Code Reviews in the AI Age: Yes, or No?!
I have been an advocate for code reviews for decades.
When I first entered professional software development in 2005, an experienced software engineer reviewed every line I wrote. What might feel cumbersome to some was, in reality, the best thing that could have happened to me.
Getting instant feedback on the code I wrote allowed me to improve quickly. Even though we didn’t know it then, we applied agile methods.
With each iteration, we slightly improved the product. In this case, it was my code.
Code Review Software Tooling
Later in my journey, software development became increasingly structured, and the agile movement shaped how we approached software development as an industry.
Tooling like Jira for task management, CI/CD tooling for deployment, and other tools for assuring code quality became the standard.
For code reviews, I worked for companies that used Crucible and Gerrit. While those tools had helpful features, such as side-by-side code comparison, they (in some cases) could slow down the development process.
Some of it was due to how the tooling worked. Another reason was strict policies that didn’t allow code to be merged before a code review was completed.
I remember that at some point, every commit had to go through two code reviews before being merged. The result was a dramatic reduction in development speed.
It’s the same everywhere. When tooling enters an industry, people overuse it in the beginning until it finds its place and becomes second nature.
My latest experience as a regular employee was on a team performing manual code reviews. While it takes some level of expertise and discipline, I was a friend of those informal code reviews. It reminded me of the time I got started with software development.
Code Reviews in the AI Age
With the transition into the AI age, we get new opportunities in software development. Cursor AI, GitHub Co-Pilot, Visual Studio Code Completion, and countless new tools help writing software.
But what about code reviews?
While the previous generation of code review tools often created a bottleneck blocking the merging of a commit, AI code review tools offer a different, much less intrusive, and more efficient process.
For example, the sponsor of this week’s newsletter, Korbit, uses AI to detect and report issues in pull requests using their trained models.
Korbit reviews pull requests in minutes and catches logic bugs, security vulnerabilities, performance issues, and more! It also learns from completed code reviews and uses that knowledge in future code reviews.
Korbit also provides understandable and actionable explanations for issues found in your code. Rather than pointing fingers, it helps developers understand the issue and improve their skills.
If you don’t understand the explanation or what is wrong with your code, you can use the built-in interactive chat and ask for clarification.
In my eyes, AI code review tools are taking code reviews to the next step. They close the gap between the chaos of manual code reviews and the overhead caused by traditional code review tools.
​Try Korbit today and get a 14-day free trial - no credit card required.
.NET 10 Preview 2
Recently, the .NET 10 Preview 2 was released. I’m currently going through the changes, and I will let you know if there is anything important coming up for us .NET developers next week.
PREMIUM CONTENT
