LLMs – What Experienced Practitioners See (dr-knz.net)

A recent guide based on interviews with experienced tech professionals explores the nuanced impact of large language models (LLMs) on software engineering. Conducted by a former software engineer and mentor, the guide reveals mixed feelings among practitioners about LLM output, productivity, and when newcomers should adopt these tools. The insights stem from a curated survey of eight industry veterans who collectively bring decades of experience, indicating a complicated landscape where LLMs are beneficial for specific tasks like ideation and refactoring but not necessarily transformative across the board. Significantly, while many respondents report modest time savings and productivity gains with LLMs, this varies by use case. A crucial takeaway is the importance of understanding an LLM’s mechanics; practitioners emphasize that users should grasp the model’s limitations—namely, that it is fundamentally a next-token predictor trained on vast text data. Hence, successful use of LLMs hinges on effective task management and code review—skills that separate those who reap benefits from their use from those who do not. This guide serves as a valuable resource for both newcomers and experienced developers looking to navigate the evolving relationship between AI tools and software development.