Bottom-up programming as the root of LLM dev skepticism (www.klio.org)

A recent discussion delves into why some developers remain skeptical about large language model (LLM)-driven programming despite its proven success among many users. The author explores various reasons for this skepticism, ultimately suggesting that it stems from the distinction between two coding approaches: top-down and bottom-up programming. Top-down programmers can leverage LLMs effectively because they already have a clear vision of the code structure they need. In contrast, bottom-up programmers, who evolve their designs through iterative writing and discovery, may struggle to utilize LLMs, as they don't know what the structure should be until they write it. This insight is significant for the AI/ML community, particularly as it highlights the need for tailored tools and approaches that accommodate different programming styles. The author argues that with advancements in tools like GPT-5.2, even top-down coders find it easier to generate code, while bottom-up programmers might find LLM-generated code less useful or misaligned with their emergent design processes. This reflection may encourage developers and tool creators to reconsider how LLMs can be better integrated into various workflows, ensuring that all programmers, regardless of their approach, can benefit from AI advancements in coding.