LLMs in Professional Software Engineering (knorpelsenf.me)

In a recent talk at the Waterkant Festival 2026, a software engineer shared insights on integrating Large Language Models (LLMs) into professional software engineering, specifically addressing the challenges of solving high-performance computational problems like the Vehicle Routing Problem. The engineer recounted a systematic approach taken in 2026 to leverage LLMs in creating a distance matrix service, critical for calculating routes efficiently. Employing LLMs drastically reduced the amount of manual coding needed; while only a few lines were crafted directly, around 15,000 lines were generated with LLM assistance, showcasing the model's potential for automating grunt work in coding. The significance of this case lies in the paradigm shift it presents for AI in software development. Instead of viewing LLMs merely as coding tools, the engineer emphasizes their role as "semantic translation machines," capable of transforming concepts and ideas across representations. This perspective helps clarify when and how LLMs can be beneficial, highlighting their strength in handling iterative, detailed tasks while still requiring human insight for more abstract problem-solving. The successful implementation illustrates LLMs' value in enhancing productivity, enabling engineers to focus on complex challenges rather than getting bogged down by routine coding tasks.