A security researcher built a deliberately vulnerable book review app and spent $1,500 to test if Large Language Models could successfully hack it. The experiment aimed to determine if AI can reproduce common classes of software exploits typically found in real-world applications.