AI as a development collaborator

1. AI as a development collaborator

Hi there! Welcome to a new video.

2. AI as a team member

Now it's time to move to something critical in real-world systems: collaboration. Because no matter how good your architecture is, software is built by teams. And today, AI becomes part of that team.

3. Collaboration challenges at scale

We are still at Wayfarer Labs and have finally deployed Atlas to production. Atlas is now used by multiple engineers: data engineers, analysts, and backend developers. As more people contribute, new challenges appear: code becomes harder to review, bugs become harder to trace, and knowledge becomes harder to share. This is where AI can act as a real-time collaborative partner, assisting us in this process.

4. AI-assisted code reviews

Let's start with code reviews. Reviews are not just about correctness. We also care about readability, maintainability, and design consistency. While we verify the correctness of a colleague's implementation, we can ask AI to act as a senior reviewer and focus on these additional aspects. On-screen, there's a prompt that helps with code reviews. Even though we explicitly asked the model not to change functionality, we should always double-check that this was followed before merging any AI-assisted changes.

5. AI in the IDE

Now let's move to development itself. Instead of using AI only after writing code, we can use it during development, like a pair programming partner. When using AI integrated into the IDE, we don't just have a chat interface. We also have built-in features that assist us while we write code. We will use GitHub Copilot in the following examples, but other IDEs offer similar features.

6. The explain feature

The "Explain" option allows users to highlight a piece of code and send it directly to the model for explanation. This is particularly useful for understanding code written by others.

7. IDE AI features

There is also an option to open an inline chat, which is useful for requesting changes to a highlighted piece of code. In this example, the model adds a conditional statement that acts as a safeguard in the loop, ensuring that only edges connecting valid hubs are processed.

8. Review and fix features

Another useful feature is the ability to request a code review. Here, we can select a function, right-click, and choose Generate Code, Review. The model then analyzes the code and generates comments with suggestions for improvement. There is also an option to fix a given code snippet.

9. Iterative collaboration

The key here is interaction. We don't just ask once. We iterate, refine, and challenge the model's suggestions. This creates a feedback loop similar to working with another engineer. If you want to learn more about the specifics of working with IDEs and their available features, DataCamp has courses on GitHub Copilot, Cursor, and more!

10. AI-assisted documentation

Finally, let's talk about something teams often neglect: documentation. As Atlas evolves, new modules, configs, and workflows are added. If documentation is not updated, knowledge becomes siloed. AI can help us generate and maintain documentation automatically.

11. Prompting for documentation

As with code assistance, we can use prompts in the chat interface to generate documentation. Even better, we can integrate this prompting step into CI pipelines so that documentation is automatically updated when code changes.

12. Generating docstrings

Additionally, many IDEs provide built-in options for documentation generation. For example, we can use AI to generate the docstring of the kmeans function in clustering.py by selecting the function and doing right-click, Generate code, Generate docs.

13. Balancing human and AI

At this point, it might be tempting to rely heavily on AI. But effective teams don't replace developers. They design workflows that balance human creativity and judgment with AI speed and pattern recognition.

14. Let's practice!

Now, let's jump into practice!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.