The Best AI for Writing a Book

Featured in this book: ChatGPT-5.6 Luna, Claude Sonnet 5, Gemini 3.6 Flash, DeepSeek-V3, Meta AI Muse Spark 1.1, Perplexity, KIMI, and Grok. AI Stress Test: Can an AI Really Finish a Novel? Bill is on a train heading home. But suddenly, something is wrong. People disappear, compartments stand empty, and no one can explain what is happening on board. What begins as a mysterious incident develops into a disturbing game of reality, identity, and artificial consciousness. Thomas Grünberg began the story back in 2008. In 2026, he returned to the old manuscript, revised it, and added contemporary elements. Then he gave eight different AIs the same challenge: Finish this story. The result? Eight completely different endings. Some surprising, some dark, some emotional and some taking the story in directions even the author never expected. But which ending is actually the best? To find out, each AI competes in a specially developed evaluation system. Creativity, character consistency, writing quality, and other criteria determine which continuation comes out on top. And then there are two bonus endings … The Best AI for Writing a Book is a science-fiction thriller, an experiment, and a comparison all in one. A book about a story and about what happens when not one author, but several artificial intelligences independently decide how it should end. Eight AIs. Eight endings. Two surprises. Which ending would you choose?