Cakes and bakes for the plans worth keeping.
Ask about a bakeA Fazore build course · Three sessions
Turn an idea into a working product with AI in your terminal.
We give one AI coding agent a folder, teach it how to work with us, and build The Sunday Oven prototype inside it. Every setup click and command is shown. Each build chapter gives you the exact prompt, approval points, and result to check.
In our fictional example, Maya finishes with a customer website and a local, unpublished teaching planner. Its orders, stock figures, and costs are invented so the workflow can be tested without real customer data.
- Fundamentals
- Setup, working rules, and the project brief
- Build path
- Website, planner, stock, costs, and a reusable skill
- You need
- A computer and permission to install one tool
Maya already knows how to ask AI questions. Now she wants help creating what her business needs.
The whole method
One folder becomes the workshop.
- 1
Explain the result. Give the agent the facts and limits.
- 2
Inspect the plan. Correct bad assumptions before they become files.
- 3
Approve a small build. Let the agent create and test one useful piece.
- 4
Open the result. Check what changed, fix it, and leave a checkpoint.
Fundamentals, then the build path
We learn only what the project needs next.
The terminal can do thousands of things. We introduce each command only when Maya needs it, then let one AI agent read, create, edit, run, and check the files inside her project folder.
Fundamentals
Chapters 1 to 3
- 01Open the project folderCreate the folder and start one agent inside it.
- 02Set the working rulesDecide how the agent explains, asks, edits, and reports.
- 03Turn the idea into factsCreate the brief and safe sample business files.
Customer build
Chapter 4
- 04Build the customer websiteCompare three designs, choose one, and create the page.
Operations build
Chapters 5 to 7
- 05Build the kitchen plannerUse invented orders in a local, unpublished teaching prototype.
- 06Add stock and cost estimatesConnect recipes, inventory, and weekly demand.
- 07Save the weekly method as a skillTest it, fix one problem, and leave a checkpoint.
Before chapter one
Let the AI work on the project, not just talk about it.
Maya can already ask AI for ideas and advice. A terminal agent goes further. It works with the actual files inside one project folder, so it can create them, connect them, run the result, find problems, and fix them. Maya still decides what gets built and approves anything that needs more access.
The AI tells Maya what to make.
It gives her words, code, and instructions. She then has to put every piece in the right place and work out what connects to what. This is usually where a simple job develops seventeen tabs and a small administrative afterlife.
The AI helps Maya make it.
It can build a page, run it, inspect an error, correct the files, and report exactly what changed. Maya sees the result and keeps control of the decisions.
Where the full course ends
The permission setup is only the front door. The remaining chapters guide Maya as she creates and checks a customer website, a local, unpublished teaching planner using fictional orders, stock, and costs, and a reusable weekly workflow saved inside the project.
PowerShell on Windows
What this setup needs
A computer, permission, and the right account.
Terminal and PowerShell are already installed. We add one AI agent that can work with the files in a folder.
- 1
A Mac or Windows laptop or desktopA phone or tablet cannot complete this course.
- 2
Permission to install one AI agentA work or school computer may require its owner or IT team to approve this.
- 3
Internet access and a supported accountThe internet is needed for installation and sign in. The agent cards below explain which account each one accepts.
You do not need to write code yourself. You will copy one command at a time and check what appears. Chapter one creates the empty project folder.
Before we build
Install the tool first. Open the project later.
The installer adds the command and its supporting files outside the Sunday Oven project, using its configured command location. If the Codex installer asks Start Codex now?
, choose No or press Return. That keeps installation separate from the agent session, which we start later from the checked project folder.
- 1Install the command
- 2Check the version
- 3Create the project folder
- 4Open Terminal there
- 5Start the agent there
Why the distinction matters: the folder tells the agent where this project lives. The permission or sandbox setting decides what it may do without asking. Chapter one checks both before a file can change.
What leaves the folder: this course does not publish the finished files. Your computer may still sync Documents according to its own settings. While the agent is running, it sends your prompts and relevant project context to its provider under that provider's data terms. Use only the fictional course data here.
Prefer a normal app window? Three ways to reuse the method. The guided course below remains terminal-specific.
These apps can reuse the course prompts and decisions. Their setup screens and file handling differ, so the step checks and completion gates below will not match them.
Important: these are alternatives for using the method, not alternate completion routes. The guided steps below require Terminal or PowerShell.
1 · Choose this computer
2 · Choose one AI agent
Setup guidance checked against the providers’ official documentation on 28 July 2026. Their screens and commands can change. Use the linked setup guide if yours differs.
0 of 7 complete
Keep real data out of the prototype
Real workflow. Fictional customer data.
This is a local, unpublished teaching prototype. The website can prepare an email request, but it does not take payments or store public orders. Maya, The Sunday Oven, and every customer name, order, stock figure, and cost shown in the course are illustrative.
Before you paste anything into the project, replace real names and contact details with invented ones. Leave passwords, payment details, home addresses, health information, private messages, and anything you lack permission to use out completely.
Official product guidance used in this course
Normal app alternatives: ChatGPT Projects and Claude Desktop Code.
Codex: CLI guide, AGENTS.md guidance, CLI commands, and skills.
Claude Code: setup, CLAUDE.md guidance, skills, and permission modes.
Gemini CLI: setup, GEMINI.md guidance, skills, and commands.