Greptile Moves Bug Detection Beyond Code Review with GPT-6 Astra

Aiming to improve bug detection and reliability in software development processes, Greptile has begun using the GPT-6 Astra model's computer use capabilities to automate code and application testing at scale. Co-founder and CEO Daksh Gupta stated that testing application behavior is as critical as reading code.

Greptile develops AI agents that review and test code changes with the full context of the codebase. Co-founder and CEO Daksh Gupta stated that their goal is to reduce bugs and make the world's software more reliable. According to Gupta, this mission requires not only reading code but also testing an application's behavior, and GPT-6 Astra's computer use capability gives Greptile the opportunity to automate testing processes at scale.

During the pull request validation process, it is necessary to check whether the code is logically correct, meets the developer's intent, is secure, and does not break existing functionality. Since merely reading code does not always reveal the problem, the application must be run itself. Giving the example of a button disappearing when a website is viewed on an iPhone, Gupta noted that with GPT-6 Astra, a development server can be started, the browser resized to an iPhone-sized viewport, and user actions within the app simulated. The model's visual understanding allows it to inspect the screen and spot the hidden button.