Bolt.new and similar tools like Lovable, V0, and Replit offer the intriguing capability of using AI developers to create software based on user descriptions, simplifying the process of building applications. While these tools enable rapid development of functional prototypes, they often fall short when attempting to make more complex changes or refine the software, as the AI may struggle or introduce bugs. This limitation is partly due to the lack of runtime context that human developers naturally use to diagnose and fix issues. Recent developments aim to overcome this by integrating runtime data capture, such as network requests and user interactions, using tools like Replay to enhance the AI's capabilities. A modified version of Bolt has been developed to incorporate this data and uses OpenHands, a flexible AI, to work efficiently on larger codebases, potentially overcoming the 70% completion barrier previously encountered. This advancement promises to make software development more accessible and efficient, with ongoing efforts to refine these tools further.