What Is Code? (martinfowler.com)

A recent discussion highlights the evolving nature of coding in the age of large language models (LLMs), emphasizing that code serves dual purposes: it instructs machines and represents conceptual models for humans. As LLMs enable rapid code generation from high-level descriptions, the value of coding may shift from mere production to the intricate process of creating meaningful abstractions and vocabulary tailored to specific domain needs. This perspective urges developers to engage deeply with their code, collaborating closely with domain experts to build local vocabularies that reflect both technical and business nuances. The implications for the AI/ML community are significant, particularly regarding how LLMs interpret and generate code. Precise vocabulary enhances model performance, ensuring generated code aligns with intended concepts and reduces the risk of cognitive debt—a phenomenon where rapid code generation outpaces team understanding. As developers leverage LLMs, it becomes crucial to maintain a shared language and conceptual clarity, ensuring that the structures and abstractions within the codebase foster effective communication and collaboration. This evolving landscape highlights the necessity for thoughtful coding practices and deliberate engagement with programming languages as tools for discovery and understanding.