AWS unveiled CodeGuru at re:Invent on Tuesday, announcing its new machine learning service for automated code review and application performance recommendations.
The service seeks to take the “peer” out of peer code review—or at least augment the process with a machine learning sidekick, which does sound cooler. CodeGuru pulls from Amazon’s code bases—which span hundreds of thousands of projects—along with over 10,000 open-source GitHub projects to train its machine learning models. It then leverages this knowledge to find code issues and comment on pull requests, providing links to relevant documentation and making suggestions for remediation. CodeGuru also receives feedback as it is used to continually improve its code review process—no big surprise in the realm of machine learning.
Image courtesy of AWS
In theory, CodeGuru could significantly boost development productivity by minimizing the human element of code review, while also boosting optimization. The impact CodeGuru might have on developer workflow, and its real-world effectiveness at optimizing “the most expensive lines of code,” remains to be seen. Still, the product seems poised to potentially shake up one of source code management's most tedious elements.