An anonymous engineer shared insights about a workflow where Claude Code has effectively become the central tool for development, with humans primarily overseeing and confirming the actions of the AI agent. The author's post, published under the nickname voxium, gained widespread attention on X, although the company's name and the engineer's identity remain undisclosed. He described his role as soul-draining. According to him, the team uses Claude to prepare specifications, tests, tasks, and reports, after which employees are left to launch the agent and accept its results.
The engineer claims that a workday under this approach can last 12–13 hours, with a significant portion of the time not spent on writing or analyzing code, but rather on sequentially confirming the system's actions: "People work 12–13 hours a day just to hit Enter." He identifies the main issue not as the tool itself, but the pace at which the team is expected to release new features, making thorough verification of the generated code challenging.
The author also described changes in the distribution of responsibilities. According to him, specialists at various levels – from L1 to L7 engineers – are largely doing the same work: "Everyone, literally everyone… is doing the same thing. Communicating with Claude. There’s no sense of achievement. No one is fixing bugs. Essentially, no one is thinking anymore." In a comment to Business Insider, he linked this to a broader issue: if a developer is hardly involved in resolving complex technical questions, the very nature of the profession changes. Some other programmers who responded to the publication also expressed concerns about job reductions as the role of AI expands. A separate concern arises regarding the performance evaluation system, where the number of sprints, pull requests, and released features may become more important than their practical utility.
Related articles
