How to keep enjoying programming in a world of LLMs

In a world where large language models (LLMs) are becoming increasingly prevalent, it's essential to find ways to maintain enjoyment and productivity in programming. This article explores strategies for keeping the creative process alive while leveraging the power of LLMs without becoming overwhelmed or burnt out.

First, it's crucial to acknowledge the ethical concerns surrounding frontier LLMs developed by big tech companies. However, this discussion is not focused on those issues. The author emphasizes their human-written content and encourages readers not to mistake them for AI-generated material.

The author draws an analogy between LLMs and a computer composed of human components, highlighting the gradual degradation of programming roles. They express concern about the decreasing ownership of codebases and the increasing reliance on LLMs. To counteract this, the author suggests keeping a balance between using LLMs and maintaining personal coding efforts.

One way to preserve coding enjoyment is by treating LLMs as bookkeeping tools. Instead of allowing LLMs to generate code, use them to handle mundane tasks like converting conversations into to-do lists, organizing test results, and planning defect fixes. This approach allows you to focus on the more creative aspects of programming while still benefiting from LLM assistance.

When using LLMs for research tasks, it's essential to understand the limitations of their knowledge and avoid blindly accepting their results. Make them ask for clarification when they lack context or become exhausted. By actively participating in the research process, you ensure that you maintain a deep understanding of the domain and can make informed decisions.

The author also cautions against letting LLMs take on crucial decision-making tasks. Instead, use them to gather information and present it in a structured manner, such as with markdown files containing frontmatter for planning items. This way, you can review and analyze the generated content before making any final judgments.

To prevent technical debt and maintain a strong grasp of the domain, the author advises conducting your own research alongside LLM assistance. Actively engage with search engines and other resources to build a comprehensive understanding of the subject matter. By doing so, you can ensure that your agents have a solid foundation of knowledge to work from, reducing the likelihood of errors and misunderstandings.

In summary, to keep enjoying programming with the rise of LLMs, it's crucial to strike a balance between leveraging their capabilities and maintaining personal involvement in the coding process. Treat LLMs as bookkeeping tools for repetitive tasks, actively engage in research, and maintain a deep understanding of the domain. By doing so, you can enjoy the creative aspects of programming while still reaping the productivity benefits of LLMs without becoming overwhelmed or losing your passion for coding.