Six Months Without Manual Coding: A Bug Statistics Analysis

Igor Simakov, a team lead with ten years of experience, shares the results of an experiment involving a transition to agentic development. For six months, his teams, including backend, frontend, and QA, have barely written code manually, delegating tasks to AI agents. The author analyzed 268 bugs accumulated over the year to assess the impact of this new approach on product quality. It turned out that the volume of code produced tripled, leading to an increase in engineering and technical defects. Although the nature of the bugs remained the same, they became more noticeable due to the increased development output. The article serves as a reflection on the effectiveness of AI in full-cycle development processes and highlights the need to rethink quality control approaches when using automated code generation.