Cognition’s Scott Wu: AI Coding Agents Should Enhance Human Efforts, Not Replace Them

Scott Wu, the CEO of Cognition, made headlines this week as his two-year-old AI coding startup secured $1 billion in funding, bringing the company’s valuation to $26 billion. Cognition is behind Devin, one of the first and arguably the most successful AI coding agents in the industry. As Wu states, Devin “naturally owns tasks from start to finish.”

In a blog post announcing the funding, Cognition expressed its vision by stating, “we are progressing toward a future of self-driving software development.”

So, is there a chance that Devin could replace a mid-level L4 programmer? The answer is both affirmative and negative, Wu explained to TechCrunch. “We’ve never seen it as a means to replace humans. I recognize that people speculate about this, but that has never been our stance.”

In this unpredictable year of 2026, where tech CEOs often announce layoffs resulting from AI integration, Wu clarifies his commitment to ensuring coders do not lose their jobs. “We are all programmers ourselves,” he shared. “I started coding at the age of nine.”

Indeed, Wu has been recognized as one of the most accomplished young competitive programmers in history, as highlighted in a recent feature by Colossus. At just eight, Wu won a national math competition designed for seventh-graders, which set him on a path to numerous math and programming contests throughout his youth. This journey also connected him with other prodigies who went on to establish AI tech startups, including Alexandr Wang of Scale AI.

Therefore, he stresses to TechCrunch that the intention has never been to render human programmers obsolete.

“When we began creating Devin, it’s a bit amusing,” he reflected, “but we genuinely envisioned it as your partner who helps you achieve more.” He even displayed a small stuffed toy holding a computer, his own Devin teddy bear, which sits on his desk. He considers it a physical representation of the Devin AI coder: “This is my partner that helps you create more.”

Wu is dedicated to ensuring that AI agents do not take away the joy of programming.

“It’s well known that most software engineers love building software, right?” he stated. “If you ask them why, they’ll typically say, ‘Well, it’s like I get to create things from scratch. I can transform an idea into a product. I can turn it into an experience.’”

Just as visual development environments made software creation more accessible, he views agents as another layer of abstraction that facilitates the transition from conceptualizing a software product to actualizing it.

However, Cognition notes that Devin is responsible for delivering nearly all of the software within the organization. They report that 89% of the code generated by its engineers was produced by Devin, with the remainder sourced from local agents of Windsurf, their AI coding competitor acquired last year.

Wu points out that Devin primarily addresses the long-tail maintenance tasks that many programmers typically find tedious: updating outdated software and migrating applications across platforms. Agents will liberate programmers “from much of the drudgery, allowing them to delve more into the creative aspects,” he assures.

Consequently, Wu resists the idea of Devin “replacing” human coders. While he maintains that it can operate independently, it functions “somewhere between a junior and mid-level engineer” depending on the specific task.

As for the concept of self-driving software, where the agent learns and improves its abilities to eventually perform at higher levels (“recursive” is the latest buzzword in AI), Wu states, “I believe we are headed for an exciting journey.”

He envisions agents branching out into various sectors, acquiring skills from customer service to healthcare, while hoping that the focus remains on enhancing human workers in those fields as well.

“Code and software have been the primary areas to evolve, but we will see this trend extend into many other industries,” he forecasts. “One thing that has been clear to us from the outset is that it should always be the human who determines the course of action … you’ll notice this in software engineering, but I believe it applies to all other professions as well.”