Imagine effortlessly spinning up complex cloud infrastructure with just a few lines of code, all thanks to AI. This dream is closer than you might think, but is it truly flawless? Researchers recently explored the potential of Large Language Models (LLMs) like ChatGPT to automatically generate Infrastructure as Code (IaC), specifically focusing on AWS CloudFormation. They designed a feedback loop, where the LLM-generated code is checked for errors using a tool called cfn-lint, and then the error messages are fed back to the LLM for correction. Initially, this feedback loop showed promise, with the LLM successfully correcting a significant number of errors. However, the researchers discovered a plateau effect. After several iterations, the LLM struggled to make further improvements, sometimes even introducing new errors. This intriguing finding raises a crucial question: Can AI fully grasp the nuances and complexities of IaC, or does it require human oversight to bridge the gap? While LLMs excel at generating code based on patterns, they seem to stumble when it comes to understanding the deeper meaning behind the code and how it interacts with the underlying infrastructure. The study suggests that although AI can significantly boost developer productivity in IaC generation, it's not yet a perfect solution. Future research could explore more sophisticated feedback mechanisms and focus on teaching LLMs to reason about the semantics of infrastructure, paving the way for truly autonomous IaC generation.