AI · Knowledge track

Understand
coding agents.

Coding agents move beyond autocomplete: they can inspect a repository, plan work, edit files, run tools and iterate on results. This track explains the architecture and the controls that make that useful rather than reckless.

Start here

What is an AI coding agent?

Build a precise mental model of agentic coding without treating every AI code feature as an agent.

Read the concept →
Distinction

Agent vs copilot

Compare inline assistance, chat, agent mode and asynchronous coding work.

See the differences →
Architecture

How the agent loop works

Goal → inspect → plan → act → observe → verify → repeat.

Follow the loop →
Control surface

Context, tools & permissions

Why repository context, tool boundaries, approvals and verification determine what an agent can safely do.

Understand the controls →
EDITORIAL SCOPE
This track explains durable concepts first. Product interfaces change quickly, so product examples are used to illustrate mechanisms rather than as a permanent feature matrix.