AI in Tring: a museum-hand approach

There is a museum in Tring that should not exist.

Half a mile from the railway station, set back behind the High Street, is a Victorian gallery containing the largest private natural history collection ever assembled in Britain. Walter Rothschild built it in the 1890s. It holds the only thylacine you can see in person without leaving England, giant tortoises he is said to have ridden through the park, a polar bear in the front hall, and shelves of birds collected from places nobody had been. The man kept a zebra-drawn carriage. He published over a thousand scientific papers. He never had a proper job.

Most of what he loved looked, at the time, useless.

If you live in Tring, you are within walking distance of a long argument that being unreasonably curious about something can pay off in ways nobody quite predicts. That is the right frame for AI in Tring. Not the productivity hacks. The museum hand.

The labelling cabinet

Rothschild's obsession was naming. Two and a half million specimens, each with a label, a date, a location, a binomial. He paid people to do nothing else.

The AI in your pocket is essentially the same instinct, made cheap.

It can label your photos by what is actually in them, not just when they were taken. It can tag the emails you keep meaning to come back to. It can group your last six months of receipts into the categories your accountant wanted. It can read a year of meeting notes and tell you which three things keep coming up. None of this is heroic AI. It is the museum hand, applied to the contents of your day.

The boring use case is also the most reliable one. The shoebox of receipts. The four-year photo backlog. The unread folder you stopped opening in February. AI is at its best on these. ChatGPT in Tring is being used most often, by the people I know who have tried it, on exactly this kind of work.

The contradictions room

The most interesting specimens at the museum are the ones that do not fit. The kakapo, a fat green parrot that forgot how to fly. The cassowary, which can disembowel a horse. The polar bear in the vestibule, where you do not expect a polar bear in Hertfordshire.

The use case AI is undersold for is the contradiction-finder.

You have a forty-page planning brief and you suspect there is a quiet contradiction somewhere — between what the front summary promises and what the appendix actually permits. You have a contract and you cannot tell if the indemnity clause matches the warranty clause. You have a long email thread where someone has changed their position three times and is acting as though they have not. Drop it in. Ask one question.

Most lawyers in town can do this. Most residents cannot, and pay lawyers who can. AI does it for free, in thirty seconds, on a phone, while you are queuing at the Town Hall post office. That is not a small change.

The eccentric experiments wing

Rothschild rode a giant tortoise. He kept a herd of cassowaries. He sent collectors to New Guinea for birds nobody had described. Most of his "useless" eccentricity turned out to be where the value lived.

The same is true for AI. The first useful thing you do with it is rarely the productivity case the headlines promise. It is something stranger.

Ask it to translate a French menu you photographed at a restaurant in Aldbury. Ask it to write your dad's eightieth birthday speech in his own slightly grumpy voice. Ask it to argue back against your portfolio before a London pitch, in the voice of a hostile client. Ask it which three songs from a friend's last playlist suggest they might like a band you discovered. Ask it to draft a Wikipedia-style entry on your favourite Tring street, then check what it got wrong.

None of these are the productivity case. All of them are how you find the corner of AI that is genuinely yours, and from there the practical uses arrive without any pushing.

The right way to use the museum hand is to give curiosity its head and see what gets labelled later.

Where to start

Open Claude or ChatGPT on your phone. Pick the labelling job you have been avoiding, or the document you suspect contradicts itself, or the eccentric ask you have been keeping to yourself. Spend twenty minutes.

If you commute, our guide to Claude on the Euston run has three things to try on tomorrow's train. Five miles down the line, the Berkhamsted guide takes a different angle. The Hemel guide too. New to all of it? AI on your phone is the entry point.