Escort Economy: How We Fed the Middle Class to Moloch

Recently, a friend and I were having a beer and talking about serfdom in the Russian Empire. We got to the legend of the "right of the first night": supposedly, a landowner could turn up at a serf's wedding and sleep with the bride.

This was not an established legal right in Russia. But the story stayed with me. Someone can have the right to refuse without being able to survive the consequences. At that point, writing the dependency into law is almost redundant.

We tend to judge the future by the technology available to us. If there is a machine in your pocket that can answer any question, the future must have arrived. You might as well judge freedom by the internet speed inside a room.

I want to know whether they can leave the room.

Literacy you can rent

I've worked in programming and technology for more than ten years. For much of that time, people paid me to explain things to machines that other people couldn't explain.

You had to understand the problem, identify the entities, describe their relationships and rules. The machine had no idea what you wanted. You had to build that understanding by hand, then spend a long time discovering exactly where it had taken you literally.

Knowing that grammar paid well. A scarce skill gave you bargaining power. As long as getting the result without you was difficult, you had something to negotiate with.

Now some of that grammar is available by subscription. Code, architecture, documentation and workflows are becoming machine services. A model doesn't need to learn from every conversation we have with it. It only needs to solve a problem we used to hire a specialist to solve.

I see this in my own work. One designer handles the work of a small team. One person in sales runs processes that used to need several employees. I do engineering, product and operational work I would previously have shared among other people. A result that once took five programmers no longer always requires five.

And I have no intention of giving that up. My competitors have the same tools, and my moral discomfort isn't in their budget.

Moloch doesn't care

In the biblical tradition, Moloch is associated with child sacrifice. Researchers disagree about what the name originally referred to: a deity or a ritual. I'm using the cultural image here. You give up what matters most because, as someone has explained, things will be even worse if you don't.

Scott Alexander used it in his 2014 essay Meditations on Moloch. He examined situations where people act sensibly and together produce an outcome none of them likes. Changing your own behaviour from inside the system is difficult: the first person to surrender an advantage may also be the first to lose.

Imagine ten companies, each employing a hundred people. These are hypothetical numbers, a thought experiment. One company learns to produce the same output with twenty employees. Its costs fall. It can cut prices, increase profit or invest in growth. The other nine need to respond.

If they copy it, the same output will need two hundred workers instead of a thousand.

That doesn't mean eight hundred people will remain unemployed forever. Demand and output may grow. New products and jobs may appear. But none of these companies is obliged to make sure the new work arrives quickly enough, in the right place, at comparable pay. Each has its own accounts to answer for.

In engineering terms, local optimisation doesn't guarantee a global optimum. A company measures profit. A worker tries to preserve an income and some control over their life. Sometimes those goals align. Sometimes the worker discovers that their income was the very cost the company has successfully optimised away.

This is why "technology has always created new jobs" doesn't reassure me. Perhaps it will again. But there can be a rather inconvenient interval between the birth of a new industry and someone's ability to pay the rent.

No secret council of billionaires is required. I automate because it helps my business. My competitor does the same. An employee learns the next tool to avoid losing out to another employee. Everyone protects their position. Together, we may weaken the position of everyone who lives by selling their labour.

Moloch isn't evil. He just doesn't give a fuck.

Preserving our independence isn't a separate task on his list. We have to decide how it will work if producing the goods and services we're used to needs far fewer people.

You don't even have to fire anyone

In London, I meet people in private equity who buy existing companies. One model they discuss is straightforward: acquire a business and make it more efficient through automation.

You don't have to announce mass redundancies. You can stop filling vacancies. People change jobs, retire or leave for their own reasons. Their work gets divided between the remaining staff and machines.

A few years later, the business may serve the same number of customers with a smaller team. The investor is happy. So is the owner. The customer may get a cheaper service.

Meanwhile, a university graduate discovers that nobody abolished their future job. It was simply never advertised.

To become a senior, you first need somewhere to be a junior. To accumulate capital, you need an income. The company can manage without your first step into a profession. You can't, at least for now.

This change needn't arrive in one terrible year that makes everything obvious. First there are fewer vacancies, then worse terms, then less bargaining power. Education is still offered as a route upwards. But the existence of a ladder doesn't mean you can reach its bottom rung.

Your business, someone else's terms

One foundation of the middle class was the ability to turn work into ownership. Get an education, find a job, buy a home, build savings, purchase shares. That path was never open to everyone. But for many people it provided a reserve of independence: eventually, the next payday could become a little less frightening.

What remains for someone who hasn't accumulated anything yet, if their hard-won expertise is no longer scarce?

You can start a business. AI makes development, design and many other tasks cheaper. I can hardly deny it when I benefit from it myself. But making a product and finding people willing to pay for it remain different problems.

When almost anyone can write an app, having written one isn't much of an advantage. You need customers, reputation, capital and access to an audience. A platform where buyers and sellers already meet has value you cannot reproduce by copying its code.

Cheap tools give newcomers a chance. Distribution owners can still control who gets to realise that chance. Production becomes more accessible. Income need not become equally widespread.

British pubs offer a useful everyday example. According to a House of Commons briefing, in 2023 roughly half of UK pubs belonged to breweries or large pub companies. The other half were in the independent sector, including small chains.

Some tenants operate under a beer tie: their contracts require them to buy certain products from the owner. You serve customers, hire staff and carry the business risk. The asset owner sets some of the conditions your earnings depend on.

Tenants have rights and legal protections. A literal comparison with serfdom would be absurd. I'm interested in the arrangement itself: you run your business while someone else controls crucial conditions of its existence.

The arrangement is familiar in the digital economy. A seller depends on a marketplace, a developer on an app store, an author on a recommendation algorithm. You can build an AI business and discover that its economics depend on the pricing of a model provider and its computing infrastructure.

Usually, both sides benefit. The question becomes harder when alternatives disappear and leaving becomes unaffordable. Ownership then lets one party set the terms on which everyone else can earn a living.

Escort Economy

Now we can return to that conversation over beer. What interests me about the landowner's story is the possibility of refusing. If you can leave an employer, switch platforms, reject an investor's terms and survive the consequences, you have bargaining power. If leaving means losing your home and the money you live on, your "no" starts to mean something rather different.

My hypothesis is that, as skilled work becomes a less reliable route to independence, some social mobility will depend increasingly on the ability to be liked.

Appearance, visibility, charisma and loyalty become more important. So does entertaining people, holding their attention and getting into the right circles. People have to offer the market their biography and personality where the results of their work used to be enough.

The name that came to mind was Escort Economy: an economy where gaining access to resources increasingly requires the personal favour of those who own them.

This doesn't mean content creation isn't real work, or that an influencer cannot build an independent business. They can. But attention is limited. We cannot all be famous relative to one another, and patronage, by definition, goes to the people who get chosen.

If a good specialist can make a decent living without becoming a public figure or joining someone's inner circle, they have their own basis for negotiation. If competence is no longer enough, they also have to prove their personal appeal to whoever controls access.

Taken far enough, it becomes a familiar courtly arrangement. You were noticed, brought closer and given an opportunity. Now you mustn't fall out of favour. At a digital platform's court, you can spend years seeking an audience with an algorithm that doesn't even know the two of you have a relationship.

Technofeudalism and Escort Economy describe different sides of this hypothesis. The first asks who owns the assets and sets the terms of access. The second asks how someone without an independent livelihood finds a place in that arrangement.

Being chosen becomes economically more important than being competent. I see this as a risk, not the inevitable future of every profession in every country.

What I can hold on to

Execution is getting cheaper first: code, layouts, routine copy. Models can also propose ideas and combine them in unexpected ways, so I wouldn't defend the human contribution by pretending a machine can't think of something new. I'd look elsewhere. Someone has to decide which problem to solve, test an idea against real people and real money, work through the consequences, and commit to delivering it to a client. My hypothesis is that this work may become more valuable as execution gets cheaper.

That still doesn't get you a job. A valuable skill isn't an income. Much depends on reaching buyers: the narrower that access, the weaker your bargaining position. You can choose the right problems and take responsibility for the results, while a platform, a sole client or the owner of the computing infrastructure still holds most of the power. Personal qualities don't undo concentrated ownership.

That's why I'd look for specific ways to strengthen my bargaining position. Owning a product or a stake in an asset means some income needn't depend on a single employer's decision, although a platform can still change the terms. Several direct clients give you alternatives to one channel. Data and processes you can move make changing supplier a nuisance rather than a catastrophe. Shared ownership and common rules let a group negotiate from a position an isolated person doesn't have.

None of this guarantees freedom or an income. Each move may reduce a particular dependency without removing it, and leave you a better chance of saying no. Escort Economy suggests another route: make yourself a product for an algorithm and dance until somebody notices. Attention is a market too, and sometimes you have to dance. But if your only asset is somebody else's favour, you're still a tenant at their court.

This is also where personal advice reaches its limit. These options are most available to people with time, a financial buffer and room to survive a couple of failures. If the rent is due tomorrow, spending years building an asset is a harder proposition. A cooperative needs shared rules and a market it can enter, and sometimes finance, though it doesn't need a loan simply to exist. Data portability works when a rule guarantees it and somebody enforces that rule. Across a country, those individual choices depend on institutions and access to ownership.

I feed this machine too

I use Claude, ChatGPT, AI agents and automation every day. They give me capabilities I have no intention of surrendering. So I cannot take the comfortable position of someone observing from outside and condemning other people's greed.

I benefit from needing less hired labour for my work. I also dislike a society where people find it harder to gain independence through their own work. Both thoughts belong to the same person. They cause no contradiction in the profit and loss account.

Learning faster is useful advice for an individual. It won't answer the whole question. You can learn to work with the machine and still end up in a world where machines do so many people's work that the old income is no longer available to everyone.

I don't want to ban AI or preserve pointless work just to keep paying wages. Freeing part of our lives for children, friends, science and our own projects would be a wonderful result of automation. But we will have to organise the distribution of both free time and income ourselves.

That raises questions about competition and antitrust, access to computing and the taxation of economic rent. About workers and entrepreneurs owning productive assets. About whether an independent company can be built without handing the conditions of its existence to a single ecosystem.

I don't have one measure that will fix everything. But I see no reason to consider those questions settled because another company has reported higher productivity.

Perhaps in five years we'll find that we overestimated the pace of change. New markets and professions may emerge. Some intellectual work may retain its price. That future is entirely possible. I'm much less certain that the old path of "learn, earn, acquire property" will survive without our involvement.

I want a future where AI frees people from selling most of their lives to survive. For now, each successful piece of automation I introduce also frees a business owner a little more from having to pay people. I understand perfectly well why I take the next step. I'd also like to understand where the road leads.

At the end, we may find some familiar kings, aristocrats and peasants. Their technological capabilities will be unprecedented. Their ability to say "no" may be much as before.

The peasants will have free ChatGPT. And a subscription to everything else.