In 1668 a priest walked backwards from a wall and found the hole in his own eye. It is large enough to swallow a face at conversation distance. It has been there your entire life. You have never seen it, because the system that reports what is out there is the same system that would have to report what it cannot.
That is the situation this book is about.
Every bounded agent runs a model of the world smaller than the world, and cannot measure by how much. From that one condition follows a specific and expensive error: converting my model cannot currently generate a route to X into there is no route to X. Usually it is not even a conclusion. The option is never produced, so no judgment is made, and nothing appears in the record.
The Belief Protocol takes that error apart across twenty-one chapters, arguing from documented cases: a laboratory that discarded its cultures two days too early for years, twenty-four radiologists who looked directly at a gorilla and did not see it, a telescope that spent seventeen years failing to identify a microwave oven, an astronomer who was right about the method and wrong about the planet.
And it supplies a brake, because an argument about possibility without one is the argument every exploitative system already sells. Possibility determines what deserves consideration. Evidence, cost, and reversibility determine what deserves action.
This is not a book about believing harder. Chapter Nineteen is written against the book and lists what it has not solved.