Build AI employees that run on a schedule, not on your memory.
A few narrow assistants that read your real numbers on a schedule and hand you a meeting you can act on. You have never written code, and you will not start here.
One-time payment. Lifetime access, and the lessons stay correct as the tools change. Last verified 2026-08-26.
8
Modules
26
Lessons
6
Templates
The method
One loop, taught until it's a habit
Every module, lab, and the capstone run this same sequence. By the end it should feel automatic, including the parts most people skip.
- 01
Inspect
Read the code before changing it. Make Claude prove it understands.
- 02
Plan
Agree on the approach in plan mode, before a single file is edited.
- 03
Build
Small, reviewable changes, not a thousand-line drop.
- 04
Review
Read every diff. You are still the engineer of record.
- 05
Test
Prove it works with tests and a real browser pass.
- 06
Ship
Deploy behind a plan you can undo.
One path, and one optional add-on
The course is the board path. The developer material is separate, so you are not paying for lessons about SQL you will never open.
Included, this is the course
The Board Path
Go from an empty folder to an AI board that reads your real numbers on a schedule and hands you a meeting you can act on.
- Start from an empty folder, not a codebase
- The four files your board reads
- Define seats and what each may not do
- Put a run on a schedule
- Permissions, before anything runs unattended
- Tell a useful run from a plausible one
You run something small, you have never written code, and you want AI doing standing work instead of one-off chats.
$57
Optional add-on
The Dev Pack
The developer curriculum: the disciplined agentic coding workflow, on real code in real stacks.
- Working with files, real dev workflows, and stack-specific tactics
- The full guided capstone build, start to finish
- Code review, testing, and git discipline with an agent
- Everything unlocks inside the course you already have
You write software too, and you want an AI workflow you can defend in review.
$97 · not on sale yet
Outcomes
What you'll be able to do
Concrete capabilities, not vibes. Each one maps to a lesson you can point at.
- Install and log in to Claude Code the way the current docs actually recommend
- Pick the right permission mode for the task instead of approving everything
- Run long sessions without losing context or your place
- Write a CLAUDE.md that measurably improves what Claude produces
- Review an AI-written diff critically enough to sign your name to it
- Debug a real failure instead of regenerating and hoping
- Write tests that would actually catch the bug you just fixed
- Package a repeatable workflow as a Skill your team can share
- Wire Claude Code to your own tools with MCP
- Split work across subagents and agent teams when a task is too big for one context
- Ship a complete feature end to end in the capstone, with a rollback plan
Curriculum
Every module, in order
Written lessons with per-lesson quizzes and progress tracking. Facts are checked against the official documentation and dated.
Start Here: Your AI Board
Go from an empty folder to a working AI board. What a board is, where it lives, how to open it, and how to tell a useful run from a plausible one. No code.
Core Concepts
How context actually works, how to write a prompt that gets the change you meant, and how to manage a session over hours instead of minutes.
Project Context & Scope
Write a CLAUDE.md worth loading, chain work across steps, run review passes, and stop scope creep before it lands in your diff.
Practice & Safety
Daily habits, a security mindset for an agent that can run commands, usage awareness, and shipping with a review you trust.
Extending Claude Code
Skills, hooks, MCP, subagents and agent teams, plugins, and running Claude Code beyond the terminal, IDE, desktop, web, and CI/CD.
Professional Practice
Using this workflow in client and team settings: scoping honestly, internal tools, and turning a repeatable process into a service.
Memory: How the System Remembers
Four files in a folder work for about six weeks. Tiers, an index, an inbox that drains, dated archives, and the split between what your board proposes and what you promote.
Seats: Charters, Goals, and Promotions
Who a seat is, what it may do alone, what floors it is measured against, and how it earns more autonomy without you getting burned.
Fit
Who this is for, and who it isn't
We would rather you skip this than buy it and feel misled.
Buy this if you're…
- Solopreneurs and small-business owners who have never written code and want AI doing standing work, not one-off chats
- Operators and consultants who want to run this on their own business before they run it for anyone else
- Anyone tired of AI that only helps on the days they remember to open it
Skip it if you're…
- Anyone who wants AI to run the business for them. This builds a board that produces analysis and drafts. The decisions and the work stay yours
- Anyone looking for passive income or a guaranteed outcome. This teaches a system, and what you do with it is on you
- Anyone who wants to learn to program. The Dev Pack assumes you can already read code, and nothing here teaches you to write it from zero
- Teams needing Anthropic's official enterprise training or support. That comes from Anthropic, not from us
Maintenance
The part that usually rots
These tools change weekly. Most courses quietly go wrong and nobody tells you. This one carries a date you can check before you buy.
Verified against the official docs
2026-08-26
Not a launch date. The last time every factual claim in the curriculum was re-checked against the source and the changes were published.
What that actually catches
The last pass found that sessions now start in a different permission mode than the course described. That is the first screen a new reader sees, and it would have contradicted the lesson before they finished it.
It also found a correction that had itself gone stale: a note saying a command had been removed, when the command had come back. Being wrong about your own fix is the failure mode nobody checks for, which is exactly why the check is scheduled instead of remembered.
See the live numbers, including the unflattering onesThe Build Lab
Live · no date set yet
A live, small-group session where we build one real feature end to end and you watch every decision, including the ones that go wrong.
No date is set yet. Join the list and you'll hear when there is one, before it goes anywhere else. The waitlist costs nothing and holds nothing, no deposit.
Join the waitlistPricing
One price. The whole course.
No tiers, no subscription, nothing held back for a higher tier. There is one other thing we sell, the live Build Lab, and the course is complete without it.
One-time payment. Lifetime access, and the lessons stay correct as the tools change.
- The board path: from an empty folder to a board that runs on a schedule
- Written lessons, kept current against the official docs and dated so you can check
- The four board files as downloads, ready to fill in
- Per-lesson quizzes and progress tracking
- Lifetime access, and corrections whenever the tools move under a lesson
Before you buy: this course drives an AI coding agent, and the one it is written against needs a paid plan billed separately by Anthropic. The free tier will not run it. Everything else you need is a folder and a text editor.
Secure checkout by Stripe · 14-day refund, just email us
The Build Lab is sold separately and is not required for anything in the course.
AI by Design is Terry Scott's practice helping teams put AI to work in real production systems. My AI Board is the self-paced course arm of that practice: the same workflow, written down.
FAQ
Straight answers
Including the questions that don't flatter us.
Do I need an Anthropic API key?
No. Claude Code requires an account, and most learners log in with a Claude subscription (Pro, Max, Team, or Enterprise). A Claude Console account with pre-paid credits also works, as does access through a cloud provider your employer configures. The course covers logging in; the subscription or credits are billed by Anthropic, not by us.
Is this affiliated with Anthropic?
Independent educational product by AI by Design. Not affiliated with or endorsed by Anthropic. Everything here is our own material, and we link to the official documentation as the authoritative source.
Are there videos?
Not yet. The written lessons are complete and are the course. We are recording original walkthroughs now, and they will be added to the relevant lessons as they ship, included at no extra cost, because they are lessons you already bought presented another way, not new material. We would rather ship no video than someone else's video.
Will this go out of date?
Parts of it, constantly, Claude Code ships changes weekly. That is why every factual lesson links to the official docs and carries a last-verified date, and why updates are included for life. Lessons were last verified against the official documentation on 2026-08-26.
What experience do I need?
For the board path, none. It starts with an empty folder and assumes you have never written a line of code. The Dev Pack is the other half, and that one does assume you can read code and use a terminal.
I already use ChatGPT every day. Why do I need this?
Because ChatGPT cannot start work without you. It waits for you to open it, and it cannot read or write the files your business actually lives in. This teaches you to build something that runs on a schedule against your real numbers and hands you a decision when you sit down. If what you want is a better chat, keep the one you have. This is for when you want work to happen on days you never log in.
Is this only for one AI tool?
No. The system is a folder of markdown files and a schedule, so it runs on any agent that can read your folder and act on a schedule. The lessons are written and verified against one implementation so the instructions are concrete rather than vague, and each one names the capability it needs so you can map it onto whatever you use.
Will this still be right in six months?
That is the actual product. These tools change weekly, so the course carries a verification date, the corrections are published rather than quietly patched, and you can check the date before you buy. It has already caught its own stale correction once: a note saying a command had been removed, when the command had come back.
Do I need to use a terminal?
No. The Desktop app is the recommended way in, and it looks like an ordinary application. The terminal is one of three doors and the board path never requires it. If you want to use it, the Dev Pack covers it properly.
What is the Dev Pack?
The developer half of the material: working with real codebases, debugging, tests, Git, stack-specific tactics, and the full capstone build. It is a separate add-on rather than part of the course, because a solopreneur building a board should not be paying for lessons about SQL indexes.
What about the live Build Lab?
It is a separate, live, one-off session, we build one real feature end to end and you watch every decision, including the ones that go wrong. It is not scheduled yet, so there is a waitlist rather than a date, and the waitlist costs nothing. The $97 course is self-paced, complete on its own, and does not depend on the Lab.
Can I get a refund?
Email terrysc107@gmail.com within 14 days and we will refund you. No form to fill out and no questions designed to talk you out of it.