Reading the State (2024-ongoing) asks viewers to decide whether a headline was published or generated. The work is not asking which headlines are “true,” as a published headline can still omit context, use biased language, or shape how an event is understood. Instead, the project asks why certain headlines feel believable in the first place.

The generated headlines are made with a Markov chain, a simple text-generation system that predicts each next word based on the word or phrase before it. Drawing from real headlines as source material, the program recombines patterns of political and journalistic language into new headlines that may sound plausible.