🧠 Why Claude Code needs a second brain

O Claude Code is an AI agent that lives in your terminal. Unlike a regular chat, it acts: reads project files, edits code, runs commands, and chains steps to complete a task. When we talk about its "memory," we mean everything it can see when it responds.

The agent's memory has two sources: the current conversation (what you said and what it read in this session) and what you points for it to consult (files, rules, indexes). The problem: the conversation disappears when the session ends, and pointing it to “the entire repository” rarely fits. That’s the gap this course solves.

A context window is how much text the model can read at once. It’s measured in tokens — word fragments (about 3 to 4 characters each). It’s large, and newer models can handle up to 1 million tokens, but it’s still finite: an entire repository, with hundreds of files, rarely fits all at once.

↑ Only a handful of files fit in the window; the entire repository overflows. That’s why the agent needs to choose well what to read — and a map helps with that choice.

When the agent doesn’t have a map, its tool is the grep: literal search for a word inside the files. It finds where the term appears — but it doesn't know what it relates with what or why. “Finding” isn’t “understanding.”

↑ Grep returns a pile of locations; the map returns the concept e your neighbors. It's the difference between a list of addresses and a neighborhood map.

"Second brain" is a personal productivity term: a knowledge base external and organized that you consult instead of trying to remember everything. Here, we’ve adapted the idea for the agent: the second brain is the graph + o vault that preserve a project's understanding outside the temporary session.

Claude Code isn’t amnesiac by default. It already reads a file CLAUDE.md at the project root with instructions and preferences, and has its own memory. That’s great for rules ("always run the tests," "use Portuguese")—but it isn't a map of an entire repository.

# Projeto X

- Sempre responda em português.
- Rode os testes antes de dizer "pronto".
- A documentação de referência está em ./docs.

Notice: the CLAUDE.md says “the docs are in ./docs,” but doesn’t say how the concepts in the docs connect. That's where Graphify comes in — it doesn't compete with the built-in tool; it complements: the native graph stores rules; the graph stores the map.

Putting it all together: the Graphify creates the map from a repository or documents; the Obsidian saves this map as navigable markdown; the Claude Code consult the map in your project context. Together, they’re the second brain.