Jason discovered a 30% misplacement rate during his community recycling experience in San Francisco. He used his weekends to combine IoT and convolutional neural networks to create a smart recycling bin prototype. Built-in cameras and weight sensors could automatically identify various materials like paper, plastic, and glass; misclassified samples were continuously updated through online learning modules. After multiple rounds of community pilot testing, sorting accuracy remained stable above 95%. Local property management decided to promote deployment, and this project also became an important highlight for Jason's Berkeley application.