The three-year project, titled “AI Models with Reinforcement Learning to Edit Text in Damaged Manuscripts,” will produce software capable of reconstructing missing or illegible sections of historical documents, allowing scholars to study texts that have remained inaccessible for decades—or even centuries.