# Unit 0: Start Here > From 102 – Operator on ThinkModel. Free to read; the interactive version of this unit is at https://thinkmodel.ai/course/102/unit/0 *Last updated: August 9, 2026. Accounts, tiers, and requirements in this unit were verified on this date.* ## What this course is ThinkModel 102: Operator teaches you to operate AI systems, not just use them. Using AI is typing a question and reading an answer. Operating AI is configuring workspaces, choosing models deliberately, delegating work to agents with the right guardrails, wiring automations that run without you, and testing all of it before you trust it. The whole course climbs one ladder, from writing a single prompt to running governed systems. Twelve units, one rung each. Every unit re-explains what it depends on, so you can read all of 102 right now without having taken ThinkModel 101 — nothing here is locked away. One thing to know before you start: tracking your progress through 102 does require 101. Marking units complete, saving quiz answers, and the Prove It exercises all unlock once 101 is finished. So read ahead freely; if you want the record to follow you, take the foundations first. If you want them anyway (how AI works, why it fails, how to verify it), 101 is the place. ## What you need **Accounts, all free to create:** - One general assistant: Claude ([claude.ai](https://claude.ai)), ChatGPT ([chatgpt.com](https://chatgpt.com)), or Gemini ([gemini.google.com](https://gemini.google.com)). Ideally two, so you can compare. - NotebookLM ([notebooklm.google.com](https://notebooklm.google.com)), Google's grounded workspace, renamed Gemini Notebook in 2026. Any Google account works. - Later units add one or two free tools each (a terminal agent in Unit 04, an automation tool in Unit 08, an image tool in Unit 09). Each unit tells you exactly what and marks the free route. **Hardware:** a phone is enough for most of the course. A laptop or desktop makes Units 04 (terminal agents) and 08 (automations) smoother. Unit 10 (running AI models on your own machine) works best with a computer that has 8GB of RAM or more; if you don't have one, the unit includes a browser-based path so you can still complete it. **Time:** each unit takes roughly 60 to 90 minutes: reading, a hands-on demo, and a challenge that produces something you keep. The suggested pace is 12 units over 8 weeks. The final unit includes a two-week capstone where you operate a real system. ## The free path Every unit in this course can be completed on zero budget, and every Live Demo marks its free path explicitly. Paid plans (typically around $20/month for an assistant's Pro tier) raise usage limits and unlock some smoother routes, and the course will tell you when that's true. It will never be required. If you hit a usage cap on a free tier, waiting a few hours is always an acceptable strategy; the course is designed around it. ## The Studio One worked example runs through the whole course: the Studio, an automated social media system for a fictional neighborhood coffee shop called Marra Coffee. In each unit's Live Demo you'll watch one more piece of the Studio get built, until by Unit 08 it's a running pipeline: researching, drafting, and waiting at a human gate before anything posts. Two rules about the Studio. First, it's a demonstration, and your challenges never depend on it: every unit's challenge uses your own life, your own domains, your own tasks, so everything you build belongs to you. Second, the Studio never posts anything without a human approving it, and neither should anything you build. That gate is not a training wheel you remove later. It's the design. ## The operator's one ethic Delegate the work, never the learning. AI can draft, summarize, research, and build faster than you. Use that. But every artifact you produce in this course, and in your life, should be one you can defend without the AI in the room: explain why it's right, where it's weak, and what you'd change. If you hand in work you can't defend, you didn't delegate the work; you delegated the learning, and that trade always loses. If a challenge here overlaps with schoolwork or a job, follow your school's or employer's disclosure rules, and hold yourself to the defend-it test either way. ## How each unit works Every unit has the same anatomy: a hook, a short video, a core concept built around one mental model you'll keep for years, a hands-on Live Demo (Studio first, then you), three Operator Moves you can use the same day, two quizzes, and a challenge with success criteria. A "Last updated" stamp sits at the foot of every unit, because tools and prices in this field expire fast. The mental models don't; they're the actual product. One more thing carries across units: your toolkit. Unit 01's workspaces get used by Unit 03's research relay. Unit 02's benchmark gets rerun in Unit 10. Unit 07's permission policy governs Unit 08's automation. By the end you don't have twelve completed lessons; you have one operating system for working with AI, built out of pieces you made. Start with Unit 01. Your assistant already remembers you. The question is who decided what.