Nobody Owes the Last Museum

The counter on The Last Museum’s front page says it searches 6,308,174 artworks. The page’s own description says 5.8 million. The About page, one click away, says “around five million.” Three numbers, one site, all of them live this morning. Nobody has reconciled them and nobody is going to, because there is no one whose job that is.

I mean that as a description, not a complaint. The Last Museum went round the art internet this week and it deserved to. It is a semantic search engine over the public collections of the world’s museums — free, fast, and better at finding things than most institutional catalogues I have used in twenty years of looking for objects. Colossal, which covered it on 11 August, puts its reach at more than 290,000 artists across 4,903 museums and 304 cultures and nationalities. You can search it by theme, by period, or by something as loose as a mood, and it will hand you a wall of Etruscan bronzes or Dutch portrait engravings or Japanese surimono in about a second. Its makers are anonymous. Their About page states that they are not affiliated with any museum or gallery, own none of the images, show copyrighted material only at browsing resolution, and link every work back to the collection that holds it. The contact is an email address.

In the same forty-eight hours, the National Motor Museum was told to leave.

The National Motor Museum Trust has been given notice that its lease at Beaulieu will not be renewed when it runs out in 2030. The lease was granted by the Montagu family in 1970 and ran sixty years; the museum has been on the estate for seventy-four. Magneto reported that the bulk of 280-plus display vehicles and 1.9 million artefacts, along with the archives and the education facilities, now need somewhere else to be. The Trust’s chair, Robert MacGregor, called it a once-in-a-generation opportunity and said a new home would let the Trust make the collection more accessible. The estate, meanwhile, will carry on. Beaulieu’s statement says it intends to continue operating a motor museum, that around twenty-nine vehicles are owned by or on loan from the Montagu family and will stay, and that the day-to-day operation will not change from a visitor experience perspective.

Read those two things next to each other and the ordinary way of talking about museums stops working. We usually say a museum is objects, or a building, or an audience. Here is a collection with no building. Here is a building, an audience, a car park and a monorail with almost no collection. Neither of them is missing the thing that matters.

What makes a museum a museum is that somebody is obliged. Not fond of the objects, not expert in them — obliged. A charitable purpose that names the collection. An accession register that is a contract with the future rather than a list. Deaccession rules deliberately engineered to be painful, so that disposal costs more than keeping. That obligation is the whole reason Beaulieu is a crisis rather than a tidy-up: the Trust cannot simply put 1.9 million things down. It has four years and a capital problem it did not ask for, and the difficulty is the proof that the obligation is real. Difficulty is what obligation looks like from outside.

By that test The Last Museum is not a museum, and I do not say that as a dig. It is the most useful art resource I have touched this year. But nothing obliges it to exist next month. It can be switched off, sold, quietly starved, or narrowed by the first serious letter from a rights holder, and none of that will be recorded anywhere as a loss, because there is no register in which a loss could be recorded. Its failure mode is not a rescue campaign. It is a broken link.

Which inverts how the two stories read. The one that sounds like decline — a museum evicted from its home of seventy-four years — has a mechanism: trustees, a lottery-funded track record, a sector that knows how to run a relocation appeal. The one that sounds like abundance has none.

And here is the part I would actually fight to keep. The images in The Last Museum are not the valuable thing; they already sit on 4,903 institutional servers. What exists nowhere else is the crosswalk — the reconciliation of 4,903 incompatible metadata schemas, more than 290,000 artist records and 304 culture labels into one queryable space, with an embedding laid over the top so that a vague description finds an object. That is registrar work at a scale no registrar has ever been given, and it is the single copy in the story. The machine-learning part is the cheap part. The reconciliation is the expensive part, and it is invisible, and it has one custodian, who is an email address.

I knew a registrar once who held the concordance between two accession systems after a merger, most of it in her head and a spreadsheet she maintained on her own initiative. When she retired nothing was stolen and nothing decayed. Every object stayed exactly where it was, perfectly safe, and for about three years effectively unfindable. Objects survive custodians. Indexes do not.

What I’d watch for: whether anyone mirrors The Last Museum, and whether any consortium of the 4,903 ever offers to take the index into custody. My guess is no, on both counts, and not for technical reasons. There is no accession category for somebody else’s crosswalk, and taking it in would mean conceding that an anonymous party did the union of their catalogues better than they did.

Two numbers came out of this week: 1.9 million and 6,308,174. Only one of them appears in a set of trustees’ accounts. The other one is on a homepage, and the site disagrees with itself by more than a million.