Zero to shipping, in 14 steps
A hands-on guide to building with AI, from zero to shipping.
One path for developers, read in order: how to ask, what the model sees, how to work with agents, how to build AI into your own software and prove it works, and how to ship it safely with a team. Each step builds on the one before. Already know the basics? Jump in anywhere.
The path
Part 1
Understand
What AI is good for, how to ask, what it sees, and which model to use.
Part 2
Work with agents
From asking questions to handing over whole tasks.
Part 3
Build
Put a model inside your own software, and prove it works.
Integrate
Put AI into a real application — the API, features, your data.
Step 9Projects
Build an AI tool end to end, then a small agent of your own.
Step 10Evals
Know whether a change made it better — with a test set, not a feeling.
Step 11MCP Servers
Build a Model Context Protocol server, step by step.
Step 12Scaffolder
An MCP server that builds MCP servers for you.
Part 4
Ship it
Keep it safe, and bring a team along.
The shortcut: let AI build the AI tooling
The Scaffolder is an MCP server that uses Claude to generate a complete, runnable MCP server from a plain-English description. Describe the task — get working code.
Meet the Scaffolder →Built by
Floyd Baker
Bakeriot is built and maintained by Floyd Baker — a developer who has spent years shipping production web software, IoT, and robotics projects. This site is where he distills what actually works when building with AI into something practical you can use today.