personal-os
$ _

How do you build an AI that runs your life?

People are solving this problem in wildly different ways. Some use Claude Code with MCP servers. Others chain GPT with Zapier. A few write custom agents from scratch.

self.md documents what actually works. We test approaches, compare tools, and publish the setups that survive real use.

No single “right” answer. Just working systems from people who built them.

Case Studies

How real people build their AI workflows
Boris Cherny
Boris Cherny Claude Code creator, parallel agents
Andrej Karpathy
Andrej Karpathy Vibe coding inventor
Sahil Lavingia
Sahil Lavingia Gumroad CEO, 41% AI code commits
Lance Martin
Lance Martin Self-learning memory system
Pieter Levels
Pieter Levels AI indie hacking, $3M+/year solo

All 144 case studies →

Where to start

Depends on your situation:

If you’re a developer — Start with Claude Code setup. Terminal-native, MCP servers for integrations, runs locally.

If you want something quick — Quick start guide gets you a basic working setup fast.

If you’re evaluating options — Read the tool comparison first. Different tools suit different workflows.

If you tried before and it didn’t stick — Check what to delegate. The problem is usually scope, not tools.

Concepts

What is a personal OS? The idea, not tied to any specific tool Architecture patterns Five layers that appear in most working systems What to delegate Where AI helps vs. where it gets in the way Tool comparison Claude Code vs. GPT vs. Gemini vs. custom

All 48 concepts →

Guides

Claude Code setup MCP servers, custom instructions, terminal workflow Memory systems Different ways to give AI persistent context MCP servers Calendar, tasks, web — what's available Quick start Fastest path to a working setup

All 38 guides →