This post discusses the nuanced experiences of staff engineers regarding the use of large language models (LLMs) in software development. Many engineers share their views on LLMs, revealing a split between those who find them invaluable for rapid prototyping or generating boilerplate code, and those who express concern over dependency on such tools, arguing that they foster bad coding practices. Commenters note that while LLMs can assist in finding bugs and generate ideas efficiently, they often lack the capability to understand complex business logic or to write idiomatic code. Some have had positive experiences utilizing LLMs for mundane tasks, such as typing corrections or creating utility scripts, but overly depend on such systems is seen as a pitfall. The consensus hints at a cautious coexistence where LLMs are beneficial but should not replace human intuition and coding craftsmanship.