Course Map — GenAI with JS, as a Reading Track

This track replaces watching with reading. It follows the session order of the GenAI with JS 2026 cohort, one chapter per session (multi-part sessions are merged into one chapter), so you can read the chapter, build the checkpoint, and only open a recording when you want to see a specific demo.

These chapters are written from first principles in the same style as the AI Engineering track. They are not transcripts of the sessions. Concepts and project scopes follow the course syllabus; the code is our own. Where the AI Engineering track already explains the theory in depth, the chapter links to it and spends its words on the JavaScript implementation instead.

---

How to Use Each Chapter

Every chapter has the same three layers:

1. The mechanism. Why the thing exists, what it does physically, and where it breaks. Read this first, without a laptop.
2. The build checkpoint. A numbered set of steps for that session's piece of the project, each with a "done when…" condition. Do this without the chapter open where you can — rebuild from understanding, not from copying.
3. Practice problems. Failure analysis and design questions. If you can answer them out loud, you own the chapter.

A chapter takes roughly 25–35 minutes to read and 2–4 hours to build. That is faster than a 2–3 hour recording plus the build, and you retain more because you build without a video to lean on.

---

Session → Chapter Map

Chapters 1–4 cover sessions you have already watched. They focus on the JavaScript implementation and lean on the AI Engineering track for theory:

• Session 1 theory → AI Ch 1 — The Frozen Oracle
• Session 2 theory → AI Ch 2 — Prompting Is Programming
• Session 3 theory → AI Ch 5 — From Answering to Acting
• Session 4 theory → AI Ch 4 — Memory the Model Was Never Given

---

Schedule to Finish by December

Two chapters a week, one on a weekday evening and one on the weekend, finishes the main track in the first week of December and leaves three weeks for the bonus material, deploying the projects, and exam review.

---

The Two Gaps to Fill Yourself

The course syllabus is broad and current, but two production topics get little time. The AI Engineering track already covers both, so apply them to every project in this track:

• Evals — every project here should ship with a small eval set: 20–50 real inputs and a check for each. See AI Ch 7.
• Prompt injection — every project that reads documents, repositories or web pages reads untrusted text. See AI Ch 8.