Pathlight·Playbook

The Meeting Cortex

Win 4: every call becomes a filed, searchable note without you typing a word. A recorder captures the conversation. The /meeting-cortex skill reads it, works out what kind of call it was, and files the note plus the useful bits where each belongs. Wire it once, and your notes write themselves.

This is the habit that pays for the whole guide. In the Codex chapter (Win 3) you started a meetings/ folder and filled it by hand. This chapter makes it fill itself.

Why this is the win to install

Codifying your business by hand has one honest problem: nobody types up every call. You finish a supplier call, mean to note the three things that mattered, then move on. The transcript sits in a tool you never open again, and by next week the details are gone.

The Meeting Cortex closes that gap. Wire it once, and every recorded call becomes a filed note within minutes. Your Memory (the Codex from Win 2 and Win 3) stops being a thing you maintain and becomes a thing that fills itself. That is why it feels less like a feature and more like getting an assistant.

One rule drives it: no useful thing said in a meeting should die in a transcript. Every point worth keeping lands in the one place you will actually act on it.

How it works: capture, classify, route

One call goes in. A note and its follow-ups come out, filed where they belong.

Your call
   │  a recorder captures and transcribes it
   ▼
/meeting-cortex  reads the transcript
   │  works out what kind of call it was
   ▼
A filed note  +  action items in your task inbox  +  ideas and facts in the right file

Three moves:

  1. Capture. A recorder (Fathom or Granola) sits on your calls and writes the transcript automatically. You do nothing.
  2. Classify. The skill reads the transcript and decides what kind of call it was: a supplier call, a client call, a planning session. That decides which folder the note lands in.
  3. Route. The skill pulls the durable signals out of the note and sends each to its home. Action items go to your task inbox. A content idea goes to your notes. A fact worth keeping goes to your knowledge file.

The note itself always lands in your meetings folder with clean labels on top (date, who was there, links). The transcript stays there too. Only the short, sourced signals travel onward, and every one links back to the note it came from, so you can always check the source.

It proposes. It never overwrites.

Handing a skill the keys to your files sounds risky. It is safe because of one design choice: the Cortex suggests, it never silently changes anything that matters.

  • Anything that would change a core file (a decision, a positioning line, an SOP) shows up as a suggestion you approve, through the same confirm step you use everywhere else (/vault-ingest). Nothing important changes until you tick it.
  • Action items and ideas land in an inbox, never straight into a prioritized list. Sorting them is a separate, deliberate step you control.
  • A fact worth keeping gets added to your knowledge file, flagged for a look, never painting over what is already there.

The Cortex is a distributor, not an author. It moves signals to the parts of your system that already have an approve step, and lets each one do its job. The worst it can do is suggest, and a suggestion is free to reject.

Re-running is safe (what “idempotent” means)

Idempotent is a fancy word for a simple promise: run it twice and you get the same result, not double. Before the Cortex writes anything, it checks whether this call was already filed. If the note exists, it updates or skips it. It never makes a second copy. So you can re-run it on the same call without cluttering your folders, and you never have to remember whether you already processed a meeting.

No recorder yet? Start by hand.

You do not need a recorder to start. Write a five-line summary of a call yourself, drop it in your meetings folder, and run /meeting-cortex on that. The routing still works: the note gets filed, the action items reach your inbox. The pattern is the value. The hands-free capture is what you add once you wire up Fathom or Granola.

Maya’s version

Maya runs an online store and a small service business, plus a content channel. Here is a normal week once the Cortex is on.

A supplier call. She talks through a restock and a price change. She never touches the transcript. Minutes later there is a filed note in her meetings folder and two action items in her task inbox: confirm the new unit price, reorder the bestseller. The price change shows up as a suggestion for her decisions file, waiting for her to approve it.

A client call. Her service client walks through what they need next. The note files itself under client calls. The three things she promised to deliver land in her task inbox as action items, each linked back to the moment in the call where she said them.

A content idea from a chat. Mid-conversation, someone asks a question that would make a great video. The Cortex catches it and drops it into her notes as an idea, waiting for her the next time she plans content, instead of being forgotten by the end of the call.

Notice what Maya never did. She never typed a note, never sorted anything into folders, never copied an action item into her task list. She had the conversations she was already having, and the filing happened on its own.

You’ve got it when

The first three prove it works. The last three prove it is safe and effortless.

The skill that helps here

One command does all of this: /meeting-cortex, which turns a recorded call into a filed note plus routed follow-ups. It is in the starter kit, and the full list of commands with what each does is in 01-skills-and-commands.md. (/granola-sync is the alternate capture path if you record with Granola.)

→ Next: Win 5 - the command center (../05-command-center.md). The single dashboard where all this filed knowledge, your tasks, and your meetings finally show up in one view.