Intro: AI-Assisted Development

Hello, everyone! Welcome to this course designed to empower our Development and Quality Assurance teams to build software with AI.

AI coding tools represent a major shift in how we approach software development. They move beyond being a passive editor and become an active partner that can understand our codebase, draft changes across many files, and dramatically increase our velocity.

This is a course about a skill, not a product

We will use Claude Code as a running example throughout — it's popular, concrete, and easy to show. But the goal is not to make you good at Claude Code. The goal is to make you good at AI-assisted development, so your skills transfer the day you switch tools or your team runs several at once.

Today's assistants and agents — Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Windsurf, Cline, Aider, Continue — are all instances of the same category. They share the same primitives: instructions/rules, context, a chat/agent loop, and tool integrations. Learn the primitives and every tool becomes "the one you already understand, with different buttons."

🧭 North star: agents.md — an open, cross-tool standard for telling any agent how to work in your project. We'll build on it in Lesson 02.

The honest picture: AI shifts the bottleneck, it doesn't remove it

It would be easy to sell you pure speed. Here's the more useful truth:

AI moves the hard part of engineering from writing code to reviewing and verifying it. The AI can produce a plausible diff in seconds — but plausible is not the same as correct. 2026 evidence is sobering: AI-assisted code introduces roughly 2× the critical vulnerabilities and logic errors of human-written code. It fails differently — often syntactically perfect and semantically wrong.

So the productivity is real, but it comes with a job description change:

🎯 You own correctness and security. The AI is accountable for nothing. You direct it, and you verify it. Speed without verification is just faster bugs.

This is why later lessons on verification (11) and security (12) matter as much as the prompting lessons — they're where a competent AI engineer separates from the rest.

The mental model for the whole course

Keep this picture in your head every time you use one of these tools:

Treat the AI as a fast, literal, tireless — but un-accountable — junior engineer.
It is astonishingly quick and knows a huge amount, but it will do exactly what you say (not what you meant), it will confidently guess when unsure, and it never carries the blame. Your job is to direct it well and verify its work — the same as with any junior on your team, just faster.

Everything in this course is a way of directing better (rules, context, prompts, patterns, specs) or verifying better (review, tests, security).

The Goal: A Unified Boost in Efficiency and Quality

By the end you will be equipped with a durable, tool-independent workflow that makes all of us more productive, more precise, and more accountable than before — a workflow that survives the next UI redesign and the next model release. Let's start building smarter.