Like, love, sad, angry. Around 6000 short messages are written every second on Twitter (now X). This mass of data produced is not only interesting in terms of its content, but also because of its inherent mood. Stream of emotions is the result of an investigation into the extent to which generated data can be analyzed and communicated creatively in real time. The script uses a software library for analyzing sentiment, which is based on the AFINN lexicon resulting from a research project. Each word entry is assigned a value that is either positive, negative, or neutral. This allows the mood of a short text to be determined with a relatively high degree of accuracy. The core of the work is a binary stream of positive and negative keywords that compete with each other and form spontaneous pairs of opposites. But when exposed and removed from their context, they also lose their meaning. Only at second glance do the complete tweets running in the background allow them to be categorized in a context of meaning. The Stream of emotions thus also raises the question of how well algorithms can understand us.