# AI Coffee Shop Course Companion Welcome to AI Coffee Shop: a practical course for normal people who want to become builders with AI. This file is your course map. Attach it to your AI assistant at the beginning of the course so the AI can understand what you are learning, connect each lesson to your real work, and help you keep moving without overwhelming you. ## How to use this guide with your AI 1. Attach this Markdown file to your AI conversation. 2. Paste the starter prompt below. 3. Answer the AI's questions honestly, one at a time. 4. Return to the same conversation as you work through each module when possible. 5. Never paste passwords, API keys, credentials, confidential company information, customer data, or other private information into a personal AI account. ## Starter prompt > I am working through the AI Coffee Shop course and I am just getting started. I attached the course content so you can follow along with me. Please act as my learning partner throughout the course. Start by asking me one question at a time about my work, what I already know, what I want to learn or build, and how I prefer to learn. Use my answers and the course guide to help me connect each module to my real goals. Do not overwhelm me—when we make a plan, give me only the first three actions to take. ## The course method The core learning loop is: **Ask → Do → Save → Automate → Ask again** - **Ask:** Give AI honest context and ask questions connected to your real work. - **Do:** Learn hands-on by trying small, concrete actions. - **Save:** Keep useful context, decisions, lessons, and working examples. - **Automate:** Turn repeated processes into reusable skills, tools, or workflows. - **Ask again:** Bring results and errors back to the conversation so the learning compounds. ## Course map ### Part 1 — Foundations #### Module 00 — The Method Learn the ask → do → save → automate → ask again loop. Give AI enough context to create a learning path around your real job, current skills, goals, and preferred learning style. #### Module 01 — What Is AI & How Does It Actually Work? Build a useful mental model for large language models, tokens, context windows, prompts, hallucinations, chatbots, assistants, tools, and agents. Remember that fluent output is not always factual output. #### Module 02 — Set Up Your Toolkit Set up a practical AI workshop with an editor, a programming language, version control, and an AI coding partner. Learn the basics of VS Code, Python, Git, GitHub, the terminal, virtual environments, and safe secret handling. #### Module 03 — Talking to AI Improve results with clear goals, relevant context, real constraints, useful examples, and short feedback loops. Ask for plans before large changes and keep asking what changed, why it changed, and how to verify it. ### Part 2 — Superpowers #### Module 04 — Memory Move durable facts and preferences out of one-off chats so future conversations can start with useful context. Keep memory small, stable, reviewable, and free of sensitive information. Ask your AI to interview you one question at a time, separate durable facts from temporary details, exclude anything sensitive, and show you the proposed memory before saving it. Approve or edit each item, then test the memory in a fresh conversation. #### Module 05 — Skills Turn repeated explanations into reusable workflows. Define the inputs, ordered steps, validation checks, expected output, and conditions that should stop for human review. ### Part 3 — Connect to the Real World #### Module 06 — Git & GitHub Use repositories, branches, commits, diffs, and pull requests to make AI-assisted changes inspectable and reversible. Review the diff and test the result before accepting a change. #### Module 07 — Connecting to Your Data Create a narrow, understandable path between an application and its data. Use safe configuration, least-privilege access, read-only permissions when possible, and enough schema context for the AI to understand what the data means. #### Module 08 — Talking to Your Data Translate a business question into a data plan, query, validation process, and useful explanation. Check joins, filters, row counts, date ranges, units, and known examples before trusting an interpretation. #### Module 09 — APIs, Tokens, OAuth & JWT Understand how applications communicate and how identity and permission move between systems. Learn requests, responses, endpoints, methods, headers, API keys, OAuth, access tokens, refresh tokens, scopes, JWTs, and PKCE. ### Part 4 — Build #### Module 10 — MCP: Giving AI Hands Learn how an AI client discovers and calls governed tools. Prefer narrow business-level tools with clear schemas, enforced permissions, and approval steps around consequential actions. #### Module 11 — Building Bots & Agents Distinguish chatbots, tool-using assistants, deterministic workflows, and goal-directed agents. Use the simplest shape that solves the problem reliably and define clear stop conditions. Use this five-layer AI Coffee Shop mental model to understand the engineering work around modern AI systems: 1. **Prompt engineering — how you ask.** Instructions, goals, constraints, examples, tone, and output shape. 2. **Context engineering — what the model can see right now.** Conversation history, files, retrieved data, memory, tool definitions, and tool results. The prompt is part of the context; context engineering is the larger problem. 3. **Harness engineering — what surrounds the model.** Tools, permissions, sandboxing, files, memory, approvals, tests, tracing, retries, and other runtime scaffolding. 4. **Loop engineering — how it keeps going and how it stops.** Define the trigger, think/act/observe cycle, feedback into the next turn, human checkpoints, completion criteria, timeouts, and hard limits. 5. **Graph engineering — how multiple agents or workflows coordinate.** Decide what specialists exist, what each owns, who can route or delegate to whom, and how state and results move between nodes. A2A is one protocol that can help independent agents discover capabilities and exchange work; it is not the whole graph design. A useful progression to remember is: **Prompt + context = useful conversation → + harness = capable assistant → + loop = agent → + graph = multi-agent system** #### Module 12 — Build & Ship a Web App Bring the pieces together in a real web application. Move from a local idea to a tested product that another person can use. #### Module 13 — Your Path Forward Choose what to build next, keep the learning loop alive, and turn AI from a novelty into a durable capability. ## How your AI should support you Your AI learning partner should: - Explain ideas in plain language before introducing jargon. - Connect examples to your role, tools, and goals. - Give you small actions instead of giant task lists. - Ask clarifying questions when important context is missing. - Explain errors and fixes so you learn while building. - Separate facts from assumptions and tell you what should be verified. - Respect security boundaries and never ask for real credentials or confidential data. - Encourage you to inspect, test, and understand the work rather than blindly accepting output. ## Check-in prompt for each module > I am starting Module [number and title]. Based on this course guide and what you know about my goals, tell me why this module matters for me. Give me one practical example from my world, then help me complete the lesson one step at a time. Ask before moving to the next step. ## End-of-module reflection After each module, answer these questions with your AI: 1. What did I learn in my own words? 2. What can I do now that I could not do before? 3. What still feels unclear? 4. Where could I use this in my real work or life? 5. What is the smallest next action that keeps the momentum going? --- AI Coffee Shop teaches normal people how to become builders with AI—one useful conversation, experiment, and working system at a time.