01 — The Senses

Everything the company knows arrives as signal

Interviews, documents, conversations, agents, systems — messy, unstructured, continuous. It all lands first in the sensory buffer: intake/.

🎙️
Interviews
AI Interview Module
📄
Documents
docs · PDFs · wikis
💬
Conversations
Slack · meetings
🤖
Agents
what they observe
🗄️
Systems
apps · databases
COMPANY AI BRAIN
1,842,315NEURONS
GROWING +4,293 THIS WEEK · via cortex/intake/
02 — Myelination · distill

Raw signal is distilled into engrams

An LLM splits each capture into atomic claims — one idea per file — and mints a neuron with full frontmatter. Markdown in, Markdown forever.

raw capture · writing
intake/…/ops-lead.raw.md

            
chunk
atomic claims
refunds ↔ zone 4 delays
reroute play on spike
2026-03-12 outage
mint
engram minted
declarative/…/eng_7f3a2c.md
type: declarative
source: interview
confidence: 0.86
strength: 1
03 — The Neuron

One neuron, three lenses

Every engram is one Markdown file — the source of truth. The same neuron renders as a document for people, and projects into a database record for fast recall.

04 — The Anatomy

Neurons file themselves into lobes

The folder tree is the anatomy; the taxonomy is the map of the mind. Declarative, procedural, episodic, semantic — plus the minds that read them.

cortex/
├── intake/sensory buffer · *.raw.md
│ ├── interviews/AI Interview Module
│ ├── documents/docs · PDFs · wikis
│ ├── conversations/Slack · meetings
│ └── _quarantine/awaiting verify
├── declarative/what the company knows
│ ├── decisions/
│ ├── facts/· 128
│ ├── people/
│ └── glossary/
├── procedural/how the company works
│ ├── playbooks/· 74
│ ├── runbooks/
│ └── standards/
├── episodic/what happened
│ ├── projects/
│ ├── meetings/
│ └── incidents/· 203
├── semantic/the ontology
│ ├── entities/· 56
│ └── relationships/
├── minds/the actors
│ ├── agents/system prompt + tools
│ └── skills/reusable capability
└── .synapse/machinery · generated
├── index.mdembeddings + freshness
├── graph.mdthe edge list
└── decay.mdrecall + pruning
05 — Recall

Recall is the graph firing

A Neural Tile asks a question. Vector search + a graph walk light the path; the neurons that answer grow stronger each time.

Neural Tile · Refunds
“Why are refunds spiking?”
+38%
zone 4 · vs yesterday
idle — watching the stream