There is a verse in the Quran that says of a guarded thing: "Over it is nineteen." It sits in chapter 74, at verse 30. For most of fourteen centuries it has been read past. A few chapters earlier in the structure of the book, twenty-nine of its one hundred and fourteen chapters open with stray combinations of Arabic letters, the Muqatta'at, whose meaning has been argued over since the beginning and never settled. The question this project began with was almost childishly simple: what if the verse and the letters are talking about the same thing, and what if you could just count.
The thread
The thread was picked up in 2017. That was the year Michael Caswell took his shahadah, and the same year he first encountered the old claim that the number nineteen runs through the Quran's structure. The claim had a bad reputation, for good reasons covered elsewhere on this site. But the core question would not let go of him, and he carried it, mostly privately, for nine years. Not as a scholar. As a Vancouver concert photographer with no training in mathematics, no Arabic, and no way, yet, to test what he suspected.
The sprint
What changed was not him. It was the tools. In the spring of 2026 he sat down with Claude, the AI assistant made by Anthropic, and started directing it to do the counting he could not do by hand: tally specific letters across specific chapters, write the code, build the means to check the code. He cannot write Python. He told the machine what the program needed to prove, read back what it found, caught where it was wrong, and refused to let it stop at "close enough."
The insight that the mathematical structure and the linguistic structure lock every letter in place was his, stated in his second message. It took eight days of work to formally prove it.
The result was a single equation. Group the disconnected letters across the twenty-nine marked chapters by their shared initials and you get thirteen totals, every one divisible by nineteen, summing to one number. Count instead every word of those same chapters, a completely different measurement, and you reach the same number again.
The research was finished on April 1, 2026. The proof, the source text, and the script that checks all of it went public on GitHub the next day, April 2, 2026. No institution, no funding, no peer review board. A zip file and an instruction: unzip it, run it, upload the output to any AI and ask it to check the claims. Three steps. No trust required.
Who found it
For its first weeks the work carried no name, on the principle that a claim about a number should not lean on a person. That principle still holds. But the person is worth naming now, because the shape of him is part of what makes the result strange.
Michael Caswell is a concert photographer from Vancouver. He has number dyslexia. He cannot read Arabic. He holds no academic credentials in mathematics or theology. He types with two fingers. He has aphantasia, no visual imagination at all. He had written code before but was never good at it; what he builds now, he builds by directing AI. By every conventional measure he is the last person who should have been able to surface a structural result in the Quran, and that is precisely the point: this did not require privileged access or rare genius. It required one stubborn question, a tool that could count, and the refusal to fudge a single figure. Everything he did, anyone can redo, because the method is in the open and the answer checks itself.
You can see who else is part of the work, as they consent to be named, on the team page.
What it asks of you
Nothing about this asks for belief. It asks for arithmetic. Download the proof, run it, change the inputs, and watch what breaks and what refuses to. Then decide for yourself what it means that a text names a number inside itself and the number turns out to be true. That last step is yours alone, which is where the book leaves it too: there is no compulsion in religion (2:256).
29:69 "Those who strive for Us, We will guide them to Our ways."