Imagine an autonomous vehicle that not only navigates roads but also reasons about its actions, much like a human driver. This is the promise of DualAD, a new AI framework that combines the strengths of rule-based systems and large language models (LLMs). Traditional autonomous driving systems excel at following pre-defined rules, but they struggle with unexpected situations. DualAD addresses this by adding a layer of reasoning on top of existing motion planners. Think of it like a seasoned driver supervising a learner. A rule-based system converts driving scenarios into simple text descriptions. An LLM then analyzes this text, offering driving suggestions. For instance, if a child unexpectedly runs into the street, the LLM could suggest braking harder than the initial plan. This allows the system to react dynamically to unforeseen events. Testing DualAD in a realistic driving simulator showed promising results, with significant improvements compared to traditional models. Surprisingly, even with a less powerful, open-source LLM, DualAD showcased significant improvement in safety and responsiveness. The most substantial gains appeared when paired with simpler rule-based planners, highlighting the value of LLM integration. While still under development, DualAD presents a unique path toward more intelligent, responsive, and adaptable autonomous vehicles. Future versions of DualAD aim to enhance the system by allowing LLMs to steer the vehicle and give it more data to process and analyze. This would give the AI an even greater ability to assess changing conditions and adjust behavior accordingly, truly mimicking the anticipation and real-time decision-making of a human driver.