Cognition Labs introduces Devin, an agent-based artificial intelligence that plans, codes, observes results, and solves problems autonomously until the task is completed. Devin functions in a closed environment with a personal server shell, code editor, and browser, providing real-time progress updates. Essentially, Devin operates like a solitary developer within a team, with the ability for supervisors to guide workflow.
Within Cognition Labs, Devin showcases its capabilities by creating and deploying games on Netlify, debugging user-submitted code, enhancing AI models, and developing computer vision models for projects sourced from Upwork until successfully completed. Testing with the SWE-Bench suite, Devin achieved an outstanding 13.86% score, surpassing Claude 2 and SWE-LLaMA.
Currently, Cognition Labs is exclusively recruiting testers for Devin, with plans to disclose additional technical information in the future.
TLDR: Cognition Labs introduces Devin, an autonomous AI agent that operates independently in a closed environment, showcasing exceptional performance in various tasks and testing scenarios.
Leave a Comment