Tips & Tricks: Get the Most Out of GitHub CoPilot | DevReach 2023

The speaker, Todd England, starts the session with a warm welcome and light-hearted remarks about the audience's possible activities the night before. He introduces the topic of the day: tips and tricks for getting the most out of GitHub Co-Pilot, an AI pair programmer. Todd mentions a past event, Dereach 2013, and engages the audience with a selfie to continue a new DevReach tradition. He then provides a brief introduction about his role at Microsoft as part of the Cloud Advocacy team and encourages the audience to connect with him on social media for further discussions.

Todd explains the underlying technology of GitHub Co-Pilot, which is based on large language models (LLMs). He clarifies that these models work by predicting what comes next based on existing code patterns. The speaker also discusses the limitations of GitHub Co-Pilot, such as its training data constraints, the fact that it is not a compiler, and its inability to read the user's mind, emphasizing the need for context when using the tool.

The paragraph focuses on the importance of context when using GitHub Co-Pilot. Todd outlines several ways to provide context, including keeping relevant files open, using top-level comments to set goals, including in-file references, using meaningful names, writing specific comments, and providing sample code. He emphasizes that good coding practices and clear communication of intent to the AI are crucial for receiving accurate and useful suggestions.

Todd elaborates on the various ways GitHub Co-Pilot can aid developers, such as writing basic code faster, helping to remember coding details, extending and refactoring existing code, explaining unfamiliar code, understanding and fixing errors, and even adding tests to a project. He encourages the audience to explore these capabilities and to stay tuned for future functionalities.

The speaker transitions into a live demonstration using GitHub Co-Pilot within GitHub Code Spaces. He navigates through a simple script file and provides tips on how to use GitHub Co-Pilot effectively. The live demo showcases the process of receiving and accepting suggestions from the AI, verifying the code's correctness, and iterating on the suggestions to achieve the desired outcome.

Todd demonstrates the process of iterating on GitHub Co-Pilot's suggestions to refine the code. He emphasizes the importance of not blindly trusting the AI and instead using developer knowledge to verify the suggestions. The live coding session continues with experimenting with different prompts to generate a function that removes HTML tags from a string using regular expressions.

The speaker shows how GitHub Co-Pilot can be used to document code and add tests. He uses the 'slash' commands in the inline chat to generate documentation for a function and to create test cases for the script. Todd also discusses the process of installing a testing framework and running tests, all within the integrated development environment (IDE).

Todd moves on to demonstrate creating a CSS stylesheet with the help of GitHub Co-Pilot. He uses comments within the CSS file to guide the AI towards creating styles for the header and footer. The speaker also addresses the limitations of generating large amounts of content like Lorem Ipsum through GitHub Co-Pilot and instead uses a more efficient method within the editor.

The paragraph covers how GitHub Co-Pilot can assist in refactoring code and adding new features to an existing project. Todd uses the chat feature to ask for help in making the main content area scrollable and the footer always visible. He also discusses the importance of context when using GitHub Co-Pilot, showing how the tool's suggestions become more accurate when the relevant files are open.

The speaker demonstrates using GitHub Co-Pilot to apply CSS features such as linear gradients and transitions. He uses comments within the style tags to prompt the AI for specific CSS effects and accepts the suggestions after reviewing them. The paragraph also highlights the importance of iterating on the code and using the AI's chat feature for clarification and assistance.

Todd illustrates how to add a new service to an existing project using GitHub Co-Pilot. He emphasizes the impact of context by comparing suggestions made with and without an open file that relates to the project. The speaker shows how the AI can generate code that aligns with the project's conventions when given the proper context.

The paragraph concludes with a demonstration of using GitHub Co-Pilot for refactoring code and debugging. Todd uses the inline chat to request code that will cause a deliberate crash in a console application. After the crash, he shows how to use the AI to understand and resolve the exception that occurred, showcasing the tool's capabilities in assisting with debugging sessions.

In the final paragraph, Todd recaps the key points discussed during the session, including the importance of using single, specific, and short prompts, trusting but verifying the AI's suggestions, and the critical role of context. He also looks to the future, mentioning upcoming features for GitHub Co-Pilot and encouraging the audience to explore the tool's potential through features like GitHub Co-Pilot for pull requests and Co-Pilot Adventures. Todd concludes by reinforcing the idea that AI tools like GitHub Co-Pilot are meant to assist, not replace, developers, and that understanding and effectively using these tools leads to more productive and enjoyable programming.