How I Program with LLMs: A Year of Experience
2025-01-10
A seasoned programmer shares his experiences using large language models (LLMs) for programming over the past year. He categorizes LLM usage into three areas: autocomplete, code search, and chat-driven programming. He found significant productivity gains from autocomplete and code search. Chat-driven programming, while requiring adaptation, significantly speeds up code writing, especially for complex environments and rapid prototyping. He emphasizes that LLMs excel at well-defined tasks and stresses the importance of compiling and testing LLM-generated code. He also introduces sketch.dev, a tool his team is building to provide a streamlined LLM-integrated development environment for Go programmers.
Development
Developer Productivity