According to Ethan Mollick on X, GPT ImageGen-2 now reliably renders readable text in images, enabling slide decks, academic-style pages, and structured visual documents from prompts, crossing a quality threshold he did not expect (source: Ethan Mollick, Apr 21, 2026). As reported by Ethan Mollick, the model passes his "otter test"—a stress test for dense, text-heavy compositions—indicating improved layout fidelity and font rendering suitable for professional use cases. According to the tweet, this step-change expands applications to automated marketing collateral, lecture slides, and report mockups, creating workflow opportunities for education, consulting, and content teams that need fast iteration on visual documents. As noted by Ethan Mollick, the capability suggests downstream value for enterprises seeking brand-consistent assets, prototype white papers, and internal training materials directly from prompt-to-image pipelines.