Software Engineer Says 80% of His Pushed Code Is Now AI-Generated

The role of the software engineer is undergoing a fundamental transformation as generative AI moves from a simple assistant to a primary producer of source code. For some developers, the profession is shifting from the act of writing to the act of architectural oversight.

Andrew Caldwell, a 26-year-old software engineer leading a team of five at a financial services firm, reports that approximately 80% of the code he now pushes is AI-generated. According to Caldwell, this shift has inverted his daily workflow. "Before generative AI hit the scene, I spent more time writing code than planning, reviewing, and testing it," Caldwell told Business Insider. "Now, the opposite is true."

The Rise of the Generalist

This transition is part of a broader industry trend driven by the integration of coding agents and assistants like ChatGPT and Claude into professional workflows. These tools are blurring the traditional boundaries between specialized roles, such as front-end, back-end, and QA engineering.

For Caldwell, this capability has allowed him to expand his technical reach into entirely new domains. By leveraging AI, he ventured into mobile development, enabling his team to develop and submit an application to both Apple and Google within a three-week window—a timeline that would have been significantly more challenging under traditional manual coding constraints.

The Architect's Burden

While productivity has accelerated, the shift introduces systemic risks that make senior-level expertise more critical than ever. The reliance on AI increases the probability of "hallucinations," where the AI suggests non-existent libraries or introduces subtle security vulnerabilities into the codebase.

Consequently, the engineer's primary value is shifting from the ability to write syntax to the ability to audit. Rigorous human oversight is now the primary defense against these flaws, requiring a deep system understanding to ensure that AI-generated efficiency does not come at the cost of stability or security.

Guarding Against Atrophy

Beyond technical bugs, there is a growing concern regarding "skill atrophy" among developers who rely too heavily on automation. As the manual labor of coding diminishes, the fundamental engineering skills required to debug complex issues from first principles may erode.

To combat this, Caldwell maintains a disciplined approach to his craft, intentionally hand-coding for approximately two hours a day. This practice ensures that while AI handles the bulk of the production, the human developer retains the core competency required to oversee the machine.