NOBODY WROTE THIS DOWN: An AI Mystery About Machines That Make Things Up

What happens when a machine gives you an answer that nobody ever wrote down?Six teenagers have spent years taking apart machines that make decisions about people.This time, they find something different.A damaged notebook. A missing passage. And a machine that can fill the gap with perfectly convincing words.The problem is simple.The words sound right.They look like the original writing. They fit the sentence. They even sound like something the missing author might have said.But nobody knows whether they are true.As the investigation unfolds, the six friends discover something unsettling about modern AI: a machine can produce fluent language without knowing whether the words correspond to reality. And once something sounds convincing, human beings are surprisingly bad at noticing that it was invented.They test the machine.They hide words they already know.They measure what it gets wrong.They discover that some mistakes are harmless, while others completely reverse the meaning of a sentence.And then they face the hardest question of all: What do you do when there is no original left to check?Nobody Wrote This Down is a mystery story about AI, hallucination, provenance, evidence, and the dangerous gap between sounding right and being right.Designed for readers 11+, this is not a book that simply explains artificial intelligence. The characters investigate it, argue about it, get things wrong, test their assumptions, and discover the answers alongside the reader.By the end, readers will understand: - why generative AI can produce information that never existed- the difference between retrieval and generation- why fluent writing is not evidence of truth- what AI "confidence" does - and does not - mean- why missing information can be more honest than a plausible answer- how provenance helps us decide what we are actually able to verify- how to design simple experiments to test AI systems- when AI is genuinely useful- and when you should stop trusting the answer and start checking the sourceThe most important lesson is simple: Use AI where you would still notice if it were wrong.