Cognition’s Scott Wu says AI coding agents shouldn’t replace humans

Cognition, the two-year-old startup behind the AI coding agent Devin, just raised $1 billion at a $26 billion valuation, and the funding round has reignited a question that has followed the company since its founding: is Devin meant to take over programmers' jobs? Chief executive Scott Wu addressed that question directly, telling TechCrunch that replacing human developers has never been the company's goal, even as Cognition's own product messaging leans heavily on automation.

Devin was among the earliest entrants in the crowded field of AI coding agents and remains widely regarded as one of the strongest performers in that category. Unlike a simple code-completion tool that suggests the next line as a developer types, Devin is built to take a task and carry it through from start to finish with minimal hand-holding. Wu has described this end-to-end handling of assignments as a defining trait of the product, distinguishing it from earlier generations of AI programming assistants that required constant supervision.

That framing sits somewhat uneasily alongside language Cognition used when it announced the new funding. In a blog post tied to the raise, the company said the industry is moving toward "a world of self-driving software development," a phrase that evokes full autonomy and, to some observers, sounds a lot like software engineers being phased out of the loop entirely.

Pressed on whether a tool like Devin could realistically stand in for a mid-level engineer — the kind of role companies typically label L4 — Wu gave a deliberately hedged answer. "We've never thought about it as replacing humans," he said, acknowledging that people have floated exactly that scenario but insisting it was never how Cognition envisioned the technology being used.

The tension in Wu's answer reflects a broader debate playing out across the software industry as AI agents grow more capable. Investors have poured billions into coding-agent startups on the premise that autonomous software generation could reshape how technology companies build products, yet the executives running those same companies are often careful, publicly, to frame their tools as collaborators rather than replacements — a distinction that matters both for morale among engineering teams and for how the technology is regulated and adopted inside large organizations.

Skeptics note that the "self-driving" language Cognition itself used is difficult to square with reassurances that no jobs are at stake, and that a $26 billion valuation implies investors are betting on Devin taking on substantially more of the work currently done by paid engineers, not merely assisting them. Whether that bet pays off, and how it affects hiring at software companies, remains an open question as adoption of coding agents spreads.

For now, Cognition's position is that Devin is designed to absorb tasks so engineers can focus elsewhere, rather than to eliminate engineering roles outright. How that plays out in practice — at Cognition's own customers and across the broader industry — will likely depend less on the company's messaging than on how businesses actually choose to deploy the technology once it's in their hands.

Source: TechCrunch