CoPilot: Can GitHub’s New Tool Really Automate Coding?
The Stamped development team shares their first impressions of GitHub CoPilot.
Stamped focuses on automation within accounting, but also explores ways to automate other workflow components. GitHub’s CoPilot application caught the team’s attention as a software coding automation tool. The team decided to test CoPilot and share their initial impressions.
Brice Sanchez, Back-End Developer
CoPilot parallels Stamped’s mission: developing AI software to simplify accountants’ work. The tool provides useful code suggestions, but we must keep in mind the overall picture of our project and think about how these suggestions can help us. Human validation remains essential.
Jean-Simon Royer, Junior Software Engineer
As an early beta tester, I appreciate CoPilot’s automation of repetitive code blocks, enabling faster work and focus on complex code sections. Currently, approximately 50% of suggestions prove useful.
Dominic Bilodeau, Back-End Developer
CoPilot excels at solving small coding puzzles. Coding resembles assembling a puzzle requiring the construction of our system’s architecture before assembling pieces. Developers may spend less time implementing small pieces, dedicating more time to designing and architecting features.
Jonathan Rochette, Tech Lead
Remaining open to new tools is important, but there’s a risk of losing the ability to judge code. Code suggestions aren’t automatically correct; developers must maintain a critical mind by taking time to read, understand, and evaluate code appropriateness.