Course Overview¶
Build With AI is structured as a complete builder path, not a prompt cheat-sheet. The goal is to help you build judgment around AI agents — not just memorize tricks, but understand why an agent behaves the way it does and what to do next when it goes sideways.
Every lesson trains the same loop you will use on real work:
- Vision — decide what you're building and what "done" looks like before you type a word.
- Context — give the agent what it needs to succeed: goals, constraints, examples, memory.
- Build — let the agent work; direct, don't micromanage.
- Verify — inspect the result with relevant human and automated checks. Treat the agent's report as a claim to test.
- Ship — put it in front of real people, capture what you learned, and feed it back into your context.
Audience¶
This course is for:
- Curious non-technical people who use ChatGPT or Claude and want to actually build — no coding background assumed.
- Creators and entrepreneurs who have ideas and an audience but no dev team.
- Software engineers whose AI usage is still copy-paste chat and who want the full agentic stack.
- Technical folks from data, IT, security, or ops who want to direct agents, not just read about them.
If you can type in a browser and follow instructions carefully, you have enough background to start.
Prerequisites¶
None for the no-code lane. The engineer lane assumes you can read code and use Git, but every tool is introduced before it's used. If you hit an unfamiliar term, check the Glossary first — it exists so vocabulary never blocks you.
You will need:
- A computer you control (Mac, Windows, or Linux).
- An eligible Claude plan or API-backed Console account — the free Claude.ai plan does not include Claude Code. The course teaches transferable concepts through this concrete implementation.
- Usually 45–60 minutes for Setup before Chapter 3, longer if account or platform troubleshooting is needed.
Time expectations and routes¶
These planning ranges come from the July 23, 2026 course inventory: 54 lessons, 49 practice sections, 11 checkpoints, two diagnostics, setup, and the capstone. They are estimates, not completion promises. Reading speed, tool access, revision, accessibility needs, and project scope can move them substantially.
| Route | Essential work | Planning range |
|---|---|---|
| Reading/reference | Read all lessons and use the unscored understanding questions. No setup, assessment, videos, or builds required. | 12–18 hours |
| Assessed practice | Reading, entry and exit diagnostics, setup or simulation baseline, one bounded practice artifact per track, all 11 chapter checkpoints, scoring, and revisions. | 30–50 hours |
| Assessed practice + capstone | Everything in assessed practice plus one scoped capstone and peer review. | 40–75+ hours |
Planning arithmetic: the reading range allows roughly 13–20 minutes per lesson. The assessed range adds 18–32 hours for diagnostics, setup/simulation, four selected practice artifacts, 11 checkpoints, scoring, and revision. The capstone adds a 10–25+ hour scoped build and peer pass. These are workload assumptions to replace after a real timed pilot, not measured completion data.
Linked videos, additional "Try it now" builds, repeated tool experiments, and portfolio polish are optional extensions unless a checkpoint explicitly requires their evidence. They commonly add 15–30+ hours. New builders may choose more of that practice; experienced builders may test out through strong checkpoint evidence.
The assessed route is the minimum viable evidence path, not a claim of mastery. Pause after Chapters 3, 6, 9, and 11 to review your evidence and decide whether to continue, repeat a weak competency, or stop with a useful partial course.
Course tracks¶
| Track | Chapters | Focus |
|---|---|---|
| Foundations | 1–3 | The agentic shift, how AI actually works, terminal + Git + Claude Code setup |
| Context engineering | 4–6 | The agentic loop, CLAUDE.md, memory layers, second brain, everyday builds |
| Agentic engineering | 7–9 | Skills, MCP and tools, hooks, validation, and guardrails |
| Orchestration and shipping | 10–11 | Subagents, agent teams, parallel work, launching, and selling |
For the complete chapter sequence, use the Syllabus. When you finish the chapters, the Capstone is where you prove it to yourself — and produce the receipt you can show anyone else.
Your evidence path¶
Use the course as a measured practice, not a reading streak:
- Take the entry diagnostic before reading worked answer patterns. Save the dated response and disclose any prior pattern exposure.
- Complete Setup, then establish the Starter Project baseline.
- At the end of every chapter, complete its performance checkpoint. The scoring guide is available before you begin; open worked patterns through Review After Your Attempt only after saving the first attempt.
- Use the Recovery Guide whenever a lab or agent change fails.
- Complete the paired exit diagnostic, which uses different situations on the same competency scale. Compare evidence rather than confidence, then decide whether to begin the Capstone.
The competency framework explains exactly what the course measures. Scores are feedback about the next practice step, not credentials or predictions of job performance.
How the two lanes work¶
Lessons are written for everyone. Where the path genuinely differs — setup steps, practice work, depth of code — you'll see tabbed sections:
- 🌱 No-code lane — plain-language steps, zero assumed background.
- ⚙️ Engineer lane — deeper mechanics, code, and edge cases.
Read your lane. Skim the other one anyway — knowing how the other half works is part of the skill.