Vibe coding is becoming a key part of software engineering, especially in working with large language models (LLMs). Currently, success in vibe coding is often measured by how quickly engineers can produce results. As companies increase their use of LLMs, the focus will likely shift from just whether something works to how the costs associated with these workflows grow. This means that teams will start to think about the "vibe complexity" of their coding practices, similar to how they evaluate algorithm efficiency with Big-O notation. The tension lies in balancing speed and cost. Teams that avoid vibe coding tend to be slower, missing out on productivity. Conversely, those who overuse it—especially through one-off prompts and retries—can face high token costs without clear insight into their spending. The article illustrates that the effectiveness of vibe coding can vary significantly. One engineer may use many loosely defined prompts, leading to high correction costs, while another might employ a few well-structured prompts that limit token use and drive efficiency. As organizations begin to track vibe coding effectiveness, metrics will likely evolve to include speed relative to token usage. The best vibe coders will be those who not only produce results quickly but do so with minimal token expenditure. However, this focus on efficiency risks overshadowing the exploratory aspects of vibe coding. Many engineers use this approach as a brainstorming tool, which is inherently less efficient but essential for creative problem-solving. If companies prioritize efficiency too heavily, they might stifle the messy, iterative processes that yield valuable insights and innovation.