Have you ever heard of the infamous puzzle book, “Maze: Solve the World’s Most Challenging Puzzle” by Christopher Manson? Published in 1985, it’s been a subject of debate and fascination for decades. Despite the official solution being revealed years ago, enthusiasts still gather online to discuss and try to solve it.
Recently, I decided to put ChatGPT 5 Pro to the test. I uploaded a high-quality PDF of the book to the chat, without giving it access to external sources. My goal was to see if the AI could solve the puzzle on its own, using only the information provided in the book.
As the AI struggled to find the solution, I started feeding it excerpts from the internet. But here’s the twist – I told it that these excerpts might contain “noise” or changes, just like in a real-world puzzle. I wanted to see if it could still find the right path.
My biggest concern was that the book might be present in ChatGPT’s training data, which could give it an unfair advantage. But despite this, the results were impressive.
You can see how it went down [here](https://chatgpt.com/share/68a01489-c67c-8006-b557-1f6ea0ff9347).
This experiment raises interesting questions about the capabilities and limitations of AI. Can they truly solve complex problems on their own, or do they rely on their training data? And what happens when you add a dash of creativity and “noise” to the mix?
## The Future of AI and Puzzle-Solving
As AI continues to evolve, it’s exciting to think about the possibilities. Could we see AI-assisted puzzle-solving tools that can help humans crack even the toughest challenges? Or will AI become the ultimate puzzle-solver, leaving us humans in the dust?
One thing’s for sure – this experiment has sparked a new wave of curiosity and debate. And who knows? Maybe one day, we’ll have an AI that can solve the world’s most challenging puzzles in a snap.