Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

Learn AI Coding

via Boot.dev

Go to class Write review

Overview

Stuck in Tutorial Hell? Learn Backend Dev the Right Way

Boot.dev teaches Python, SQL & Go through projects you won't want to stop. 25% off with code CLASSCENTRAL.

Get 25% Off

Learn to use agentic AI tools effectively as a software engineer, not just a vibe coder. You'll understand how models, tools, prompts, context, rules, skills, MCP servers, and plugins all work under the hood. Coding agents let you ship code faster than any human could before. But without a talented human developer overseeing them, AI will make codebases worse, and bad code plus AI means more bad code.

Syllabus

Agent Harness — Install OpenCode and see how coding agents differ from chat and autocomplete, whether you run them in a terminal, a GUI, or on a remote server.
Models and Providers — Pick the right model for the job without lighting money on fire. Compare providers, token costs, reasoning models, and open-weight options you can host yourself.
Tools — Agents are only as useful as the tools they can call. Use the built-in tools, write your own, and lock things down with permissions and sandboxing.
Prompting — Write focused, well-scoped prompts, and learn when to plan, steer, cancel, or queue work while an agent is running.
Agents — Build custom agents and subagents for specific jobs, each with its own prompt, tools, step limits, and permissions. Plus, when (rarely) to touch temperature.
Context — Context is the agent's working memory, and it fills up fast. Manage sessions, compaction, forking, and undo so your agent doesn't lose the plot.
Rules — Give every agent session the same project knowledge with a well-written AGENTS.md file that you keep up to date.
Feedback Loops — Type checkers, formatters, and tests let an agent catch its own mistakes before you have to.
Skills — Package your repeatable workflows as skills, custom commands, and dynamic prompts, and figure out when a skill beats a tool or a dedicated agent.
MCP — Connect agents to outside services with the Model Context Protocol, including OAuth and code mode, without stuffing your context window full of tool definitions.
Computer Use — Let agents click around in a real browser using accessibility trees, vision models, Playwright, and Chrome DevTools.
Plugins — Extend OpenCode with your own hooks, transforms, and tools, then publish your plugin for other developers to install.
Best Practices — You're still accountable for every line an agent writes. Debug with evidence, refactor safely, ship small changes, and run multiple agents without making a mess.

Taught by

Dax Raad

Reviews

Start your review of Learn AI Coding

Go to class

Taught by

Never Stop Learning.