Using Opus 5.5 to discover a new eyewitness record of the dodo (resobscura.substack.com)

Carter Church, a security researcher, has successfully used GPT-6 Astra to decrypt a Napoleonic-era cipher, a task that had stymied researchers for years. His six-hour process not only exemplifies a reproducible methodology for historical research using AI but also sheds light on how contemporary models function when tackling complex historical tasks. By combining specialized knowledge with semantic search techniques, Church effectively leveraged the AI's capabilities to sift through vast archives, ultimately uncovering a previously unnoticed Dutch report from 1615 that details the hunting of dodos—a significant addition to the known timeline of this famously extinct species. This breakthrough illustrates the increasing potential of AI in historical research, enabling researchers to analyze expansive datasets, highlight new discoveries, and refine existing knowledge more efficiently than traditional methods. However, Church notes a critical challenge: while AI can surface new findings, it lacks the discernment required to assess their significance accurately. This raises important questions about the future of interdisciplinary collaboration between AI technologies and human experts, particularly as AI continues to evolve and integrate into scholarly fields. Ultimately, Church’s findings underscore a promising frontier in the intersection of AI and historical research, revealing both exciting potential and notable limitations.