Code reviews – peer reviews of code that help developers improve code quality – are time-consuming. After According to 1 source, 50% of firms spend two to 5 hours per week on it. Without enough staff, code reviews may be overwhelming and distract developers from other necessary tasks.
Harjot Gill believes that code reviews may be largely automated with the assistance of artificial intelligence. He is co-founder and CEO of CodeRabbitthat analyzes code using AI models to supply feedback.
Before founding CodeRabbit, Gill was chief technology officer at data center software company Nutanix. He joined the corporate when Nutanix acquired his startup Netsil in March 2018. CodeRabbit's other founder, Gur Singh, previously led development teams at white-label healthcare payments platform Alegeus.
According to Gill, CodeRabbit's platform automates code reviews using “advanced artificial reasoning” to know the “intent” behind the code and supply “actionable,” “human” feedback to developers.
“Traditional static evaluation tools and linters are rule-based and sometimes produce high false positive rates, while peer reviews are time-consuming and subjective,” Gill told TechCrunch. “CodeRabbit, alternatively, is an AI-first platform.”
These are daring claims with a variety of buzzwords. Unfortunately for CodeRabbit, anecdotal evidence suggests that AI-powered code reviews are likely to perform worse compared to those who involve humans.
In a blog post, Greg Foster of Graphite Conversations about internal experiments applying OpenAI's GPT-4 to code reviews. While the model caught some useful things – like minor logical errors and spelling mistakes – it produced a variety of false positives. Even attempts at fine-tuning didn’t drastically reduce these, in keeping with Foster.
These aren’t revelations. A recent Stanford study found that engineers who use code-generating systems usually tend to construct security holes into the apps they develop. Copyright is an ongoing Worriesin addition to.
Using AI for code reviews also comes with logistical drawbacks. As Foster notes, more traditional code reviews force engineers to learn through meetings and conversations with their fellow developers. Outsourcing reviews puts that knowledge sharing in danger.
Gill sees it in another way. “CodeRabbit's AI-first approach improves code quality and significantly reduces manual effort within the code review process,” he said.
Some people consider the sales pitch. Gill says there are currently around 600 organizations paying for CodeRabbit's services, and CodeRabbit is running pilot projects with “several” Fortune 500 firms.
There's also investment: CodeRabbit today announced a $16 million Series A funding round led by CRV and supported by Flex Capital and Engineering Capital, bringing the corporate's total raised to almost $20 million. The latest money shall be used to expand CodeRabbit's 10-person sales and marketing function and product offering, with a give attention to improving its vulnerability evaluation capabilities.
“We will spend money on deeper integrations with platforms like Jira and Slack, in addition to AI-driven analytics and reporting tools,” Gill said, adding that Bay Area-based CodeRabbbit is within the strategy of organising a brand new office in Bangalore as it’s going to roughly double the dimensions of the team. “The platform may even introduce advanced AI automation for dependency management, code refactoring, unit test generation, and documentation generation.”
