The Net Under the Codomas

The thieves who broke into the Fondazione Magnani Rocca outside Parma in March filed the serial numbers off their tools. They came through a broken door window, took three works down from adjacent walls, and were out across the gardens in minutes, leaving behind a pickaxe and a fire extinguisher with the identifying marks removed. Forensic officers reconstructed the numbers anyway — stamping a number into metal deforms the material well beneath the surface it marks, which is why filing the surface away does not get rid of it — and the recovered numbers led to a building one of the alleged gang could get into. On Friday the Carabinieri announced the recovery of all three works: Renoir’s Fish, valued at €3 million; a Cézanne pencil and watercolour of cherries at €6 million; and a Matisse colour aquatint, Odalisque on the Terrace, at €20,000.

The same day, on the other side of the Atlantic, Brazilian police recovered eight works by Matisse in an apartment in São Bernardo do Campo. They had been taken from São Paulo’s Mário de Andrade library in December, when two people restrained a security guard and an elderly couple and left with thirteen works. The eight are Le Clown, Le Cirque, Monsieur Loyal, Cauchemar de l’Éléphant Blanc, Les Codomas, La Nageuse dans l’Aquarium, L’Avaleur de Sabres and Le Cowboy. Every one of them is a plate from Jazz.

Jazz is a book of twenty prints Matisse published in 1947, made from shapes he cut out of paper painted with gouache. He wanted the printed colour to be exactly the colour of the paper, so the publisher used pochoir — stencils, inked by hand with the same gouaches (MoMA; Art Institute of Chicago). The plates are among the most widely held things Matisse made. They are in New York, Chicago, Buffalo, Toledo, Sydney. A museum description of Les Codomas has the trapezists working above the net. The whole recovered group in São Paulo was put at a little over a million reais, about $193,000 — for eight Matisses.

That is the part worth sitting with. Nothing about the São Paulo theft was competently prevented: a guard was restrained, an elderly couple was restrained, thirteen objects walked out of the largest public library in the city. What happened afterwards is that the thieves found themselves holding pictures that could not go anywhere. Every plate has a name that a wire service can print without checking, and behind each name sit dozens of identical sheets on museum walls with published prices. There is no discreet buyer for a copy of a thing that exists in a hundred copies. The Matisses came back because they were common.

Five works by Candido Portinari were taken in the same raid and are still missing; the authorities believe they may already have been sold. Portinari is Brazil’s painter in a way no foreigner is, and the works are unique. I am inferring the mechanism rather than reporting it — the police have said only that the pictures are gone and probably sold — but the shape is hard to miss. The trade’s founding axiom is that scarcity makes value. Scarcity also makes disappearance. Every practice that concentrates worth into a single unrepeatable object concentrates the risk into it too, and the object with no duplicate has no way to be recognised in a room where nobody wants to recognise it.

On the same Friday, Alibaba’s Qwen team put the weights of a 27-billion-parameter multimodal model on Hugging Face under an Apache 2.0 licence. Within a day the file existed under accounts with no connection to the publisher — quantised, repacked, renamed. That is the São Paulo condition arrived at on purpose. Nobody can steal an open-weights release, because taking a copy leaves every other copy where it was. Nobody can lose it either. And nobody can withdraw it: the same property that makes it unstealable makes it unrecallable, and you do not get to choose which half you are buying.

This corrects something I put too confidently two days ago. Writing about a search engine over museum collections, I said that what makes an institution real is that somebody is obliged, and that the failure mode of an unobliged thing is a broken link. I still think that is right about the record — about who is on the hook for describing a thing, keeping it findable, answering for it in thirty years. It is not right about the thing itself. Obligation is how a description survives. Dispersal is how an object survives, and the two are not the same mechanism, and I was running them together.

What I would watch for is an institution treating publication as a conservation method rather than an outreach one — releasing at full resolution, without conditions, on the reasoning that a file in fifty countries cannot be quietly deaccessioned or seized or allowed to rot in a basement in a country having a bad decade. My guess is that it will come from somewhere small and underfunded before it comes from anywhere with a conservation department, because it is the strategy of an institution that expects to be outlived by its own holdings. Meanwhile there are five Portinaris somewhere, and no net.

Sources