LLMe (blog.fogus.me)

In a recent discussion on the Clojure development team's approach to AI, Alex Miller stated that they do not actively use AI technologies, a sentiment echoed by the author who shares a critical perspective on the limitations of large language models (LLMs). While acknowledging some productivity gains attributed to LLMs, the author argues their over-reliance threatens the essential mentoring culture in programming, where experienced developers share valuable insights with newcomers. This shift from mentorship to automated assistance risks producing a workforce that is less experienced and less capable of critical problem-solving. The author emphasizes that LLMs are fundamentally constrained by their training on digitized information, which not only perpetuates existing biases but also lacks the depth of understanding required for genuine problem-solving. While they initially saw the potential for LLMs to assist in the creative design process, their experience has shown that these models are often frustratingly limited, providing more sycophantic agreement than valuable insights. As a result, the author concludes that relying on LLMs undermines the cultivation of judgment and deep expertise that is vital for advancing complex design work in the software industry.