Notes from the studio.

One of my favourite places to buy books is at the shops attached to museums. For a long time, I lived walking distance to the Tate Modern, and to this day it is still my favourite place to buy books. (On the same walk I also love the second hand tables outside the BFI.)

To explain why, a little context about book buying and selling in a physical bookstore. I managed an independent bookstore years ago and learned why what you decide to carry is so challenging. If you let a book sit on your shelf for too long, called a turn, then are giving up a place for a book that might sell, take fewer turns, and depending on the sales of books in your neighbourhood that can impact the ability of the bookshop to exist at all. Also, bookstores can run with their tiny profit margin only because new books are returnable, and if you keep them too long, you risk not being able to return them and then you eat that cost.

Which is why, if you like to browse the weird and wonderful in (new) books, you have to go to a bookstore where they have a reasonable expectation someone walking the aisles will want to buy it. (Or you can special order it from your local bookstore, but I'm talking about finding a book because it caught your eye in person, jumped out at you from a curated bookshelf.) And where do these strange people like to go? Museums.

So what do I buy from the curated books in a museum bookstore? Well I have a very specific love of small books with a lot of information I wouldn't otherwise know. Books of portable discovery. Also I love foil on covers, like a magpie. One of my favourites that ticks both of these boxes is Countryside: a report. Today I'll talk about one section in it and how the idea has grown in my practice.

The book is by AMO, the research studio attached to Rem Koolhaas's architecture firm, and it came out alongside the Countryside, The Future exhibition at the Guggenheim in 2020. The premise is that the countryside, by which they mean roughly the 98% of the earth not occupied by cities, is changing faster and stranger than people are paying attention to. It is a collection of short essays and case studies from researchers working in Siberia, Kenya, Nevada, the Netherlands and China, among other places. The chapter Villages with Chinese Characteristics, by Stephan Petermann, is where I first heard of a Taobao village.

A Taobao village is a rural Chinese village where a significant share of households are running shops on Taobao, which is China's main online marketplace. Dongfeng, in Shaji in eastern Jiangsu province, was the first one. In 2006 a young man from the village opened a Taobao shop, first selling phone accessories and then furniture. His neighbours watched, copied him, and within a few years what had been an agricultural village had quietly stopped being one. By the late 2010s Dongfeng had around 16,000 online furniture shops, with workshops bolted onto traditional courtyard houses. When the AMO team revisited a year after their first trip, the model for the planned expansion had quadrupled in scale and they had cut a new road through the village called Taobao Road.

The bit that stuck with me was the description in the book of these places as factory, IKEA, and fulfilment centre in one room. Construction, sales, and distribution all happening in the same space, without the customer ever physically being there.

That model has a problem of its own, though. The Taobao village looks like a community of makers, but the production is shaped by the marketplace, not by relationships, and the goods tend to reflect that. Taobao furniture is well known for cheap materials, templated 3D designs remixed in software, and finishes that don't hold up close. The platform rewards the lowest price and the fastest delivery, which is a different incentive from the one a maker faces when they know their customers personally and need them to come back. Decentralised production can land in either direction, and the difference between the two is the difference between a village of small factories feeding an algorithm, and a high street where people know what they are buying and from whom.

Reading it back, the part I keep thinking about is the spatial reorientation. The farmers in Dongfeng did not move to a city to join the urban economy, they stayed where they were and joined it through the web. In the older model of rural to urban migration, the rural was a place you left because the market was elsewhere. In the Taobao village model the market comes to you, but only because the web is functioning, structurally, as the city.

This feels to me like the shape of work after AI. The middle layer of the in-person office, the agency with connections, the high street shop with overhead and the need for steady foot traffic, has always depended on density creating opportunity for demand. AI is reducing that layer from one direction, and the web has been reducing it from the other for two decades. People who built careers inside corporations are being laid off and becoming solo operators, often for the first time, and that pattern is going to keep accelerating. What is left, when those middle layers shrink, is the thing someone actually makes or does, and the people who want it, with the web providing the connection between them.

Which brings me back to the Taobao village model. My theory is that with the rates of unemployment rising, being able to be your own little factory of goods or services is going to become more common. Whatever you love to do, or have a unique interest in, or just can stomach doing for many years, you can control access to opportunity. A handyperson who works through a webpage rather than a shop with overhead, someone with a 3D printer in a spare bedroom making things to order, a small design studio publishing books from Devon.

Crucially, this is not a picture of everyone staying home on the web. It is more that the means of production and the choice of where to sell now sit with us. The web killed a lot of businesses but also made a different kind of business possible, and being replaced by AI can similarly evolve into using AI yourself, to run your admin and distribution, or even just to make a quick website. So when we do go to the high street, the bookshops and the shops are carrying more unique things, things we controlled ourselves. My own high street in Totnes is proof that buying intentionally from unique sellers works.

And the handyperson, working through their webpage instead of a shop, has the money to come in and buy a book because you could do a local Google search. Maybe that bookstore will be able to carry the weird little tomes I love.

Our old safety net is disappearing, and that loss is real, so we need to build a different kind of net out of the small shops and the people we keep going back to. If we harness the new technology to make this net, we need to think about how it can be stronger than it was before by prioritising peer to peer community.

Countryside, A Report by AMO and Rem Koolhaas, was published by Taschen in 2020 as the companion to the Guggenheim Museum exhibition Countryside, The Future. ISBN: 978-3-8365-8331-2

There is a kind of publishing that moves at the pace of the reader, not the platform. We try to do that kind.

Most of what we make is asked to last. A book that costs forty pounds to produce will be on someone's shelf a decade from now; a publication printed on newsprint will be folded, torn, and pressed flat again before it's recycled. Either way, we are designing for time, not for an algorithm.

Working slowly does not mean working fewer hours. It means letting the work tell you when it is ready, rather than the other way around.

One. Consent is the simplest test we know. If a person would not knowingly hand over their words, their voice, their drafts to a third party — would not, in writing, sign that contract — then the system asking for those things is not the system they should use.

Two. Locality is the next test. A model that runs on your machine, without a server, is a different kind of object from one that calls home. The first is a tool; the second is a relationship. Both can be useful, but they are not the same.

Three. Most of the AI systems on the market today are pitched as the first and built as the second. Choosing carefully — and reading what the model actually does, not what its homepage says — is increasingly the work.