An Autonomous AI Agent Called ‘Devin’ Plans and Executes Complex Coding Tasks

By
IBL News
-

IBL News | New York

Cognition AI, which builds AI teammates, introduced this week an autonomous AI software engineer called Devin.

This AI agent can independently write entire software projects from scratch based on simple text prompts.

Devin can plan and execute complex coding tasks with hundreds of steps.

The autonomous agent can code while learning, recall relevant context at every step, fix errors, and collaborate with users in real time.

“With our advances in long-term reasoning and planning, Devin can plan and execute complex engineering tasks requiring thousands of decisions,” said the company.

Cognition AI has equipped Devin with common developer tools including the shell, code editor, and browser within a sandboxed compute environment.

This agent can report on its progress in real-time, accept feedback, and work together with the user through design choices as needed.

In the demos shown below, Devin built complete websites and apps in under 10 minutes. It also successfully completed real gigs posted on Upwork by itself.

On a coding benchmark, the AI agent solved 13.86% of real-world GitHub issues end-to-end, crushing the previous SOTA benchmark of 1.96%.

Funded with a $21 million Series A led by Founders Fund, Cognition AI is dedicated to building AI teammates with capabilities far beyond today’s existing AI tools by solving reasoning.
.