GitHub Copilot is presented as an AI pair programmer, that plugs into the user's editor. It then turns natural language prompts into code, offers multi-line function suggestions, speeds up test generation, filters out common vulnerable coding patterns, and blocks suggestions matching public code.
$10
per month
IBM Bob
Score 8.3 out of 10
N/A
IBM Bob is an AI-assisted development tool designed to support developers across the software development lifecycle. It helps interpret developer intent and provides assistance when working with real-world codebases by using contextual awareness of the surrounding project.
Copilit is fantastic at the following: 1. Solving simple, well-defined problems, such as implementing an algorithm, manipulating a data structure, or string manipulation and regex. 2. Implementing simple APIs that are mainly CRUD in nature, with moderate business logic inside them, which may involve some processing or passing the data through an algorithm. 3. Implementation of well-defined activities, such as implementing a connection to an Oracle DB using Hibernate or JDBC, or implementing boilerplate code for a backend service to listen to Kafka events. It is not that great when it comes to understanding and implementing code in a proprietary DSL. It struggles when implementing a major feature across a complex codebase. I believe developers should also adopt the trust-but-verify paradigm when expecting highly secure or regulated code from GitHub Copilot.
The first time using it feels like magic. Its ability to analyze an existing codebase and adapt accordingly to questions, feature requests, issues, etc. is remarkable. However, it is not limited to an existing codebase - I've had it create new applications based on styling from another application from scratch. It has quickly gone from an amazing novelty to something that is relied upon daily.
I feel that GitHub Copilot's overall usability is good due to its tight integration with Visual Studio and the workspace. However, developers expect greater ease of use, as there is a learning curve to realize productivity gains with the tool fully. I think there is room for improvement in GitHub Copilot's UI integration within Visual Studio.
It works well, for the most part, but asking for authorization for multiple steps or tool installation and use sometimes becomes bothersome. You can grant blanket authorizations but I do prefer to have a tradeoff there so not all is bad, but sometimes I get the impression that we get asked the same question several times during the same workflow.
It is useful that copilot integrates so well with vscode, which is a very common IDE. I used Tabnine for a little while but it was not that intuitive, and did not seem as helpful as GitHub copilot was. I have enjoyed GitHub copilot a lot, especially the ease of hitting the tab key and seeing quick progress in my tasks.
ChatGPT just did not have the breadth and depth of insight for my coding needs. To get an approximation of what I wanted I had to restate my prompt multiple times, and it looked almost like it did not learn from one time to the other.
