GitHub Copilot and Codium.ai Misconceptions.

We recently had the opportunity to present an interesting topic through Abstracta, “Usando Github Copilot y Codium AI para pruebas”. This is a follow up based on the audience great questions and inquiries.

As my father used to say, ‘The worst question is the one that was never made’. When presenting, one of the things that I like the most is getting questions from the audience! And this time, I want to share some of theme here :)

a. “Que pasa si tengo un proyecto con capas y encapsulamiento, Copilot también nos podría apoyar?”

For me this was one of the best questions, and one of the functionality / feature misconceptions. GitHub Copilot has the ability to understand code and logic outside of the current file if there are dependencies on other classes, packages, etc. However, based on our experience, and the latest information available, GitHub Copilot’s capabilities are limited to the current files and opened tabs in your IDE.

This means it cannot automatically access and read all your project files, and layers just by installing it, this is not how it works.

b. “Con respecto al manejo de datos, recomendarías usar copilot en proyectos con código muy sensible?” And “Cuando uso ChatGPT soy cuidadoso en que código estoy ingresando a la herramienta, pero al instalar un plugin que hago?”

Another great pair of questions! It’s important for us to know that while GitHub Copilot can read files, it does not share your private code with other users. GitHub follows responsible practices to ensure that your code snippets will not be used as suggested code for other users of GitHub Copilot.

As mentioned above, this privacy is maintained by not sharing your code with other users and by providing you with control over the collection, retention, and processing of your data. For more info:

https://docs.github.com/en/copilot/copilot-individual/about-github-copilot-individual

https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement

During the presentation we also mentioned a Codium.ai misconception, Code completion. This was based on misinterpretations and misleading data. As of today, this tool does not offer a code completion feature such as GitHub Copilot does, meaning, it does not generates (or ‘throws’ as we like to say) possible completion cases right after your last line of code, comment, and/or character. For instance:

But lets also note that Codium.ai offer suggestions based on existing code, and you can access these by clicking the Code Suggestions tab. For instance:

For more info: https://www.codium.ai/blog/ai-powered-code-suggestions-for-productive-development/

The fun part of evaluating tools is not picking a winner, but identify their features, functionalities, and where these tools excel. This helps to understand and acknowledge which are their limits. I believe both of these tools offer great features and unique characteristics, and once we learn how to properly use them, we can easily increase our productivity, design, execution, and gain some precious time!

Another thought_

A colleague asked me:‘Why you do this for free? You always talk about time, but you do this for free’, and so I asked him: You only share if you are getting paid?

Massive value comes from sharing our knowledge, perspectives and opinions.