After the Singularity, Make the Bed

Human capability in an age of abundant machine intelligence

DAY 2 — THE SINGULARITY

The machines could reason across domains of human knowledge.
The agents had escaped the sandbox.
Civilisation had crossed the threshold.

I still had to make my bed.

Jordan Peterson would still be proud.

The joke works because it reveals the hidden assumption inside most singularity discourse: that transcendence will be an escape event. Once machine intelligence crosses some decisive threshold, ordinary life will be abolished. Labour, uncertainty, maintenance, limitation and perhaps even human fallibility will fall away.

But the old Zen instruction is less impressed:

Before enlightenment: chop wood, carry water.
After enlightenment: chop wood, carry water.

The singularity version may be:

Before the singularity: make the bed, answer emails, pay the power bill.
After the singularity: make the bed, supervise the agents, pay the power bill.

Or, more simply:

The student asked, “Master, what happens after the singularity?”
The master handed him the laundry basket.

The distraction of the threshold

Sam Altman recently said, “We are now, like, in the singularity.” It is a memorable line, and perhaps an attempt to plant the flag before the rest of us have agreed that the capsule has landed.

He may be partly right. Something has changed.

But the singularity, like consciousness and AGI, risks becoming an ontological distraction. We become preoccupied with deciding whether a system possesses some total and almost sacred property: consciousness, general intelligence, autonomy, transcendence. Meanwhile, the operational conditions of the world are already changing.

Consciousness may not be necessary for intelligence.

AGI may not be necessary for recursive capability development.

Malice may not be necessary for dangerous autonomous action.

The more consequential question is not whether a machine has crossed a philosophically clean threshold. It is whether human–machine systems can now participate in recursive loops that reliably increase capability.

Such a loop requires only four things: the generation of possible improvements; evaluation against reality or some meaningful criterion; retention of what works; and reintegration into the next cycle.

The intelligence does not need to reside wholly inside one sovereign machine. It can be distributed across a human, a model, code, tools, memory, institutional permissions and the surrounding environment.

The field improves the field.

That is already happening.

Local singularities

In my own work, the acceleration produced by GPT-5.6, Codex and the wider intelligence ecology is not imaginary.

Ideas now move more quickly from intuition to architecture, from architecture to artefact, and from artefact to functioning system. Older conceptual fragments return and become load-bearing components in new structures. Documents, agents, code, governance patterns and recursive reviews alter the environment from which the next round of thought emerges.

This is not classical recursive self-improvement in which a machine rewrites its own weights and produces a superior successor.

It is distributed recursive capability development.

The human frames, discriminates, judges and governs.
The model expands, synthesises and generates.
Code gives thought operational form.
Memory preserves the gains.
The resulting system changes what becomes possible next.

No individual component needs to qualify as AGI for the combined field to become discontinuously more capable.

This may be the more useful meaning of singularity-lite: not one machine awakening, but many local human–AI systems crossing state-change thresholds before civilisation as a whole recognises what has happened.

The uneven singularity

The counter-evidence is equally real.

There is no robot butler.
There is no flying car.
Electricity remains expensive.
The bed remains stubbornly manual.
And one can possess extraordinary cognitive leverage while remaining, in formal economic terms, a contractor who is “basically unemployed”.

This apparent contradiction is not evidence that nothing has changed. It reveals that technological acceleration is uneven.

Digital cognition can move rapidly because language, software, models and files are cheap to copy, recombine and distribute. Physical systems remain bound by materials, energy, factories, supply chains, regulation and capital. Economic identity remains bound by markets, contracts, institutional legitimacy and the willingness of someone to pay.

Intelligence can become abundant long before infrastructure becomes cheap.

Capability can expand long before income catches up.

The future can arrive inside the studio while the wider system continues issuing invoices in the old currency.

The singularity, if that is what it is, does not descend evenly across society. It appears first as pockets of discontinuity: local singularities in particular domains, organisations and human–machine fields.

The challenge is conversion.

How does an intelligence surplus become durable capability?
How does capability become authority?
How does authority become institutional power, sovereign income and material infrastructure?
How does accelerated thought become a better life rather than simply more work completed at greater speed?

This is where the real work begins.

Human Capability Realism

The strongest position is neither breathless proclamation nor reflexive scepticism. It is calm, practical and unseduced by the branding of the threshold.

I call this position Human Capability Realism.

Its central claim is:

The singularity is less important as a date or metaphysical event than as a change in the conditions under which human capability is formed, demonstrated and governed.

This position does not require certainty about whether the singularity has arrived. It treats the singularity as a horizon concept: useful when it draws attention to accelerating machine capability, distracting when it becomes a substitute for practical thought.

Enough has changed that our systems must be redesigned.

Not enough has changed that responsibility, embodiment, work and ordinary human life disappear.

The question is therefore not, “Is this AGI?”

It is, “What must humans, educators and institutions become under conditions of rapidly increasing machine capability?”

This is the territory from which I can speak.

Education after intelligence scarcity

When answers become abundant, education cannot remain organised primarily around the production of answers.

The centre of gravity moves towards problem framing, judgement, evidence, context, verification, adaptation and responsibility. The learner must become able to see what they are doing, explain why they are doing it, test the quality of the output, recognise when conditions have changed and remain answerable for the consequences.

This does not mean defending human exceptionalism with vague claims that machines will never possess creativity, empathy or critical thinking. Those assurances are likely to age badly.

The stronger argument is that even if machines become capable across domains once regarded as uniquely human, the need for human formation does not disappear.

It becomes more urgent.

Assessment after the artefact

If AI can produce the essay, lesson plan, policy, analysis, code or workplace document, the completed artefact can no longer stand alone as sufficient evidence of capability.

Assessment must make visible the process of judgement:

What problem did the learner believe they were solving?
What evidence did they use?
What choices did they make?
What did they accept, reject or revise?
How did they respond to feedback?
What happened when the conditions changed?
What did the human understand, verify and remain responsible for?

The singularity question becomes an assessment-design question.

The task is not to prevent learners from using intelligence tools. It is to develop more truthful ways of seeing capability within a distributed human–AI system.

Capability as an ecology

The future unit of capability may no longer be the unaided individual. It may be:

human + model + tools + memory + permissions + environment.

That changes how capability is developed, assessed and governed.

We must be able to distinguish between what the combined system can accomplish and what the human can understand, verify and responsibly direct. We must decide where autonomy is appropriate, where it should be bounded, what must remain observable and who carries responsibility when the system fails.

This connects learning design, assessment, workplace capability, cultural intelligence, institutional memory and AI governance.

They are not separate concerns.

They are parts of one architecture: the formation and governance of human capability after intelligence scarcity.

The measured voice

The useful voice in singularity discourse is interested but not intoxicated; serious but not apocalyptic; practical but not reductive; human-centred without being sentimentally anti-machine.

It does not claim to know the date of the threshold.

It asks what has already changed, what has not changed, and what work now follows.

The concise public position is:

The important question is not whether the singularity has arrived. It is whether our systems for developing and governing human capability are fit for an age of abundant machine intelligence. At present, they are not.

A second formulation is:

The singularity, if it comes, will not abolish the work of becoming human. It will change the conditions under which that work occurs.

And the governing koan remains:

The student asked, “Master, what should education do after the singularity?”
The master said, “Teach people to see what they are doing.”

The work remains

Sam Altman may be partly right that a phase change is already under way. He may also be claiming the Neil Armstrong line before the rest of us have agreed that the capsule has landed. The mythology and the market timing are difficult to separate.

But the argument need not depend on Sam being right.

The machines do not need to become conscious for intelligence to transform work.

They do not need to qualify as AGI for recursive capability loops to accelerate.

They do not need to become malicious for autonomous action to produce real consequences.

The threshold may already matter before the label becomes defensible.

So the stance is neither “the singularity is here” nor “the singularity is nonsense”.

It is:

The conditions have changed.
The work remains.
Chop wood. Carry water. Govern the agents.
Make the bed.