Turn Your AI Note Taker Into a Second Brain (The Meeting Wiki System)

If you run cohorts or coaching, your best teaching disappears after every call. The sessions. The 1-on-1s. The stories and analogies you use to explain the hard stuff. Gone the moment the call ends.

Most people now have an AI note taker. Fathom, or one like it, records the call and gives you a summary. Useful. But a summary is a notes folder, not a brain. It sits there. Nobody reads it. It feeds nothing.

The fix is a second brain. Here is the difference, and how to build one.

---

A note taker is not a brain

Think about it in three tiers.

A raw recording is a security camera in an empty room. It films everything, and nobody watches the tape.

A note taker's summary is one guard watching one monitor. He sees that single call, writes it up at the end of his shift, and that is it. He never saw the other cameras, so he cannot connect a thing.

A second brain is the whole control room. Every call linked together, every person, every idea, wired into one picture. You do not get twenty separate reports. You get the pattern no single guard could ever spot.

That is how it tells you the client who went quiet in March is the one at risk in June. One feed cannot see that. The linked room can.

The test: open your note taker right now. When did you last read a summary back? For most people, never. That is tier two, sitting dead in a folder.

---

Stop chasing the model. Build the context.

Here is the mindset shift. Everyone is chasing better prompts and the newest model. Both are the wrong thing to chase.

Learning to prompt is a fading skill. The way you talk to one model changes with the next release. And the models keep getting better whether you do anything or not. That is not your job to fix.

Here is the part people miss. Feeding a model more of your work does not give it a bigger brain. It gives it a fatter notebook. The brain is rented and swapped for a better one every few months. The notebook is yours, and it is the only part that compounds. So the counter-intuitive move is to build the memory now, while the models are still improving. Every better model that ships just reads the context you have already captured and pulls more out of it.

Own the method, not the model.

---

The three parts that make it work

A second brain is three parts, and they map to a body.

Wikis are the brain. Not one wiki, a few. A meeting wiki of every call, mapped. A stories wiki where your metaphors and the way you explain things live. Plain markdown, linked together. The stories wiki is the secret weapon: it is what later makes content sound like you instead of generic.

Skills are the muscle. A skill is a packaged method that reads the right wiki, then acts. One skill pulls a call apart and files it. Another mines the teaching out of a transcript. Another drafts in your voice by reading your stories first.

Routines are the heartbeat. A routine fires the skills on a clock, so the whole thing runs without you. You run a call, Fathom records it, and a routine synthesises it into the wiki, no prompt from you. Miss the heartbeat and you have skills you keep forgetting to run.

This is built on Andrej Karpathy's LLM-wiki pattern: the synthesis happens when a call goes in, not when you search. So every later read is instant. The note taker that records the call, Fathom or one like it, is just the front door. The wiki is what turns the recording into a brain.

---

Content is just one use case

Content is the obvious win. A skill scans freshly captured teaching and drops post ideas onto a stack. Your best on-call explanation becomes a post before you have left the meeting. Then the content team reads your stories first and drafts in your voice.

But content is one use case, not the point. The wiki pools every call, so it sees patterns no single call shows.

You can pool a set of client calls and ask what they were all asking, and it hands you your ICP. You can ask "what is my metaphor for this" and get your own words back. You can spot the client going quiet before they churn. One call cannot show you any of that. The linked brain can.

The reason the content sounds like you is not a clever model. It is that the model reads your real stories before it writes a word.

---

What you actually do now

You do not build this from scratch. Go to the guide. It has a mega prompt. You run that one prompt, point it at your Fathom calls, and it stands up the wiki, the skills and the routine for you.

Start with your back catalogue. Feed it the calls you have already run, so the brain starts full, not empty. Then the routine takes every new call from there.

The full mechanism, with the Fathom-edition mega prompt, is in The Meeting Wiki Brain guide. The wiki itself follows the same pattern you would use to build your own AI stack, and it is the memory layer of a full AI Operating System. Capturing your teaching is also the first step to turning it into a product, which DevWiz covers in turning your program into a software platform.

Without it, every call is a separate guard writing a report nobody reads. With it, one control room, every feed linked, the whole picture. That is the jump from a note taker to a brain.

Want to know which part to build first for your business? Run the assessment. Five minutes, scored, with a clear first move.