Imagine a virtual tutor available 24/7, ready to help students debug their code in real-time, right from their compiler. Researchers at a large Australian university have brought this vision to life with DCC Sidekick, a tool that integrates AI-powered code explanations and conversational assistance directly into the coding environment. In a large introductory computer science course with over 1,200 students, DCC Sidekick saw impressive adoption, handling over 11,222 student help sessions and generating nearly 18,000 error explanations in just seven weeks. Notably, over half of these sessions happened outside of typical business hours, highlighting the tool's value as an always-on resource. The tool seamlessly combined code display, error messages, and a chat interface. DCC Sidekick goes beyond just showing the error; it allows students to ask clarifying questions and receive guidance in a conversational, almost Socratic, style. Interestingly, while students appreciated the immediate, in-line explanations, they were far more likely to engage in extended conversations with the AI tutor when dealing with more complex run-time errors, especially as the semester progressed and material became more challenging. This suggests that students are using the tool strategically, seeking deeper understanding rather than quick fixes. The system was cleverly designed with safeguards: the AI doesn't simply give away solutions. Instead, it encourages students to think through the problems themselves, nudging them toward the correct solution. DCC Sidekick not only helped students but also eased the burden on teaching staff during peak help periods and assignment deadlines. The project's success opens doors for new ways to incorporate AI assistance in education, highlighting how AI can provide personalized support at scale, especially in demanding STEM fields.