Using CoPilot to Boost Productivity

Since ChatGPT was announced, I have been using it as a pair programming buddy. I bounce ideas, look for pseudo-code, and discuss drawbacks and potential improvements with it. In my work environment, I have access to GitHub’s Copilot, which assists me in my IntelliJ IDE. Initially, I was skeptical about its performance. However, it has proven to be useful, albeit occasionally interrupting when I’m writing code that isn’t the final version and suggesting things I don’t need. Upon installing the plugin in my IDE with its default configuration, distinguishing between autocomplete suggestions and Copilot’s was challenging. To make this clearer, I had to change Copilot’s suggestions to appear in red color.

I have found Copilot particularly useful in writing tests. By splitting the screen into two windows, I can see the actual implementation while writing my tests. Although it may not fully understand what I’m doing, once I begin writing test descriptions, it can predict and even generate describe blocks and it blocks, saving me time and effort. I would prefer to write all tests myself if Copilot’s suggestions were inaccurate. However, it requires some guidance and does a good job overall. I wouldn’t entirely rely on Copilot’s suggestions for business logic since it lacks full context.

However, when conducting technical interviews where clear requirements are provided to candidates, Copilot can predict almost accurate suggestions in simple cases, given the text copied into the clipboard or split windows.The candidate should evaluate if the code fits their writing style before implementing Copilot’s suggestions. I have seen instances where a suggestion taken without considering other parts of the code broke existing logic, wasting valuable interview time.

Overall, I find Copilot useful and am also exploring ollama.com to incorporate it into my workflow. I look forward to the advancements in software development and the skills we need to upskill ourselves in the coming years.

Photo owned by Github.com