Best Practices

Having many different AI features at hand can be overwhelming.

​

When to use what:

​

Chat

Chat is the best way to have a truly in-depth conversation about anything. Things you want to discuss before implementing should be addressed in chat.

Inline Code Edit & Generation

AI can accurately handle many straightforward tasks on the first try. To optimize your workflow, you should utilize inline code edits and code generation to have the AI instantly implement such tasks. Using the chat, where AI is typically more conversational and where you would need to copy and paste the code output, could slow you down.

​

General Tips

​

Context

Provide the AI with the necessary context but avoid overwhelming it.

If your question can be answered with the context of just one other file, the response is likely to be more accurate if you simply reference that file using the ”@“-Symbol Reference, rather than including the entire codebase context with the @Codebase reference.

​

Prompting

When working with AI, it can be easy to become lazy when formulating your questions. The AI can fill out many gaps in your question by itself, but not everything. Accurate questions will lead to way more accurate answers.

​

Trust

These super-advanced AI language models are relatively new, and it can be challenging to adjust to coding with them.

You might forget that you can simply ask the AI about a problem you’re working on and receive a pretty good answer. Even when you’re dealing with a larger problem that you believe the AI couldn’t solve, it’s a good idea to let it have a try.

On the other hand, you shouldn’t trust AI at all times and accept everything it outputs. Especially when facing significant decisions that will influence the rest of your project or important security questions, it’s crucial to verify the information yourself.