CodeRabbit, an AI-powered code review platform, is scaling agentic software development by leveraging OpenAI models. CEO Harjot Gill explains that the company uses reasoning models like o1 and o3 to improve code validation. By adopting Codex, CodeRabbit has shipped 20 times more software in the last six months without adding engineers.
The platform now uses multiple subagents for navigation and planning, enabled by better price-performance from models like GPT-6 Luna and GPT-6 Astra. CodeRabbit Secure uses Daybreak and Codex to find and fix vulnerabilities. The company serves over 20,000 organizations and 150,000 open-source projects, continuously evolving its benchmark as AI agents introduce new types of bugs.
Source: OpenAI Blog · Summarized by HeadlinesBriefing