Complete one lesson every day to keep the streak going.
Su
Mo
Tu
We
Th
Fr
Sa
You earned a Free Pass!
Free Passes help protect your daily streak. Complete more lessons to earn up to 3 Free Passes.
Welcome to Claude Code, the Highly Agentic Coding Assistant. This short course is built in partnership with Anthropic and Anthropic's Elie Schoppik is back to share with us best practices for how to use Claude Code. I'm really excited about this short course. Claude Code's my personal favorite coding assistant right now, and it has boosted my and many other developers' productivity by a large margin. And it is a tool with a lot of depth to it. And so I want to get together with Anthropic, to teach, hopefully the definitive course on all of the most important ideas behind how to use it in a systematic way. Thanks, Andrew. I'm excited to be back here and start like you mentioned from explaining what the tool is, how it works, all the way towards using it in parallel with many different tools including Git worktrees and MCP servers. What we've seen over the last couple years is AI assisted coding has evolved rapidly. It started from maybe people asking LLMs occasional coding questions. suggestions to then GitHub autocomplete to then, your various tools that became more and more autonomous. And Claude Code, when it was released, was definitely a step up in terms of the degree of agency or the amount of stuff that the coding assistant could do by itself. And so I think many people were surprised that you could set a task that Claude would work on for many minutes or sometimes even more than a few minutes. And now there are developers that are orchestrating not just a single Claude instance, but even several of them. working in parallel on different parts of a codebase. But coordinating all this has to set the best practices that is not widely known and if you have not worked with people close to the best practices, I think there could be a big uplift still for mastering these best practices and knowing how they drive that amazing productivity that I'm seeing many developers have using Claude Code. So as we start to talk about those best practices, a key tip for working with Claude Code is providing clear context to help Claude code achieve the task you want efficiently. This means pointing Claude code to the relevant files, clearly describing the features and functionality that you want and making sure that you're properly extending the capabilities of Claude code with MCP servers and other tools in that ecosystem. In this course, you'll apply those best practices to three different examples. We'll start with a RAG chatbot and you'll implement the features from the front end to the back end, including refactoring code, writing tests, and then using the GitHub integration to work with pull requests and fixing issues. You'll make use of many Claude Code features like planning, thinking modes, creating parallel sessions, and managing Claude's memory. For the second example, we'll shift gears and work with Jupyter notebooks to explore e-commerce data. We'll refactor notebooks using Claude Code, remove redundant code, and create powerful dashboards with web applications. Finally, we'll move to create a visual mockup based in Figma and use Claude Code, the Figma MCP server and a different MCP server to import the design, iterate, test, and build agentically a powerful front-end application. If you're not currently a Claude Code user, I think learning this set of ideas will give you a meaningful acceleration in the way that which you can engineer systems. And even if you are a current Claude Code user, I think having Ellie share these best practices with you in a comprehensive and systematic way, will, I hope, leave you with quite a few new things you try that will be useful. your work. I'd like to thank from DeepLearning.AI, Hawraa Salami, who had contributed to this course. In the next video, Elie will share Claude Code's underlying architecture, and you might be surprised by how simple the architecture is. Claude Code relies on just a small number of tools to search for patterns within your code files, to list directories, look at files, do regex. It does not rely on semantically embedding your code in the code base or transforming it into searchable structure. And one of the things that I think has made Claude Code effective is how it agentically can read through your code to take notes in a file called code.md to figure out autonomously what is going on your codebase to then drive decision-making on how to advance your code. That's right. And because of that, and not having a need to index the codebase, you can ensure the codebase stays local. We'll talk about some of the security ramifications with that. So let's get started. And I'll see you in the next video.