As the popularity and usage of large language models (LLMs) have increased in recent years, developers have been leveraging this technology to improve their efficiency. With the ability to code, LLMs can be used to quickly create prototypes of apps and websites. This functionality has enormous potential to benefit both skilled and novice programmers, who can advance their ideas to the testable stage with more ease than ever before. In this project, we are studying how humans and LLMs can effectively collaborate to create prototypes. We have conducted preliminary studies to understand the obstacles that arise when users work with ChatGPT-4, which include issues on both the model's part and the users' part. We are developing an implementation of a prototyping assistant addressing the issues we have seen in our studies.