Artificial Intelligence, and later Machine Learning, have historically been benchmarked by scale: “how much” data was processed, “how many” moves can be computed, “how much” predictive accuracy was delivered. But today, our most urgent tests are not about scale, but real-life relevance. In this talk, I’ll introduce the first statewide AI/ML-deduced pedestrian network, built for Washington State, allowing us new opportunities to simulate and measure true access across diverse populations, including travelers with wheelchairs, older adults, families with strollers, etc. When combined with census tract health data, this graph reveals inequities in the built environment that shape health, resilience, and opportunity. I argue that AI’s civic role must be reframed from “how much” (more data, bigger models, higher sensitivity/specificity) to “how well” (do our systems expand capacity, allow for better quality of life and access). This reorientation echoes but also disrupts the legacy of Minsky and Newell: intelligence is no longer an abstract contest of reasoning, but a concrete test of whether AI improves the lived accessibility of everyday life. The real test of AI today should not be whether it scales, but whether it serves. Civic tech is just one proving ground for this shift. If AI divides and isolates us, and at the same time cannot help us cross the street, reach opportunity or connect us to our community, then the scale of its models is irrelevant.