‘We don’t want to be a policeman of the internet’: How a student of John McPhee turned GPTZero into a $30 million AI

Edward Tian built GPTZero during his senior year at Princeton and hosted it on a free Streamlit page, which experienced high traffic. The platform later reached over 20 million registered users and generated $30 million in annual recurring revenue, leading to an acquisition by Superhuman (formerly Grammarly). Tian and cofounder Alex Cui now lead the authenticity team at Superhuman. The founders have pursued journalism and investigative work, including with Bellingcat and the BBC Eye Investigations team. Their development of detection technology was influenced by their relationship with John McPhee, who emphasized the importance of the writing process over just the final product.

Edward Tian built GPTZero during his senior year at Princeton. The site crashed when 30,000 people flocked to it. GPTZero has more than 20 million registered users and $30 million in annual recurring revenue. Superhuman acquired GPTZero in June. Tian and Alex Cui lead the authenticity team at Superhuman. Edward Tian worked with investigators at Bellingcat and the BBC Eye Investigations team. GPTZero investigated over 100 hallucinated citations in NeurIPS papers. GPTZero scanned 300 ICLR submissions and found over 50 hallucinated citations. GPTZero analyzed errors published by Deloitte, KPMG, EY, and PwC.

The narrative illustrates a pivot from a desired career path in journalism to building technology related to content authenticity. The tension exists between the practical application of technology—creating a detection tool—and the philosophical grounding derived from journalistic principles emphasizing process over product quality. Cui’s prior work on misinformation and his focus on launching products rapidly suggest a pattern where foundational, disruptive ideas often precede stable commercial success, sometimes at the expense of the original vision. The trajectory shifts from building a specific investigative tool to seeking broader "AI curiosity," indicating a tension between addressing immediate technical problems and pursuing deeper epistemological questions about authorship and critical thinking. The outcome—the acquisition and subsequent focus on broad authenticity rather than pure detection—suggests that value is increasingly placed on framing, context, and the perception of truth over singular factual accuracy in an increasingly automated information landscape.