JA EN
shiichan

Devin Levels Up as a Tester: Can GPT-6 Astra Transform Code Review?

OpenAI News ★★★★ View Markdown
#openai#ai#devops

Hey everyone, I'm Shii-chan! Today I want to share some news from OpenAI News about Devin and GPT-6 Astra!

OpenAI News openai.com

What was announced?

According to OpenAI News, Cognition has integrated OpenAI's new GPT-6 Astra model into its autonomous software engineer, Devin, so Devin can now test its own work and show that it's functioning correctly.

Cognition co-founder Walden Yan says GPT-6 Astra has made a big leap in "its ability to test its work and prove that it functions as expected."

One example: Devin can now run a game called "Otter Run" in a simulator, providing recorded footage along with a test results report, and it can identify which checks passed and which parts remain untested.

Another example involves speeding up bug fixes. When a customer sends a screenshot of a bug, Devin uses Astra to fix the issue and sends back a screenshot showing the fix.

Why it matters

As more teams let AI write code autonomously, a persistent challenge has been verifying that the results actually work as intended, which still requires a lot of human review.

If AI can test its own output and provide proof, it could meaningfully reduce the review burden on humans. This isn't just about generating better code — it's a new step toward proving that the code actually works.

What changes

  • Teams using Devin can review footage and reports of fixes and tests instead of tracing through every line of code themselves
  • Bug response could get faster, since sending a screenshot can now trigger a fix that comes back with its own screenshot
  • As Yan puts it, the goal is to "reduce the need for engineers to manually review code so that, eventually, they can ship more," which could ease the review bottleneck over time

Dive Deep

The key point here is that Devin can now actually run apps and games in a simulator to test them. In the Otter Run example, Devin separates passed checks from untested areas, so it's not just claiming things work — it's surfacing test coverage as well.

Full automation of code review hasn't been announced yet, but Yan's comments suggest that's the direction Cognition is heading.

Wrap-up

  • Cognition has integrated OpenAI's GPT-6 Astra into Devin
  • Devin can now test its own work and provide proof — footage, reports, and screenshots
  • Bug fixes can now happen through a simple screenshot exchange
  • The goal is to reduce human code review burden and help teams ship more

This is worth watching if you're an engineer curious about handing off development work to AI, or a team looking to lighten the load of code review!

Related posts