Codex With GPT-5.5 Cuts Code Review Times to Minutes

Ramp engineers are now receiving substantive code review feedback in minutes, thanks to the integration of Codex with GPT-5.5, a shift that has reduced review times from hours. The company’s AI Developer Experience team is leveraging the tool to both accelerate code review and build internal agentic tooling, improving software development velocity and code quality. “Codex code review catches things that I miss and that other engineers miss and that other AI code reviewers definitely miss,” explains Austin Ray, AI DevEx at Ramp. The tool has proven so effective that it has become a part of many code review flows at Ramp, with engineers actively requesting it by name for its thoroughness and ability to reason deeply against the codebase.

Codex with GPT-5.5 Accelerates Code Review at Ramp

This level of adoption signals a significant shift in how Ramp approaches software quality assurance and developer productivity. Austin Ray, who leads AI DevEx at Ramp, emphasizes that Codex surpasses both human and competing AI code reviewers in its ability to identify critical issues. This depth of analysis isn’t merely about finding bugs; it’s about providing a level of thoroughness that most human reviewers don’t have time for, according to Ray. The system’s ability to deeply reason against the codebase allows it to detect subtle errors and potential vulnerabilities that might otherwise slip through the review process. Beyond accelerating code review, Ramp is also leveraging Codex to build internal agentic tooling, such as the On-Call Assistant, designed to alleviate the burden on engineers during on-call rotations. Ray explains that on-call work at Ramp involves significant business logic, domain knowledge, and heavy incidents, requiring substantial mental effort and focus.

Codex’s reasoning capabilities are proving invaluable in developing this assistant, enabling faster build times and increased confidence in shipped improvements. “Codex with GPT-5.5 is incredibly adept at dealing with that complexity in a way that would take me a ton of mental effort, a lot of sleep, and a lot of single-minded focus on the problem to figure out,” Ray added, underscoring the tool’s impact on complex engineering challenges. Ray believes engineers will transition into roles as directing AI tools like Codex rather than writing every line of code themselves.

On-Call Assistant Development Supported by Codex Reasoning

Ramp is actively integrating advanced artificial intelligence into its operational workflows, moving beyond simple code completion to tackle the complexities of on-call incident management. This initiative acknowledges the inherent difficulties of on-call duty, where engineers must simultaneously manage extensive business logic, domain knowledge, and high-pressure incidents requiring sustained concentration. Ray explains that the system is designed to address the numerous challenges inherent in maintaining context during prolonged incident investigations, including concurrency bugs and the delicate balance between internal and external events. The development of On-Call Assistant has accelerated, with Ray expressing increased confidence in each improvement shipped thanks to Codex’s capabilities. This isn’t simply about automating tasks; it’s about augmenting human reasoning. Ramp’s product surface area is large, yet Codex with GPT-5.5 handles it effectively. He believes this shift will redefine the skillset of top engineers, prioritizing direction and oversight of intelligent systems.

Ramp’s AI DevEx Team Builds Trust with Codex Feedback

Ramp, a financial technology company, is significantly accelerating its software development cycles through the integration of Codex with GPT-5.5, an AI-powered code review tool. Rather than simply adopting the technology, Ramp’s AI Developer Experience team has embedded Codex into core workflows, with engineers now actively requesting its input on pull requests. This level of adoption demonstrates a deliberate strategy to enhance both velocity and code quality, moving beyond initial experimentation to a fully integrated system. The impact on review times has been substantial; engineers previously waiting hours for initial feedback now receive substantive analysis in minutes. This speed is attributed to Codex’s ability to deeply analyze the codebase, identifying issues that often elude human reviewers. Ray concludes, “Codex is the real deal; it definitely helps us ship faster.”