Self-driving cars are getting pretty good at navigating simple routes, but they still struggle in complex situations. Imagine a broken traffic light with a police officer directing traffic – a human driver easily understands the situation, but a self-driving car might get confused. Or perhaps you prefer to cruise slowly in the right lane while looking for an address. Current autonomous vehicles can't adapt to such nuanced preferences. That's where Autoware.Flex comes in. This innovative system lets you give voice commands to your self-driving car, effectively merging human intuition with autonomous navigation. Autoware.Flex tackles the challenge of translating your natural language instructions into something the car can understand using a large language model (LLM) combined with a specialized knowledge base about driving. This process ensures that your commands are accurately interpreted and safely implemented. But how do you ensure your instructions don't lead to dangerous maneuvers? Autoware.Flex uses a safety validation system that checks your commands against a set of rules before allowing the car to execute them. This safeguards against potentially unsafe instructions. Tests in both simulations and a real-world prototype vehicle show that Autoware.Flex can handle complex scenarios like navigating malfunctioning traffic lights, adjusting the stopping distance from pedestrians, or even choosing specific lanes. While further enhancements are in the works, Autoware.Flex opens exciting new possibilities for the future of autonomous driving, allowing for a more natural and user-centric experience.