The author shares personal experiences using generative models (LLMs) while programming, finding them to be a net-positive on productivity. Automating repetitive steps and building tools like sketch.dev have been beneficial. LLMs help with autocomplete, search, and chat-driven programming, although the last one is challenging. The author emphasizes the importance of setting clear objectives for LLMs to produce quality code and proposes the benefits of smaller code structures. An in-depth example showcases the use of LLMs to write code and tests, highlighting the trade-offs and potential pitfalls of relying on LLM-generated code. The overall exploration demonstrates how LLMs can enhance the programming experience.
https://crawshaw.io/blog/programming-with-llms