The fuzzy-logic Nimbus OS used a decision tree with 47 “mood states,” each tied to specific sensor thresholds. If temperature rose 0.3°C in 90 seconds and barometric pressure fell and the camera saw fidgeting (low-res pixel change rate), the output was “agitation.”

By Alex Rinehart Retro Tech Chronicles

Users grew attached not despite the errors, but because of them. The SASSIE felt like a quirky roommate, not a surveillance tool. FogBank died in 1996 after a class-action lawsuit. It turned out the SASSIE 2000’s “random mood suggestions” weren’t random at all—they were pulled from a hidden 500-line text file of stock phrases written by a single overworked intern named Kevin. Kevin had never studied psychology. He just liked ambient music and horror films.

The SASSIE 2000, by contrast, used flawed, analog, environmental data. It would declare a room “nostalgic” when someone just opened an old book. It once flagged a cat as “mildly contemptuous” (accurate). Another time, it interpreted a nearby subway train as “impending doom” and started playing Gregorian chant.

A FogBank rep named Donna would walk in, sigh loudly, and slump into a chair. The SASSIE’s LED would turn deep red . After three seconds, the monitor would display: “Atmospheric shift detected. Low-pressure front + occupant fatigue. Suggest: Coffee, window ajar (humidity 62%), or Mozart K.448.” Then—and this is the part people swore was fake—the built-in piezoelectric speaker would play 15 seconds of Mozart, but only the minor-key sections . The SASSIE had allegedly “learned” that Donna preferred melancholic over energetic when tired.

Modern AI mood detectors (your phone’s “wellness” features) are boringly correct. They track your typing speed, your heart rate, your search history. They know you’re sad because you searched “why does my back hurt.”