The Not Hotdog app launched an in-story campaign that had the engineering savvy to back up the very technology it parodied. Utilizing mobile cameras and artificial intelligence, Not Hotdog accurately distinguishes whether objects are “hot dogs” or...“not hot dogs.” The app is designed to get a laugh, but there is serious programming in the backend. Developed using open source technology, including Tensorflow from Google and Keras, Not Hotdog was trained on 150,000 images.

The Not Hotdog app was announced across Silicon Valley’s official social channels. A spoof Bloomberg Technology segment served as the key announcement asset, and all posts directed fans to a download page decked out with in-story copy. After downloading the app, fans went wild, snapping photos of both hot dogs and... things that were not hot dogs, sharing their results. Following the initial wave of posts, many fans put the app’s tech to the test, going to great lengths in their attempts to trick its hot dog-distinguishing capabilities.

Not Hotdog generated a flurry of press coverage around Silicon Valley, garnering a review from TechCrunch, stating, “The engineer behind the viral spoof…probably put more thought into his AI than at least one AI start-up.” The app has over 140,000 downloads and counting since launch and generated thousands of social shares across Twitter, Facebook and Instagram.