AI Now Generates Over 90% of Code at Cognition, Signalling Shift in Software Development

Artificial intelligence has taken on a dominant role in software creation at Cognition, with the company revealing that AI systems now generate more than 90 percent of the code used in its internal development. The disclosure highlights how rapidly automation is reshaping the way engineering teams operate and redefining the role of human programmers.

According to the company, engineers no longer spend most of their time writing code line by line. Instead, AI agents handle the bulk of coding tasks, while human developers focus on setting goals, designing systems, providing context and reviewing outputs. This change marks a fundamental transition from traditional software development to a model where humans act as supervisors and decision-makers rather than primary coders.

Cognition’s leadership described the shift as a natural evolution made possible by recent advances in generative AI. Engineers now guide AI tools with high-level instructions and problem definitions, allowing machines to translate those ideas into functional code. The human role has increasingly moved toward planning, quality control and strategic thinking, ensuring that the generated software aligns with intended outcomes.

The development mirrors a broader trend across the technology sector, where companies are adopting AI-driven coding tools to accelerate product development and improve efficiency. Supporters argue that this approach boosts productivity and allows engineers to concentrate on more complex and creative challenges, while routine coding is handled automatically.

At the same time, the announcement has intensified debate about the future of software engineering jobs. While some see AI as a powerful assistant that enhances human capabilities, others worry about how widespread automation could change skill requirements and hiring practices in the long term.

Cognition’s experience suggests that the industry may be entering a new phase in which writing code is no longer the central task of engineers. Instead, the emphasis is shifting toward problem-solving, system design and effective collaboration between humans and intelligent machines, pointing to a significant transformation in how software is built.