From the Lab

Second Brain

never lose context

cursorclaude-codechatgpt
$ cat problem.md

The Problem

Context switching between AI tools was killing productivity. Every time I moved from Cursor to Claude Code to ChatGPT, I had to re-explain:

  • What I was working on
  • The codebase structure and conventions
  • Previous decisions and why they were made
  • The current state of the project

It felt like having amnesia every time I opened a new tool. Each AI assistant started from zero, wasting time on re-explaining instead of actual work.

$ cat solution.md

The Solution

Built a structured project setup that maintains context across all AI tools I use:

  • Cursor: Configured with project-specific rules and context files
  • Claude Code: Access to same codebase with skills and instructions
  • Claude Desktop: Project mode with shared knowledge base
  • ChatGPT: Custom instructions that reference the same project structure

The key is a shared foundation: CLAUDE.md files, .cursorrules, and .clinerules that every tool can read. Context persists no matter which tool I'm in.

$ cat stack.md

Technical Stack

The second brain isn't a single tool - it's an orchestrated system:

  • Cursor: Primary coding environment with AI pair programming
  • Claude Code (CLI): Terminal-based assistant for quick tasks
  • Claude Desktop: Project mode for high-level planning and design
  • ChatGPT: Supplementary for specific tasks
  • Shared context files: CLAUDE.md, .cursorrules, .clinerules in each project

Each tool reads from the same source of truth. When I update project conventions in one place, all tools benefit. No more maintaining separate instructions for each AI.

$ cat impact.md

Impact

Never Lose Context

The transformation is dramatic:

  • Switch between tools without re-explaining project context
  • Each AI knows the codebase structure, conventions, and current goals
  • Consistent responses across tools because they share the same foundation
  • New projects spin up faster - just copy the context structure

It's like having one continuous conversation across multiple tools instead of starting over every time. The cognitive load reduction is massive.

THE RECEIPTS

How Projects Are Organized

my-project/
├── .cursorrules          # Cursor-specific context
├── .clinerules           # Claude Code context
├── CLAUDE.md             # Shared project context
├── docs/
│   ├── architecture.md   # System design
│   └── decisions.md      # ADRs and rationale
├── src/
│   └── [code...]
└── README.md             # Project overview

Every AI tool reads these context files first, ensuring they understand the project before making suggestions.

How Context Persists

                Source of Truth
                      │
           ┌──────────┼──────────┐
           │          │          │
    CLAUDE.md   .cursorrules  .clinerules
           │          │          │
    ┌──────┴────┬────┴────┬────┴─────┐
    │           │         │          │
  Cursor    Claude     Claude    ChatGPT
           Code     Desktop

All tools read from same files
→ Consistent context everywhere
→ No re-explaining needed

Update context once, benefit everywhere. This is the key to never losing context.

Sample Context File

# Project Context

## What This Project Does
[Clear description of purpose and goals]

## Architecture
- Tech stack: Next.js, TypeScript, Tailwind
- Key conventions: [specific patterns we follow]
- File structure: [how code is organized]

## Current Work
- Active tasks: [what we're building now]
- Known issues: [things to be aware of]
- Decisions: [important architectural choices]

## How to Help
- Code style: [formatting preferences]
- Testing: [test requirements]
- Documentation: [when to add docs]

Customize this template for each project. All AI tools read it before responding.

← Back to the Lab