“Our AI engineer is set to write 50% of the company’s code this year,” says Cognition CEO

Devin, Cognition’s brainchild, is not just another coding assistant; it’s an AI agent capable of independently planning, coding, debugging, and deploying software applications.
Unlike traditional tools that require constant human oversight, Devin operates with a level of autonomy that allows it to handle complex engineering tasks from start to finish.
Known as the world’s first autonomous AI software engineer, it is equipped with a sandboxed environment including a shell, code editor, and browser.
Devin mirrors the toolkit of human developers, enabling it to execute intricate projects with minimal intervention.
Founded in November 2023 by a trio of competitive programming champions, Scott Wu, Steven Hao, and Walden Yan, and their formidable foundation in algorithm and problem solving has been instrumental in developing an AI capable of navigating the complexities of software engineering.
Scott Wu, who previously co-founded the AI-powered networking platform Lunchclub, brought valuable entrepreneurial experience to the team, while Steven Hao’s background includes significant work at Scale AI, where he was one of the first engineers in the company. Walden Yan’s expertise in AI technologies and competitive programming strengthened from working with the core team of Cursor, another prominent AI-assisted coding tool.
The company’s co-founder and CEO, Wu, describes the inception of Devin as a “hackathon project.”
As presented by the Cognition team, this AI engineer is designed for complete software engineering tasks. It connects with development team tools like Slack and Linear for managing tasks and GitHub for handling pull requests. This year, Devin is projected to author half of its company’s code.
Devin is like that dedicated and proficient team member, capable of both collaborative development and independent task completion for review. By handling tasks and building alongside engineers, it allows them to concentrate on more complex challenges and enables engineering teams to pursue more ambitious objectives.
Launched roughly a year ago as a junior engineer, Devin has significantly advanced and is now utilised in production by numerous companies. Engineer Scott Wu’s fifteen-person team employs Devin to further develop Devin, with each engineer using approximately five instances of Devin to accelerate their coding process.
The CEO highlighted the fact that they expect that it would pull over 50% of the code by the end of the year. Wu went on to explain that when they launched the Devin a year ago, many people did not believe that an AI-agent engineer was possible at all. Scott Wu stressed on the importance of everyone integrating and collaborating with AI agents into their workflows. When Wu made that statement on how 50% of the company’s codes will be written by Devin, he was projecting on Devin’s strong coding abilities as well.
Its rapid evolution from performing at the level of a “high school CS student” to that of a “junior engineer” within a year demonstrates the accelerating capabilities of AI in software development.
From nimble startups with just one or two team members to Fortune 100 giants, all are leveraging Devin to build their initial prototypes and products efficiently. Large public companies, including major banks like Nubank, are integrating Devin into their engineering teams to enhance productivity and streamline complex projects.
Devin functions as a collaborative AI teammate, designed to augment engineering teams’ capabilities. By integrating with tools like Slack, GitHub, and Linear, Devin seamlessly fits into existing workflows, acting as a “junior buddy” that accelerates development processes. Its ability to learn from interactions and accumulate knowledge across team members enables it to bring valuable context to each new session, effectively multiplying the team’s collective expertise.
Cognition’s CEO, Wu, envisions a future where engineers transition from “bricklayers” to “architects,” focusing on strategic design while AI handles implementation. He believes that AI tools like Devin will lead to more engineering jobs, not fewer, as increased efficiency and innovation drive demand for software solutions. Because of that, he would certainly emphasise building skills as an engineer, which helps a lot to understand the abstractions and to be able to peel the layers beneath. The integration of AI tools like Devin into software development workflows is transforming the role of engineers. Rather than rendering software engineers obsolete, these tools augment their capabilities, fostering a collaborative environment where human ingenuity and machine efficiency coexist.
In April 2025, Cognition AI unveiled Devin 2.0, introducing significant enhancements that further solidify its position in the AI development arena. Notably, Devin 2.0 features interactive planning, allowing the AI to proactively research codebases and develop detailed plans, even from vague prompts. Additionally, Devin Search and Devin Wiki empower the AI to understand and document complex codebases, providing developers with insightful answers and comprehensive documentation. The ability to deploy multiple Devin instances simultaneously enhances productivity, and a revised subscription model reduces the monthly fee from $500 to $20, making Devin more accessible to a broader range of users.