LLMs bring new nature of abstraction (2025) (martinfowler.com)

A recent discussion highlights the transformative impact of Large Language Models (LLMs) on software development, likening their influence to the shift from assembly language to high-level programming languages. While high-level languages increased abstraction and productivity, LLMs revolutionize programming by not only enhancing abstraction but also introducing non-determinism into the coding process. This new dynamic requires programmers to adapt their expectations and embrace a different workflow, as LLMs respond to prompts in unpredictable ways unlike conventional programming languages that yield consistent results. The significance of this transition lies in its potential to redefine how developers interact with technology. As practitioners begin to integrate LLMs into their toolkits, they will need to confront the challenges posed by this unpredictability, which might alter debugging, version control, and overall software quality assurance practices. This unprecedented evolution encourages a reevaluation of programming paradigms, making it essential for the AI/ML community to explore and understand the deeper implications and opportunities presented by LLMs. As practitioners navigate this new landscape, they might uncover innovative methodologies that leverage both the increased abstraction and inherent non-determinism of these generative AI systems.