I’d like to briefly introduce some use cases of Github Copilot.
Translation
Include the source text in English or other language, and write your text/program in a way that conveys your intention to translate it.
Below is code (text) suggested when I was writing a blog article, quoted English text from Stack Overflow, and started to write a Japanese translation myself.
Batch Scripts
Shell script languages like .bat files require strict attention to handling single quotes and double quotes.
What about cases where you write 2 or 3 lines of code with a certain regularity, and you have to repeat it for another 50 lines? Simple work while fearing simple syntax errors is no longer a programmer’s job.
In addition, when you want to process JSON or objects, if you write the data structure in the file as a comment, it will generate code that matches it, or if you specify the properties of the object to be passed as a function argument, it will generate code that matches it. In this regard, it can be said that when combined with a typed language such as TypeScript, it becomes a quite powerful code generation tool.
As a basic approach (this applies equally to ChatGPT and others), what they have in common is that if you provide some regularity/pattern and add minimal comments as needed to convey it, such tools can skillfully decode that regularity and sense what the prompter is looking for.
AI and Copilot do not require unnecessary additions or redundant (magical, commonly called spell-like) prompts. Or rather, since this could lead to confusion, I believe it is important to convey what you are looking for in the simplest and most abstracted form.