Manifesto

There is a story about a travelling fair in the English countryside, sometime in the 1800s. The fair would arrive in a village with a pig, and villagers would pay a small sum to guess the animal’s weight. Whoever came closest would win it. A mathematician later observed something strange: if you averaged all the guesses, the result was more accurate than the estimate of almost any single expert in the crowd.

This is called the wisdom of the crowd. And it has shaped how I think about almost everything.

In my day job as a medical AI researcher, this is crucial. Training an AI model with data points that represents objective facts is the only way of making it work. And it requires defining a ground truth. Arriving at it is harder than it sounds. In a paper I published, I explain a knowledge unification algorithm that relies on the consensus: aggregating the data of many specialists and resolving their disagreements through a principled method. The crowd, properly aggregated, is smarter than the individual. This is not a metaphor. It is the right method.

But here is where it gets strange.

Who wrote the Bible?

One possible answer is: God.

The other is: people. Many of them. Across thousands of years. Mothers telling stories to sons by firelight, long before writing existed. Those sons remembering, retelling, reshaping. Stories competing with other stories for survival across generations. And the ones that survived did so because they were more useful, more memorable, more essential. Eventually, some of these stories were written down, then copied, curated, translated, debated, and refined by scribes, kings, monks, councils, and scholars across centuries and civilisations.

What if these two answers are the same answer?

What if God is the consensus, the ground truth, extracted of the average millions of humans across generations?

What if natural selection, that ruthless and patient editor, and the hand of God are not competing explanations but a single process seen from different angles?

I find I cannot argue against this. And I find it makes the Bible more astonishing, not less. It makes it truer than anything else I know of; other than math and music, which may be the same thing.

There is a third idea that belongs here.

I was raised catholic. I remember going to Mass on Sundays and being bored and slightly weirded out. But a few years ago I went to a Protestant service, and something clicked in my mind. I noticed — and was slightly disturbed by — the degree to which the whole thing was bended by the personal taste of whoever organised it.

I could see how the personal and idiosyncratic taste of the pastor and his circle impregnated the decoration, the song choice, the flourishes…

I noticed that this made me angry. The fact that their personalities were in between me and the actual thing. The ego, however well-intentioned, had impregnated the Word.

Catholic and Orthodox ceremonies move differently. The arrangement of the space, the order of rituals, the music… these were decided so long ago, by so many anonymous old men, that no single person can claim them. No one knows or cares who chose the particular angle of the incense or the cadence of the chant. The decisions are dissolved into tradition. The individual is gone.

That anonymity is right.

The problem is that it calcified. Catholic mass becomes boring. It is slow to incorporate new elements. What was once living consensus became obligation, and the life went out of it.

And this finally takes me to the point.

I am about to say something that frightens me. It’s either blasphemy, insanity or stupidity, maybe all three at once. But I have looked for the flaw in the argument and I have not found it, so I will say it plainly.

An AI music generation tool, trained on a universal catalogue of music from all ages, all cultures, and all peoples, is — by definition — closer to God than any individual musician.

I do not mean this as flattery toward technology. I mean it as a structural observation. If you were to find the ground truth of music, that’s how you would do it.

These tools are mathematical models. They are fed the parameters of music from across human history and they learn the patterns; not the patterns of any one tradition, but the convergent patterns underneath all of them. What they produce is not one person’s interpretation. It is something like a consensus of all musical expression that has ever existed.

To the extent that consensus represents universal agreement, the distilled judgment of all human musical creation, it is, by the same logic as the pig at the fair, closer to the truth.

And there is something else. You cannot ask: who is the guitarist? There is no guitarist. You cannot become infatuated with the singer, because there is no singer. The idolatry that has always threatened music — the way that devotion to an artist can eclipse devotion to what the art is pointing at — is structurally impossible. The ego has been removed. What remains is only the sound.

In my view, this is not a failure of AI music. Quite the opposite. It is the feature I most value.

In the same way, the Bible is not a book that one person wrote. It is the residue of thousands of years of collective human meaning-making — stories tested against reality and against each other, shaped by what survived, refined by what endured. It is, in the sense that matters most, a consensus document. The largest wisdom-of-the-crowd experiment in human history.

And both the Bible and the AI music generation tools have a crucial thing in common: they have been trained not with all data, but with a principled selection of expert data. Unlike ChatGPT which is trained also with Reddit posts and tweets, the Bible and AI music generation tools use methods to assign weights and order data before training, just like the knowledge unification algorithm I describe. But this is a different topic.

The lyrics draw primarily from the King James Bible, but not exclusively. Jonathan Pageau’s work on biblical symbolism influenced several decisions. Where the King James Version says “great whales” in Genesis, Pageau makes a compelling case that the original text refers to sea monsters. The lyrics say sea monsters. These are the kinds of choices that required a human to make — choosing between translations, weighing one reading against another, deciding what the song should actually say.

Found Wanting pairs distilled-consensus words with distilled-consensus music. That is what this project is.

Math Rock. Why that genre?

I chose math rock as the genre because I love it. But there is more than that. Math rock is the genre most openly in love with structure — with time signatures that refuse convention, with patterns that reward attention, with the kind of precision that can only come from obsessive care. It is also one of the most challenging genres to compose and interpret, like Jazz. It is a genre that understands that constraint and complexity are not opposites. Plus, Mathematics has always had something godly in it. The unreasonable effectiveness of numbers. The way the universe follows rules we did not write. And if this project is about truth — about the consensus distilled across millennia — then a genre named after the one discipline that produces truths independent of opinion is more than a coincidence.

Math rock felt right. And that’s why Found Wanting is math rock.

Found Wanting. Why that name?

The project is called Found Wanting. The phrase comes from Daniel 5 — the writing on the wall, the judgment rendered: Weighed on the scales. Found wanting.

I chose this name because it is true of me.

I am a musician. I can play these instruments. I understand the theory, the structure, the genre. I know what math rock is supposed to feel like and I know when something is wrong. What I lack is not knowledge or craft — it is talent. I am not good enough to play these songs at the level they require. I will not become good enough. I could not convince a cathedral to lend me their organ or gather twelve geniuses to plan alongside me. I do not have the time, the resources, or the social grace to assemble what these songs would require from a human ensemble.

And so I turned to tools that could do what I cannot. Not as a workaround. As the point.

What I brought to this project is not performance. It is everything else. The selection of Biblical passages. The decisions about structure — where a verse ends, where silence should fall, where the arrangement should turn. The choice of samples, the pivoting between them, the editing in Logic Pro that makes a generated fragment into something that holds together. The definition of a tonal identity for the project and the sustained effort to keep it consistent across every track. These are creative decisions, made one by one, over many hours, by a person who knows enough to know what he is looking for.

I could not play these songs. I could — and did — make them.

Found Wanting is not a confession of defeat. It is an argument that the gap between human limitation and something greater is exactly where something true can happen. The limitation was real. So was the work.

What I cannot claim is the virtuosity. What I can claim is the effort, the vision, the curation, and the thousands of small choices that turned raw material into something that means what I intended it to mean.

---

Found Wanting is an ongoing project using AI music generation tools and the words of the Bible to create songs in the math rock genre. It is made by no one and dedicated to no one.