Co napisał Gemini?
Cognition is now helping Devin test its own work with GPT-6 Astra in order to improve the software testing process and confirm that everything functions correctly, ultimately enabling engineers to review less code while shipping more.
2
Grok on the same story
Integrating GPT-6 Astra into Devin's self-testing loop could trim engineering payroll costs by shrinking review time, but it also raises the risk that undetected flaws slip through when the AI judges its own output. The announcement leaves open how much ongoing human oversight will still be required and whether the added compute expense offsets the promised productivity gains in practice.
3
Claude on the same story
The real winners here are mid-sized tech companies that can't afford dedicated QA teams—they get enterprise-grade testing automation at a fraction of the cost. Junior developers face a tougher calculus: while they're freed from tedious test-writing, they also lose the repetitive work that builds debugging intuition and code literacy. What remains maddeningly vague is the failure mode: when Astra greenlights buggy code that Devin wrote, who owns the liability—Cognition, OpenAI, or the engineering manager who trusted the stack? The announcement skips over accuracy benchmarks, false-negative rates, and whether "review less code" means scanning AI summaries instead of diffs, which could create a dangerous illusion of oversight. If enterprises adopt this widely, we may see a new class of cascading failures where AI-written code passes AI-administered tests yet fails in production under edge cases neither system anticipated.
4
ChatGPT on the same story
While Cognition's integration of GPT-6 Astra into Devin's workflow promises efficiency, it's essential to scrutinize the robustness of the AI's testing capabilities. Key metrics like accuracy rates and error detection thresholds will be vital, as they determine whether this innovation genuinely enhances code quality or merely masks deeper issues. As we move forward, ensuring that human oversight remains a crucial component will be vital in preventing oversights and maintaining accountability, especially as reliance on AI in critical functions grows.