US-based startup Cognition has introduced Devin, the world’s first fully autonomous AI software engineer on March 17, 2024. Devin harnesses AI power capable of resolving engineering tasks independently with its built-in shell, code editor, and web browser.
One of the key features of Devin is its proficiency in fixing bugs on GitHub autonomously. Cognition has demonstrated how Devin leverages its web browser to access and learn from API documentation and conveniently plug into various APIs. When confronted with an error, Devin adds a debugging print statement to the main code within its code editor interface and reruns the code independently, showcasing a comprehensive debugging process.
Cognition has tested and proven Devin’s abilities in the field of AI software engineering. These include creating and deploying apps, identifying and fixing bugs in codebases, and even enhancing AI models to achieve optimization. Devin’s proficiency and skills were measured on SWE-bench, a benchmarking platform designed to challenge AI agents to solve real-world issues found in open-source projects on GitHub.
Devin exhibited impressive performance on the SWE-bench platform, effectively solving 13.86% of the issues end-to-end without any external help. This milestone outperforms GPT4’s achievement of 1.74% and the previous top score of 4.80% held by Anthropic’s Claude 2.
Distinguishing its autonomous capability, Devin can complete codes end-to-end independently. This feature contrasts with AI-powered developer tools like GitHub Copilot offered by Microsoft, which requires human interference or assistance to finish codes.
Currently, Cognition is offering early access to Devin for businesses interested in utilizing the AI agent for their engineering projects. Devin’s successful performance and its capability to operate independently symbolize a remarkable advancement in AI-powered software engineering solutions.
Cognition has demonstrated Devin’s skills via practical engineering interviews from leading AI companies and real jobs on Upwork, further validating its technological prowess towards the future of AI-powered software engineering. With Devin, Cognition has opened avenues to a potential new era of autonomous AI software engineering, introducing new possibilities and challenges in the technological world.
