How an agent works

Watch: How Claude Code Works

https://www.youtube.com/watch?v=6bs5b4FltCU&list=PLmWCw1CzcFilebjK89WLb5cAvM8K0cLB3&index=3

An AI coding agent is a language model in a loop with tools. It can read files, run commands, edit code, and observe the results — then decide what to do next. It is not "thinking about your whole project": it works through a context window, a limited working memory that holds the conversation and whatever files it has read.

Three consequences that explain most confusing agent behavior:

It "forgets." When the context window fills up, older material gets compressed or dropped. A long chat slowly degrades: the agent loses details you told it an hour ago.
It only knows what it has read. On any project bigger than a few files, the agent sees only the parts it has pulled into context. It may "not know" a function exists and write a duplicate.
It can be confidently wrong. A language model always produces plausible output. When it lacks facts, plausible and true diverge — that's a hallucination. It will not look unsure. Your review is the safety net.

Watch: Context Management

https://www.youtube.com/watch?v=eW3oTyfeWZ0&list=PLmWCw1CzcFilebjK89WLb5cAvM8K0cLB3&index=7

You are the architect, the agent is the executor

The working relationship this whole track builds on:

You decide what to build, set boundaries, review results, and own the outcome.
The agent executes: writes drafts, searches the codebase, runs commands, produces options.

The responsibility for the result never transfers to the agent. "Claude wrote it" is not an explanation for broken code in a PR — you shipped it.

The HackYourFuture curriculum is licensed under CC BY-NC-SA 4.0 *https://hackyourfuture.net/*

CC BY-NC-SA 4.0 Icons

Built with ❤️ by the HackYourFuture community · Thank you, contributors

Found a mistake or have a suggestion? Let us know in the feedback form.

Week 1

Watch: How Claude Code Works

Watch: Context Management

You are the architect, the agent is the executor