Cognition helps Devin test its own work with GPT‑6 Astra

Cognition, the company behind the autonomous software engineer Devin, has integrated GPT‑6 Astra to enhance its ability to test and validate software work. The goal is to reduce the volume of code engineers must review manually, allowing teams to ship products more efficiently. Devin is already deployed by businesses globally, including major banks and tech startups, and the addition of Astra aims to address the growing challenge of reviewing increasing amounts of code.

GPT‑6 Astra enables Devin to test its own work and produce clear evidence of the results, such as recordings and reports. For example, Devin tested Otter Run, an iPhone game, by running it in a simulator and generating a report that highlights passed checks and areas needing further testing. These outputs allow engineers to visually inspect the application’s behavior and identify untested components without manually reviewing the entire codebase.

The integration of Astra also accelerates the resolution of customer-reported issues. When a customer submits a bug screenshot, Devin can use Astra to diagnose and fix the problem, then return a screenshot showing the corrected behavior. This process reduces the time required to address customer feedback and streamlines the debugging workflow for Cognition’s engineering teams.

Cognition emphasizes that stronger testing with GPT‑6 Astra supports a more efficient code review process. By automating testing and providing verifiable results, Devin can help engineers evaluate changes with less manual examination, potentially improving productivity and reducing bottlenecks in software development pipelines.