Cognition, the company behind the autonomous software engineer Devin, is using GPT-6 Astra to make code review more efficient by having the model test its own work and present evidence of the results. As Cognition‘s engineering teams write more code, reviewing it manually has become a bottleneck. Co-founder Walden Yan highlights that one of the biggest improvements Astra brings is its ability to test and prove that its work actually functions as expected.

Cognition is applying GPT-6 Astra across its product lineup, including the core cloud agent Devin, as well as its CLI and desktop products. In one example, Devin uses Astra to test Otter Run, an iPhone game, and then returns a recording of the game running in a simulator along with a report of which checks passed and which areas remain untested. The recording shows the application’s behavior, while the report documents the scope of the testing, allowing engineers to inspect how the software works and see what still needs attention.

Astra also helps Cognition respond to customers more quickly. When a customer sends a screenshot of a bug, the team can pass it to Devin using Astra, which fixes the issue and returns a screenshot showing the result. This shortens the feedback loop and reduces time spent on back-and-forth.

Looking ahead, Cognition sees stronger testing with GPT-6 Astra as a path toward a more efficient review process. Having Devin test its work and provide evidence of the results could help engineers evaluate changes with less manual examination of the code. Yan says the team expects to manually look at less code over time and ultimately ship more, which he describes as a key source of excitement about GPT-6.