THE CONCEPT LIBRARY
One idea at a time.
AI basics you can understand and use. With a short, an example and a small exercise.
-
▶ 60 SEC
Tools
MCP Basics: Connect an AI Agent to External Tools
Learn how MCP connects coding agents to external data and functions—and why access, permissions, and results still need review.
-
▶ 60 SEC
Context & memory
Context compaction: Continue a long AI session cleanly
Learn how context compaction reduces a long AI session to a smaller working state and helps you continue a coding task reliably across the handoff.
-
▶ 60 SEC
Work safely
Review AI Code With Diffs and Tests
AI-generated code is a proposal, not a verdict. Learn a simple review loop: understand the change, inspect its diff, run checks, and decide deliberately.
-
▶ 60 SEC
Work safely
Plan Before Code: Add a Review Checkpoint for AI Changes
Ask a coding agent to investigate and propose a plan before editing. Review the approach, then inspect the diff and run relevant tests.
-
▶ 60 SEC
Better prompts
Prompt iteration: improve AI work step by step
Use a draft–review–revision loop to improve AI output against clear criteria.
-
▶ 60 SEC
Tools
Skills: Save a repeated AI workflow
Turn a repeated checklist or multi-step prompt into a named, reusable procedure for an AI coding tool.
-
▶ 60 SEC
Context & memory
Context Windows: What AI Can See Right Now
Learn what fills an AI context window, why long coding sessions can lose earlier requirements, and how to hand off a focused task with relevant evidence.
-
▶ 60 SEC
Tools
Agent permissions: instructions are not access control
Learn how coding-agent permissions separate a requested action from an action the client may actually run.
-
▶ 60 SEC
Better prompts
Use Examples and Output Formats to Make Prompts More Predictable
Show the shape of a useful answer, then state the format you need. This simple prompting pattern helps make recurring developer tasks easier to review and reuse.
-
▶ 60 SEC
Tools
Chat, Coding Agent, or Model: Choose the Right Layer
Learn when to use chat, a coding agent, or model selection—and how to set a review boundary for every AI-assisted change.
-
▶ 60 SEC
Work safely
Check AI Code Before You Keep It
A simple review routine for AI-generated code: inspect the diff, run checks, and judge whether the change fits the task.
-
▶ 60 SEC
Context & memory
AGENTS.md: Project Rules for Coding Agents
Use a small project file to give Codex durable local guidance without replacing a clear task, testing, or code review.