written. But plans change, decisions get made on calls, and last month’s document is still there saying the opposite. Your brain holds only what’s settled, and keeps it current.
How should I handle the Acme renewal?
How should I handle the Acme renewal?
Your brain brings the facts. Your AI brings the thinking.
Ask the way you always do. Your brain hands over the notes that matter, and the AI mixes them with everything it already knows. Pick one.
Write the email for Thursday’s renewal call.
- Maya, the CFO, signs off. Tom doesn’t.
- Onboarding slipped. Now due 15 December.
- A discount didn’t land. A weekly written update did.
Here’s a draft to Maya. It owns the delay, commits to 15 December and offers a short written update every Friday. It leaves out a discount, since she said last time it wasn’t about price.
Kept true, with your OK.
Plans change. When something new comes in from a meeting or a chat, your brain proposes the update, shows where it came from, and flags anything that contradicts what it already holds. Talking something through isn’t deciding it, so only what’s settled goes in. Approve it and every AI and agent gets the new version. Try throwing one out.
- NewNew
Maya will join Thursday’s renewal call. She wants a firm onboarding date before she signs.
from a Granola meeting “Acme weekly sync” · Tue 2:00 PM/meetings/2026-11-10-sync.md - ChangedChanged
Acme’s audit moved a month, so priority support now runs March to April. You mentioned it while planning the support rota.
from a Claude conversation · Wed 9:40 AM/promises/2026.md - ConflictConflict
Lina is now the weekly contact at Acme. The brain still lists Tom.
The brain already says something different here. Worth a look before you keep it.
from a Notion agent · Thu 11:00 AM/people/who-decides.md
3 of 3 updates will be applied.
Repeated notes are merged, so a busy week still reads in a minute. Throw out what’s wrong and the rest still goes in.
Plain notes. Read every line.
A brain is written and tidied for you, a folder per topic. Open any file and see exactly what your AI will be told.
# Who decides At Acme - Maya Ortiz, CFO. Signs off the renewal. Cares about cost and a clean audit. Joined the account in October. - Tom Reyes, ops lead. Our weekly contact. Friendly, but holds no budget. - Ravi Menon, head of IT. Can block anything that touches security. Ask him early. On our side - You own the relationship. - Sam runs their support. Since the October review, take anything commercial straight to Maya.
One memory for the whole team
Built-in AI memory belongs to one person. A brain can belong to a team. Hand over a client and the context stays. Someone joins and asks the brain instead of you.
- Private
- Where every brain starts. Nobody you work with sees it until you share it.
- Read
- It answers their questions in their own AI.
- Write
- They can add to it too. Changes still arrive as a review.
- Agent key
- Give an agent its own key, limited to the brains you pick.
What people keep brains about
Anything you’ll talk to an AI about more than twice. Several small brains work better than one giant one.
- A clientpeople/ · promises/ · history/
- A productvision/ · customers/ · competitors/
- A teamwho-owns-what/ · rituals/ · requests/
- Someone you managegoals/ · feedback/ · open-asks/Only you
- A renovationquotes/ · decisions/ · builders/
- Your investmentsholdings/ · theses/ · watchlist/
Start in about fifteen minutes
Connect your AI
Paste one link into the Claude app, Cowork, Claude Code or ChatGPT, and sign in.
ChatGPT needs developer mode turned on.Make your first brain
Answer a few questions, or point it at a folder or a Claude project you already have.
Carry on as normal
Ask the way you always do. It reads your brain when that helps.