How to keep enjoying programming in a world of LLMs

You keep writing code that you enjoy and remain skilled, while LLMs handle the boring parts of planning and research.

A recent article discusses how programmers can continue to enjoy their work amidst the rise of Large Language Models (LLMs) like Claude. The author, a Haskell programmer, expresses concern about the potential for LLMs to degrade the role of human developers, turning them into mere operators of AI tools rather than creators. They argue that while LLMs can assist with productivity, relying on them to generate all code risks a "codebase wasteland" and the erosion of essential programming skills. The proposed approach involves using LLMs for auxiliary tasks such as planning, research, and organizing test results, but emphasizes that the human programmer should remain the primary coder. This method aims to leverage LLMs for efficiency without sacrificing the creative and enjoyable aspects of software development, ensuring that developers retain control over their codebases and continue to hone their skills.

The core argument reframes LLM usage from a code-generation crutch to an intelligent assistant for the programmer. By offloading tedious tasks like initial research, planning, and organizing notes, developers can reclaim the joy of creative coding. This strategy aims to foster a symbiotic relationship where LLMs augment human capabilities rather than replace them, leading to more fulfilling work and sustained skill development.

This approach addresses the ethical and practical concerns surrounding LLM adoption in programming by empowering developers to maintain agency. It suggests that the future of programming involves a partnership, where humans guide the overall vision and intricate design, while AI handles the more routine and time-consuming aspects. Such a model could mitigate the risk of skill atrophy and the creation of unmaintainable code, ensuring that programming remains a craft driven by human ingenuity.