A software engineer with AI expertise, Devin AI can handle the entire coding process

In IT, a new AI model has caused concern. Twitter because of its incredible capacity to write complex code, detect and automatically fix any compilation errors, and do it in a way that’s identical to that of a human programmer. The AI startup Cognition is developing the model, known as “Devin.”

Cognition has raised $21 million in funding and is supported by prominent figures such as Tony Xu, a co-founder of Doordash, Elad Gil, a former executive at Twitter, and Peter Thiel’s Founders Fund. In addition, even though OpenAI’s well-known Copilot and other AI coding assistants have been around for a while, Devin reportedly increases the threshold by performing end-to-end programming tasks.

In the field of AI-assisted coding, Devin might signal a change if Cognition’s claims prove to be true. This AI is set to function as a stand-alone software engineer, rather than taking a backseat to human workers. Devin uses standard development tools like code editors and web browsers to plan and carry out complex technical jobs inside a secure sandbox, according to the company’s creator and CEO, Scott Wu.

Devin only needs to be instructed by a person using a chat interface. After that, the AI develops a solution on the fly, writes the actual code, solves errors as it goes, tests the system, and provides real-time updates to the user. The programmer can easily message Devin to make any necessary corrections if they find any problems.

Devin’s outstanding range was showcased by Wu in a blog post, where he went from deploying websites and online apps to fine-tuning big language models utilizing GitHub repos.

Its result on the SWE-bench test, which evaluates AI’s capacity to address actual open-source software bugs from GitHub, may be its greatest achievement, though. Devin was able to handle 13.86% of these situations completely on his own, compared to 4.8% for Claude, 3.97% for SWE-Llama, another AI, and 1.74% for GPT-4.

Devin is currently under wraps, but Cognition intends to eventually make it available to a limited number of users. The business seems to think that coding is just the beginning, suggesting that it may use its fundamental advancements in “long-term reasoning and planning” to produce AI professionals for other fields.