About this course

Why this course exists

Most writing about AI coding tools does one of two things: it sells the next tool, or it explains how the models work inside. A working engineer needs neither. You can already write code; what is missing is a reliable method for operating an agent — giving it the right context, specifying work it can carry out, verifying output you did not write, and running more than one agent without losing control. This course teaches that method and stays narrowly on it.

Who it is for

Software engineers who already ship code and want to operate AI agents at a senior level — both individual contributors and the engineering leaders bringing agents to their teams. It assumes you can read and write code, and that you have an AI coding agent and a real codebase to practice against. It is not an introduction to programming, and it is not a study of model internals. If your question is what a large language model is or how a transformer works, that belongs to a different course; this one begins where that ends.

Who wrote it

This course is written by a former senior director at Meta and Snapchat who has helped engineering teams become AI-native. The method here is the one he has used and taught across those teams: tool-agnostic, grounded in how strong engineers actually work, and focused on judgment rather than novelty.

What you will be able to do

By the end of the course you will be able to:

• read an agent's working loop and predict where it will fail, instead of being surprised by it (Module 0);
• write a project instruction file that controls what the agent sees, so it opens the right files on its own (Module 1);
• turn a vague request into a specification an agent can execute and you can check (Module 2);
• verify work you did not write using tests, evals, and gates rather than by reading every line (Module 3);
• shape the agent harness with tools, skills, permissions, and restraint (Module 4);
• run several agents in parallel on divisible work without losing control (Module 5);
• fit agents into your real workflow so that each task makes the next one cheaper (Module 6);
• decide, with honest cost math, when an agent is worth using and when it is not (Module 7).

How the material is kept honest

The course follows a cite-or-cut discipline. Specific numbers and factual claims are tied to primary sources, and anything that cannot be supported is removed rather than hedged. The result is a verifiable account of what works, with the underlying sources visible.