# AI Second Brain Cheat Sheet

A second brain is 100% about your knowledge and 0% about the AI. It is your existing knowledge (notes, emails, transcripts, saved videos, Google Drive files, CRM data) connected by topic through one index file. You can plug any AI into it.

## Gather your sources

- Notes (for example Apple Notes)
- Emails
- Transcripts (sales calls, meetings)
- Things you saved on YouTube
- Google Drive documents
- CRM data

## The LLM Wiki process (Andrej Karpathy's approach)

1. Have the AI (for example Claude Code or Codex) read each document.
2. It comes up with a list of topics.
3. It tags each document with those topics.
4. It links documents that share a topic.
5. It writes everything into one file, the index: each topic and the links between documents.
6. When a new document comes in, the AI reads it, tags it and creates the new links.

## Use it

Tell your AI once:

```
Before you answer anything to me, I want you to go and check the index of my second brain to find the necessary knowledge that we have.
```

Then ask questions about your business, for example:

```
Can you tell me who this person is?
```

```
Can you tell me what's the status of the pipeline?
```

```
Can you tell me how my content is performing?
```

## The three levels

| Level | How documents get linked | Limitation |
|---|---|---|
| 1. Keyword links (LLM Wiki) | The same word appears in several documents ("revenue" and "revenue") | Misses related terms ("ARR") |
| 2. Embeddings | Words with similar meanings sit close together, so "ARR" and a misspelled "revenue" still connect to revenue | Experimental: in my tests it adds connections but is not there yet |
| 3. Document sectioning | Another AI splits each document into sections by topic (project A, theme B, theme C in one meeting), and each section is linked separately | More processing per document |

Start with level 1. It is enough to build your second brain and see its shape.