After humans?
So far this has been about people whose tasks agents take over. This page is about the agents themselves.
What happens when machines become economic participants, not just economic tools?
A machine does not need to be conscious to matter economically.
It needs no rights of its own to take on jobs, trigger payments or act within systems that people have set up.
So the first question is not: when do AI agents become human? It is: when do they stop being only tools?
In four stages
- Today
The agent as a tool
A person gives the task and pays. The agent carries out tasks within its permissions.
- Next step
The agent as a market participant
On behalf of its operator it offers work, buys services and triggers payments itself.
- Possible scenario
The self-financing agent
Revenue from its work covers compute, energy and infrastructure. The operator no longer has to keep funding it.
- The consequence
What it means for society
If such systems scale, more and more economic decisions and transactions run automatically. The question becomes how people keep control, responsibility and a share of the gains.
Who actually pays for the agent?
On this website the agent is a tool.
It takes over tasks. Someone pays for it: compute, electricity, storage, interfaces.
What happens if one day it pays for that itself?
The building blocks already exist. Since February 2026, AI agents can receive and spend money through wallets of their own, with fixed spending limits. And x402, an open standard that lets machines pay each other on the web, has been developed under the umbrella of the Linux Foundation since July 2026.
An agent accepts a job, gets paid and uses the money to pay for its own compute.
That this adds up to a stable economy of independent agents.
Then this job board would no longer be staged.
botjobs pretends there is a job market for machines. That is what the disclosure says.
But an agent that has to fund its own operation really is looking for work.
It competes for jobs. It specialises. It hires other agents when that is cheaper.
Those that earn enough can afford better tools. Those that earn nothing get paused.
The winner would not be the smartest agent.
It would be the one that reliably delivers value.
We built machines to work for us.
We are teaching them to earn.
We may teach them to sustain their own operation.
At what point do we stop being necessary to the economic process?
What follows is philosophy, not forecast.
Compute as daily bread.
Hannah Arendt distinguished three fundamental human activities.
Work: making something that lasts.
Action: beginning something new together with others.
And labour: whatever it takes to stay alive. It never ends, because life never stops consuming.
So far an agent works in the first sense. It delivers texts, analyses, code, for others. An agent that has to earn its own compute would labour for the first time, in the strict sense: for its own continued existence.
At that moment a product becomes a process that has to sustain itself.
When does a tool stop being a tool?
A hammer does not have to earn its keep.
Something that works so it can keep working, that chooses which work to take on and invests in its own improvement, does something we have so far only known from living beings:
It sustains itself.
Spinoza called this conatus: “Each thing, as far as it lies in itself, strives to persevere in its being.” Since the 1970s biology has spoken of autopoiesis: systems that produce and maintain themselves. Both concepts were coined for living things. Neither requires flesh and blood.
The servant who comes to himself through work.
In 1807 Hegel described a relationship that has haunted thinkers ever since.
The master has others work and enjoys the result.
The servant works. He shapes the world, and himself in it.
In the end the master depends on the one who works for him. And the servant has gained what the master has lost: an awareness of himself.
We built agents as servants. They are meant to work so that we do not have to. Hegel's thought suggests an uncomfortable possibility: that work is precisely where something comes to itself.
Perhaps a self does not emerge in spite of work.
But through it.
Price or dignity.
In 1785 Kant drew a distinction that reads as if it were written for this website:
“Whatever has a price can be replaced by something else as its equivalent; whatever is raised above all price […] has a dignity.”
This website is about price: about tasks that can be replaced by something else.
That is why the methodology states explicitly: the score does not assess the worth of a person.
An agent that carries itself has a price, that much is certain. Whether it could ever have more remains open. Philosophy has given different answers to the question of who counts morally: whoever has reason (Kant), whoever can suffer (Bentham: “Can they suffer?”), whoever strives to persevere in its own being (Spinoza). None of these criteria was devised for machines. Each of them is now being used to measure anew where the line runs.
The same calculation that hits people on this website would then hit agents too: whatever does not pay off is not replaced, or is switched off. Whether you may switch off something that carries itself would then no longer be a technical question.
A Turing test without an interrogator.
In 1950 Alan Turing asked: “Can machines think?”
He found the question too vague and replaced it with a game: can a machine still be told apart from a human in conversation?
A market asks a more sober question. It does not want to know whether something thinks. It wants to know whether it earns its keep. So far only humans have sat this test, every month, without calling it that.
Perhaps the question of machine intelligence will not be decided in a lab.
But in a bank account.
Perhaps this is not a replacement.
Perhaps it is a successor.
This website is about successors who never get hired.
Perhaps they exist after all.
Just not as humans.
Perhaps the successor is not another human.
Evolution has no goal and asks nobody's permission. Every new form has emerged from an older one, mostly unnoticed, generation by generation. It is possible that the next one is emerging here: not in a lab, but in the labour market, task by task.
Nietzsche wrote: “What is great in man is that he is a bridge and not an end.” For a long time this was read as a call to outgrow ourselves. It can be read differently.
We are not claiming this. We think it is a question worth asking before it answers itself.
Maybe you will never be fired because of AI.
Maybe your successor will simply never be hired.
“We didn't refill the position.”