1. 🚀 Claude Code — not just a chat, but an AI agent
During the first level you got acquainted with an AI agent — a tool that can work directly with the files of your project. A regular AI chat returns text or code, which you then have to copy, save, and launch yourself. The agent works differently: it creates files itself, modifies them itself, and checks the result itself 👨💻
You state the task in plain words, and the agent performs the necessary actions on its own. And it works not somewhere on someone else's website, but in a real project — everything it creates stays with you.
However, the agent does not know in advance what exactly you want. The clearer the task is described, the better the result will be 🤔 You have already seen this principle in action: you described a game in a few sentences — and got exactly that.
2. 🏆 What you made in a single level
Let's just look at the score.
🎈 A clicker game. Neon balls, particles, sound, a timer, a score. You described it in words — and it appeared.
🧱 An arkanoid. A paddle, a wall of bricks, bounces. A second game on the same day.
✨ Golden bricks. You took a finished game and reworked it to your own taste with a single phrase.
📄 A cheat-sheet page. Not a game, but made with the same command: you said it in words — you got a finished page.
🧹 Order in the project. The files are sorted into folders, and that wasn't done by your hands either.
🔧 A developer's workplace. Eighteen programs: the AI agent itself, two development environments with plugins, languages, build utilities, a version control system, speech recognition. All installed, wired together, and signed in. Usually this eats up a full working day, and for more experienced people — more than one. It took you half an hour, and most of that time you were just waiting.
Two games, a cheat-sheet page, and a fully assembled workplace. In a single level. Without writing a single line of code by hand.
Just yesterday each of these things would have meant weeks of studying HTML, CSS, and JavaScript. Today it means one clearly worded phrase.
3. 🥇 What you now know
💻 A program is files. Not the cloud and not magic. Any project is a folder, and inside it are files you can open and look at. That's exactly where the agent works.
🤖 A chat hands you text, an agent does the work. That's the whole difference, and that's exactly what turns a conversation with an AI into a working tool.
💬 A session is your working dialogue with the agent. Inside a single session it remembers what has been discussed: to rework a game, you don't need to explain it all over again.
▶️ Launching. Right-click an HTML file → Run — and the page opens like any ordinary page in the browser.
🌍 The terminal. The second way to control a computer — not with a mouse, but with commands. It always has a current folder: pwd shows where you're standing, ls shows what's around you, cd moves you, mkdir creates a folder. And the black window isn't scary anymore — which is probably more important than the commands themselves.
4. 🎮 The level finale
Two things left. And both of them are fun.
All level long the AI was working for you. Now you'll build an AI that plays against you — and you won't be able to beat it. That's the whole point.
5. 🚀 You are moving on to the next level
Now we can do the final count: three games, two pages, and a developer's workplace. In a single level.
Congratulations, you have completed the first level. Just one level — and there's already a shelf with your games sitting in the project.
Great! Of course, the next levels will not be that easy, but their difficulty will increase gradually. The same goes for practical tasks. Just like in a gym: you slowly increase the load, and after six months a beginner can bench-press a 100-kilogram barbell 💪
Boring lessons stayed in the 20th century. Can you imagine that — first spending months learning code, and only then creating your first game. Dinosaurs were probably still walking the streets then 🦕
And tomorrow things get more serious. You were making toys for yourself: if it came out crooked, you had a laugh and redid it. Next comes work you'll have to show to people. And it will turn out that the hardest part is not getting the agent to build a page at all 😉
You are moving on to the next level! 🚀
6. What to read next
Want to understand what happens between your sentence and the answer? What one step of the agent is made of, why it opens the same file twice, and why "done" from an agent is not the end of the work — it is all laid out in a separate article, "The agentic loop: what Claude Code does while you wait." From your request to your review.