Building the Beautiful
Why a letter from Rome is the most practical document on AI this year
There is a man in a minor book of the Bible who is told that the walls of his city lie in ruins. This is not news. They have been rubble for a hundred and forty years; everyone knows, and knowing has changed nothing. What changes everything is that this one man cannot put the thing down. He asks his king for leave and goes. He rides around the ruin at night, before saying a word to anyone. Then he does something no prophet had done: he divides the work. Every family rebuilds the stretch of wall beside its own door — priests, goldsmiths, perfume-makers, the ruler of half a district with his daughters. Under mockery, then threats, they keep building, armed. Fifty-two days later the wall is whole.
Hold that picture. I will come back to it at the end of the season, and I will tell you then why it matters more than you think. For now I want only its first lesson: he did not write a lament. He looked at the ruin, and he organised.
The tiredness, named
I suspect you are tired, and I suspect I know of what.
Not of the technology. The technology is extraordinary, and I have spent a good part of my working life explaining how it is built. You are tired of the gap between what is said about it and what is done with it. Every vendor is “human-centred”. Every deck has a slide on “responsible AI”. And meanwhile the systems that now plan, decide and act — the ones the industry calls agents, and for once the jargon is honest, an agent is one who acts — are wired into banks, hospitals, hiring pipelines and public offices by people who will never meet the person on the other side.
And all of this lands in a world that is busy tearing itself down. Wars that nobody can end. Violence as a language. Institutions hollowed from within while still wearing their old names. It would be naïve to think the most powerful technology of the century is being handed to a gentler world than the one we have. It is being handed to this one.
So here is the question I want to put to you, and it is the question of this whole season. Seeing the beautiful is not enough. Admiring the right principles, nodding at the right conferences, is not enough. Somebody has to build the beautiful, against the ugly, on the systems that are already running. With what?
Floor and ceiling
Let me give you a model, because I think in models and I would rather show you mine than hide it.
Regulation is a floor. It is the level below which you may not go. A floor is written by people who must make it enforceable against an unwilling party, which means it has to be specific, bounded and provable in a proceeding. The EU AI Act is a floor. So, in its way, is the NIST framework, and so is ISO 42001. I respect floors. Every one of them is a serious achievement, and in the autumn I will show you exactly where each one sits.
But a floor tells you what not to do. It cannot tell you what a good system is. Nobody has ever been inspired by a floor.
The beautiful is a ceiling: the height of the room. It is not a stricter floor or a tougher grade. It is a statement about how much space there is above where you are standing — what a system could be if the people building it wanted it to be that. A ceiling does not forbid. It describes the top of the possible, and leaves you to decide where in the room you intend to live.
Nothing in law will make you climb. Nothing in this newsletter will either. The only thing a ceiling can do is make the height visible, so that a decision not to climb is a decision and not an accident. That distance — between the floor you stand on and the ceiling above you — is the room in which an organisation decides what kind of organisation it is.
Which brings me to the letter.
Where the letter stands
On 15 May 2026 Pope Leo XIV signed Magnifica Humanitas, his first encyclical and the first ever dedicated to artificial intelligence. It was presented in Rome on 25 May. I will not summarise it here; you should read it yourself, directly and more than once. What I want to do is place it, because a document is understood by what stands behind it and what gets built beside it.
Behind it, first, is a date. One hundred and thirty-five years earlier, in 1891, another pope named Leo — the thirteenth — published Rerum Novarum, “of new things”. The new things of his day were factories, industrial machines, and a working class whose lives were being reorganised by the owners of a power nobody had yet learned to govern. That letter did not stop the industrial revolution and did not try to. It named what was at stake so precisely that a century of labour law could be built on it. The choice of name in 2025 was not an accident, and neither is the choice of subject.
Behind it, second, is a lineage. The Vatican did not discover AI in May. In February 2020 the Pontifical Academy for Life launched the Rome Call for AI Ethics, and the first signatures beside it were Microsoft’s and IBM’s — then the FAO and the Italian government. In January 2023 the Abrahamic religions signed it together; in July 2024, at Hiroshima, the Eastern traditions joined. In January 2025 two dicasteries published Antiqua et Nova, a long note on artificial and human intelligence that already carried much of the letter’s vocabulary. The encyclical arrived, in other words, with six years of work underneath it.
And beside it, a structure. On 3 May 2026 — twelve days before signing the letter — the Pope approved an Inter-dicasterial Commission on Artificial Intelligence, announced on 16 May. Seven institutions, one representative each: the Dicasteries for Integral Human Development, for the Doctrine of the Faith, for Culture and Education, and for Communication; and the three Pontifical Academies — for Life, of Sciences, of Social Sciences. Integral Human Development coordinates the first year; the chair then rotates. Its mandate, in the decree’s own words, is the development of AI and “its potential effects on human beings and on humanity as a whole”. Read that composition again. Theology, culture, communication, the natural and the social sciences, bioethics, and the dicastery whose name is the letter’s own decisive test. It is not a press office. It is the shape of a watch.
So: a letter with a lineage behind it and a structure beside it. That is not what a lament looks like. That is what organising looks like.
What “the ugly” looks like at desk level
I said the beautiful must be built against the ugly, and you may have pictured the wars. Let me bring it closer, because the ugly is rarely dramatic.
A man — call him Andrea — applies for a job. A system reads his CV and returns a number: 6.4. Nobody can tell him what the number measured. It did not measure his worth; it cannot, worth is not in the data. It measured a proxy: how much his CV resembles the CVs that pleased that company in the past. There is nothing scandalous about a proxy, as long as everyone remembers it is one. The scandal begins when the proxy hardens into a verdict — when 6.4 stops meaning “this form resembles past hires less than others” and starts meaning “this man is worth less”. A score becomes a threshold, a threshold becomes a policy, a policy becomes “the system won’t let me”, and somewhere along the way the person disappeared and a resource took his place.
That is the ugly. Not a war — a number that nobody answers for. The letter has a name for that slide, and a paragraph for it, and it is from paragraphs like that one that everything I will show you this autumn descends.
The position
Here is where I stand, so you can argue with me.
A letter cannot stop a technology, and this one does not try to. What it can do — what I believe this one does, better than any governance text I have read — is name what is at stake so precisely that an engineer can be held to it. Not inspired by it. Held to it. Nineteen principles can be read out of it, each anchored to the paragraphs that ground it. Twenty-four criteria can be derived from those principles, each a sentence an engineer can act on and an assessor can check. I know, because I spent the summer doing it, and the result is a book and an instrument.
That is this season. Six issues, one link of the chain each: the letter, the book, the principles, the criteria, the evidence, and the instruments — one that exists today and one still being built. Not to admire the letter. To build from it.
Nehemiah’s first act, before he spoke to anyone, was to ride around the ruin at night and look. If you build or run an agent, you can do the same this week: the Tester at try.neemia.org reads it in ten minutes and tells you, with the evidence beside each answer, what is there and what is not. Look first. Then we build.
I will be on my stretch of wall. Tell me where yours is.
— Maurizio
When the Machine Is Not Enough — a season of six letters on building AI agents that respect the people they touch. The book is Neemia: Guarding Human Dignity in the Age of AI Agents, on Amazon. The Tester reads your agent in ten minutes, free: try.neemia.org.
If you want to take part — as a builder, an assessor, a critic, a pilot participant — reply to this email. I read everything, and I answer.
Maurizio Grassi