A Pragmatic Approach to LLMs (gracefulliberty.com)

In a thought-provoking discourse, a programmer articulates a strong stance against the use of large language models (LLMs) like Claude Code and Codex in software development. Emphasizing the importance of crafting code by hand, the author argues that reliance on LLMs could undermine essential programming skills, leading to unmaintainable codebases and increasing dependency on corporate tools. Key concerns include the loss of autonomy in software creation, potential legal ramifications surrounding LLM-generated works, and the deterioration of programming quality as many open-source projects embrace LLMs, leading to an influx of low-quality contributions. While voicing skepticism towards LLMs, the author acknowledges their limited utility in specific scenarios like vulnerability research and navigating poorly documented codebases. He advocates for ethical LLM use that preserves user control and transparency, highlighting the necessity for open-source models to empower, rather than commodify, programming skills. This perspective strikes at the heart of ongoing debates in the AI/ML community about the balance between automation and skill preservation, urging developers to critically assess their relationship with emerging technologies rather than succumb to dependency.