Cognition AI has introduced Devin, described because the world’s first fully autonomous AI software developer.

Cognition AI was developed by programming experts Scott Wu, Steven Hao and Walden Yan and was backed by $21 million in Series A funding led by Peter Thiel’s Founders Fund.

One of the founders, Scott Wu, described Devin as a “tireless, competent teammate” who was in a position to work with people or complete entire projects independently.

Devin is much more sophisticated than OpenAI and Microsoft’s CoPilot. Instead, it’s akin to an AI agent developing software from natural language prompts right into a finished project somewhat than spitting out individual code segments.

AI insight released a series of video demos Describes Devin’s skills across a wide selection of software development and engineering tasks.

Some of the applications featured by Devin from the Cognition AI demo:

  • Adaptability to recent technologies: After reviewing a blog post, Devin successfully ran ControlNet on Modal to generate images with embedded hidden messages, demonstrating his ability to creatively learn from input and adapt.
  • End-to-end app development: Devin independently developed an interactive website that simulates the sport of life. Devin managed the complete project lifecycle, from integrating features based on user requests to deploying the appliance on Netlify.
  • Autonomous debugging: Another key feature of Devin is its ability to discover and fix bugs in codebases without human intervention. Cognition demonstrated this by showing Devin maintaining and debugging an open source competitive programming book.
  • AI model training: To push the boundaries of AI’s role in software development, Devin independently built and refined a Large Language Model (LLM) using only a link to a research repository on GitHub. Yes, that is an AI model that’s capable of making AI models by itself.

Devin’s skills were rigorously evaluated within the SWE Bench coding benchmark. This difficult test asks agents to unravel real-world GitHub issues in open source repositories.

Devin’s performance was remarkable: he consistently solved 13.86% of problems accurately, an enormous leap from the previous state-of-the-art, which only managed 1.96%.

Speaking to Bloomberg, Wu said: “Teaching AI to develop into a programmer is definitely a really deep algorithmic problem that requires the system to make complex decisions and look just a few steps into the long run to make your mind up which path to take should.”

“It’s almost like this game we’ve all been playing in our heads for years, and now there’s a probability to program it into an AI system.”

This huge improvement demonstrates Devin’s advanced problem-solving skills and potential to extend productivity and efficiency in software development.

Devin’s ability to perform complex software engineering tasks autonomously offers a glimpse right into a future where manual programming is all but extinct.

Jensen Huang, CEO of Nvidia, addressed this on the World Government Summit in Dubai, telling the audience: “It is our job to develop computer technology in order that nobody has to code.” And that the programming language is human, everyone on the earth is now a programmer. This is the miracle of artificial intelligence.”

But AI can be ideal for replacing creative professions, So where are people turning?

It’s about using AI when it’s effective, while encouraging critical pondering and artistic agility – and doubtless a good amount of luck in your chosen profession path.

This article was originally published at