00 — What Cortex is, and what becomes possible
Before you install anything, you need a clear picture of the system you’re building. This chapter is the picture. Read it once. Come back to it every time a chapter feels disconnected from the whole.
What Cortex is
Cortex is an AI Operating System for a business.
Not an AI tool. Not an assistant. A system that runs underneath your business the same way macOS runs underneath the apps on your laptop. You don’t think about the OS most of the time. You think about the work. The OS makes the work possible.
A Cortex has four parts. They map to how a brain works, and how a golfer plays.
| Part | Brain | Golf | What it holds |
|---|---|---|---|
| Memory | Temporal cortex | Yardage book | What the business knows: voice, ICP, SOPs, decisions, prior work |
| Senses | Sensory + motor cortex | Clubs and scorecard | How it perceives and acts: inboxes, calendars, CRMs, channels, surfaces |
| Habits | Basal ganglia | The repeated swing | What runs without thinking: skills, workflows, agents, automations |
| Learning | Neuroplasticity | The practice round | How it gets better: rituals, lints, audits, drift detection |
A system without all four does not compound. Memory without Habits is a wiki. Habits without Memory is RPA. Senses without Learning drifts. Learning without Memory has nothing to update.
All four. Always.
The three layers
A Cortex installs in layers. Each layer is the same four parts, applied to a wider surface.
Layer 1 — Owner. The founder’s own Cortex. Your Memory, your Senses, your Habits, your Learning. The first node. Until this works for you personally, nothing scales.
Layer 2 — Team. The Cortex extends to the people you work with. The same Memory the founder reads from is the Memory the team writes to. The same Habits the founder ran by hand become workflows the team triggers.
Layer 3 — OS. The Cortex becomes the substrate the whole business runs on. New hires onboard into it. Clients touch it. Decisions get logged into it. The business stops being held together by the founder’s head.
The Playbook installs Layer 1. Layer 2 and Layer 3 come later, through a Cortex Diagnostic or a Full Cortex engagement. The architecture is the same. Only the scope changes.
What becomes possible
A working Cortex changes what a single operator can do. Concrete examples — these are not aspirational. They are install outcomes from real Cortex deployments.
Memory makes you faster than you can read
- A prospect emails. The system already knows who they are, what they care about, what you’ve discussed, what was decided last time, and what you said you’d send. You open the thread and the context is loaded before you read it.
- You ask “what did we decide about X” and get the decision plus the conversation it came from plus the document where it was logged. No archaeology.
- A new hire reads the Codex on their first morning and is operational by lunch. The knowledge that used to live in your head now lives in files they can search.
Senses make the system reach further than you can
- Meetings become structured knowledge automatically. Granola transcribes. The system classifies, extracts decisions, routes follow-ups, and updates the right files. You walk out of a call and the artifacts already exist.
- Inbox triage happens before you open the inbox. The system reads, drafts replies in your voice, flags what needs you, archives what doesn’t.
- A prospect signal lands on LinkedIn and the system surfaces it to you with context: who they are, why they matter, what you already know about their firm.
Habits make the work run without you
- Outreach campaigns run on a schedule. Discover prospects, score them, personalize the message from real case studies in Memory, queue for approval. You review and send. The system did the other 95%.
- Content gets drafted on demand. You log an idea. The system produces a script in your voice, with your structure, ready to film. You edit, not generate.
- Workflows that used to take a Monday morning now take a button press. The work still happens. You’re not doing it anymore.
Learning makes the system better every month
- Weekly rituals surface what drifted, what broke, what got better. Drift caught early is cheap. Drift caught late is a rebuild.
- Lints and audits run continuously. Stale files get flagged. Broken links get reported. Contradictions get surfaced. The system audits itself.
- Decisions get logged with reasons, not just outcomes. Six months later, the answer to “why did we do it this way” lives in a file, not your memory.
What stays human
Cortex does not replace the founder. It removes the friction between the founder’s intent and the business executing on it.
You still:
- Decide what the business is for
- Decide who the business serves
- Decide what the business will and will not do
- Hold the relationships that matter
- Set taste
Cortex handles the rest. The translation from your intent to the work getting done. The memory that what was decided stays decided. The execution that doesn’t depend on you being awake.
If you are doing work a Cortex could do, you are leaving the strategic work undone.
What “working” looks like
You know your Cortex is real when:
- You can leave for a week and the business runs. Not perfectly. But the workflows ship, the inbox stays managed, the system flags what needs you when you return.
- A new hire onboards in days, not weeks. Everything they need to know is in Memory, written down, searchable, current.
- Decisions stop getting re-litigated. When the same question comes up again, you point at a file. The file has the answer and the reasoning.
- Output goes up while hours go down. The marginal cost of one more campaign, one more piece of content, one more prospect approaches zero.
- The system catches its own drift. You don’t audit the Cortex by hand. The Learning layer does.
If three of these are true, you have a Cortex. If all five are true, you have an Operating System.
The install path
The rest of this Playbook is the install. Read in this order:
00-prerequisites.md— the toolchain you need before chapter 1 (Claude Code, an MD vault,~/hubstructure)- Memory (
01-memory/) — install the Codex, the wiki, the strategy vault. This is the longest section. Memory is the foundation everything else depends on. - Senses (
02-senses/) — wire Claude Code as your surface, connect Gmail, Slack, Notion, n8n - Habits (
03-habits/) — install the skills, the workflows, the meeting Cortex - Learning (
04-learning/) — the rituals that make the system get better instead of decay
Order matters. Memory before Senses, Senses before Habits, Habits before Learning. If you try to install Habits without Memory, you build a brittle automation. If you try to install Learning without Habits, there is nothing to learn from.
A note on scope
This is the personal install. One operator. One Cortex.
If you are running a 5 to 50 person firm and want a Cortex that
scales to the team and the OS layer, the personal install is still the
right place to start. The founder installs first. Then 99-where-this-goes-next.md
shows you the bridge to the Cortex Diagnostic, where the same
architecture extends to the rest of the business.
You cannot install a team Cortex on top of a founder who has not installed their own. The Owner layer is the foundation.
Definition of done for this chapter
You can move to the next chapter when:
- You can name the four parts without looking
- You can describe what each part holds in one sentence
- You can describe one concrete thing that becomes possible at the Owner layer
- You understand that Memory comes first and Learning comes last, and why
If any of that is fuzzy, re-read this chapter. The rest of the Playbook assumes it.
→ Next: 00-prerequisites.md