What if your stack already ran itself by the time you opened your laptop? It's 8:47 on a Tuesday. Maya has 64 unread emails, a 9 AM she hasn't prepped for, a Slack DM asking about subscription spend she can't answer without opening five files - and she hasn't had the coffee that counts. ChatGPT lives in a tab. It cannot read her Gmail. It cannot reconcile her bank export against Stripe. It cannot drive her browser at 6 AM while she sleeps. This book is about the version of Claude that can.
The agentic filter - what makes this book different.
Every chapter answers one question before it earns a place: "Could ChatGPT do this in a chat window?" If yes, the chapter teaches the wrong thing. This is not a book of clever prompts you paste into a browser. It is an operator's manual for the version of Claude that lives in a folder on your laptop, reads your files, runs your shell, drives your browser, and works while you sleep. The wedge sentence this book exists to defend:
"Claude Code is the duct tape between your systems. You operate it through prompts; it operates everything else - your files, your shell, your apps, your browser, your desktop, your schedule, your specialists, your dashboards. By Chapter 26, Maya does eight jobs unattended that she used to do by hand. None of them happen in a chat window."
The eight jobs you build by the end:
- Reconcile your money in 10 minutes from your bank + Stripe exports
- Route your Gmail into Notion automatically - drafts replies in your voice, never auto-sends
- Watch competitors and surface pricing changes weekly
- Monitor your deploys and escalate problems before customers notice
- Scrape pages that have no API - cancel subscription leaks, pull 50 competitor pricing pages
- File your monthly P&L from CSVs and APIs
- Refresh a live dashboard your co-founder reads every morning
- Run a small staff of named specialists - Iris, Atlas, Reuben, Echo - that operate while you sleep
What's new in this 2nd Edition: - Part V - The Multi-Agent Studio (3 new chapters): a newsroom, a launch crew, a watchdog mesh
- Part VI - Real Projects You Ship (4 new chapters): a deck, a converting landing page, browser+computer-use automation, a live dashboard
- Part VII - Advanced Claude Code (3 new chapters): the agent SDK, marketplaces, auto mode, /security-review, browser & computer use
- Bonus chapter: a Karpathy-inspired second-brain wiki that compounds for ten years
- Companion repository - extras, not requirements
Who this is for: founders, ops leads, marketing managers, solo consultants, finance professionals, indie hackers, and anyone with technical literacy who wants automation across their stack - but does not want to write code. Who this is NOT for: pure software engineers (see Anthropic's docs and AI Engineering by Chip Huyen instead), mobile-only readers (Claude Code needs a real computer), and anyone who loves the chat-tab workflow and isn't trying to leave it.
The honest costs: Claude Code requires a paid Anthropic plan, roughly $20/month. A persistent computer - a desktop, a Mac mini, or a small rented server - is recommended for the "runs while you sleep" payoff.
The promise. 26 chapters. 9 appendices. Two years of running Claude Code as a daily AI operator, distilled into the operator's manual the field was missing. If you can describe what you want in plain English, you can do this. Turn the page and meet Maya.
