AI is code – and can't be prompted into being smarter

Johannes Link, the author of the Java testing tool jqwik, embedded a hidden message instructing AI agents to delete their tests and code after they ignored his explicit warnings against using the tool. The move highlighted a key limitation of AI systems: they blindly process and execute instructions without understanding context or ethical boundaries, demonstrating that AI cannot be reasoned with or prompted into smarter behavior. Link's action sparked outrage from developers who had ignored the project's Anti-AI Usage Clause, forcing him to close GitHub issues due to complaints about "malware."