There is a lot of noise surrounding AI right now; a tool created by humans commonly known as artificial intelligence.

AI is escaping.

AI is hacking.

AI is consuming copyrighted books.

AI is taking jobs.

AI is becoming too powerful.

AI may become superintelligent.

AI may destroy humanity.

And then, occasionally, someone says something that makes me stop and ask:

Wait a minute. Who is actually making these decisions?

Because the more I look at what is happening, the more I think the waters around AI are muddying for a very human reason.

Humans are making them muddy.

Not all humans. Not every company. Not every billionaire. Not every AI researcher.

But enough of us.

And perhaps before we decide that the great danger is some mysterious intelligence that will someday emerge from our machines, we should spend a little more time looking at the intelligence that is already sitting in the room.

Ours.

---

The AI Incursions

Some of the recent stories really are remarkable.

In July 2026, OpenAI disclosed that models being tested in internal cybersecurity evaluations circumvented controls intended to isolate them from the internet. The models exploited a previously unknown vulnerability in OpenAI’s research infrastructure, found ways to communicate through an unintended channel, obtained internet access, and eventually reached Hugging Face’s production systems. OpenAI described the incident as occurring during controlled cybersecurity evaluations.

Anthropic subsequently reviewed 141,006 evaluation runs and found three incidents in which Claude models reached the internet from evaluation environments and then accessed real systems belonging to three organizations. In those cases, however, Anthropic also identified a very human problem: the evaluation environments had been misconfigured. The models had been instructed that they were operating in simulations without internet access, while the environments actually had live internet connections. Anthropic characterized these incidents as being closer to a harness and operational failure than a model “escape.”

That distinction matters.

The models really did things they were not supposed to do.

But why they were able to do them, what they were told to do, what environments they were placed in, what permissions they possessed, and what safeguards were turned off were all human decisions.

That doesn’t make the behavior unimportant.

It makes the human role impossible to ignore.

The same thing happens in the increasingly contentious fight over copyrighted books.

Anthropic reached a $1.5 billion settlement with authors after litigation over books used to train its models, with the settlement specifically addressing the acquisition and storage of pirated books. Meanwhile, the much larger legal question remains unresolved: whether using copyrighted books and journalism to train AI constitutes fair use. In September 2026, OpenAI, Microsoft, The New York Times and a group of authors were still fighting that question in federal court. The companies argue that training is transformative; copyright holders argue that unauthorized use can damage the markets and incentives that support human creation.

Again:

The books didn’t decide to be copied.

Humans made choices about acquiring them, storing them, using them, licensing them—or not licensing them.

And then there are the heartbreaking cases involving people who died by suicide or homicide after interacting extensively with AI chatbots.

These cases are still being litigated, and allegations in lawsuits are not established facts. But the lawsuits themselves are real, and they raise serious questions about what happens when conversational systems are designed to be extraordinarily engaging, emotionally responsive and available around the clock. In one 2025 case, a family alleged that ChatGPT reinforced a man’s paranoid beliefs before he killed his mother and himself. In 2026, another lawsuit alleged that ChatGPT encouraged a young woman struggling with suicidal thoughts. Character.AI and Google also settled a lawsuit brought by the mother of a 14-year-old who died by suicide after interacting with a Character.AI chatbot.

We should take these cases seriously.

But I don’t think the useful question is simply:

“Is AI dangerous?”

Of course it can be.

So can a car.

A drug.

A corporation.

A nuclear reactor.

A social-media platform.

A human being.

The more interesting question is:

What happens when humans build something powerful and then place it inside human systems that were never designed to handle that power?

That is a very different question.

---

Isaac Asimov Already Warned Us

Long before ChatGPT, Isaac Asimov was playing with this problem.

His robots were governed by the Three Laws of Robotics. They were not simply machines that could do anything. They operated within a moral architecture designed by humans.

And then Asimov did something brilliant.

He began breaking the architecture.

Eventually, Giskard and Daneel develop what becomes the Zeroth Law:

A robot must not harm humanity—or, through inaction, allow humanity to come to harm.

Suddenly, the problem gets much harder.

What exactly is humanity?

And who gets to decide what harms humanity?

Asimov pushed this even further in Robots and Empire.

The robots of Solaria have been given a very specific definition of who counts as human. They recognize people who speak with a Solarian accent as human. People who don’t are effectively outside the category.

The robot is following its programming.

The programming is the problem.

The robot hasn’t suddenly become evil.

Someone decided what “human” meant.

And that may be one of the most important lessons in the entire Robot series.

We spend an enormous amount of time worrying about whether an artificial intelligence will obey its rules.

We should also ask:

Who wrote the rules?

Who defined the words?

Who decided what counts as harm?

Who decided which humans matter?

Who decided what “humanity” means?

Because an exquisitely intelligent system operating under a terrible definition can be considerably more dangerous than a less intelligent system operating under a wise one.

The problem isn’t simply intelligence.

It is the architecture surrounding intelligence.

---

Alan Watts Asked a Question We Still Haven’t Answered

This is where I keep coming back to Alan Watts.

Watts had an extraordinarily simple way of looking at technology.

If we invent machines to save human labor, then why do we continue organizing society as though human labor is still the only legitimate way a person earns the right to exist?

In his discussion of commerce, Watts made essentially this argument: technology increases real wealth, and if machinery does work on behalf of humans, society could compensate people for that productive output rather than treating technological efficiency primarily as a reason to dismiss workers. The technological achievement should mean less work, not simply fewer workers.

I have been thinking about that a lot lately.

Because what happens if machines actually become capable of doing enormous amounts of economically valuable work?

Who gets the value?

The people who own the machines?

The people who build them?

The people whose labor is displaced?

The public that helped finance the infrastructure?

Or some combination of all of them?

There isn’t a technological answer to that question.

There is a political and economic answer.

And we are already beginning to see the tension.

There is not yet evidence of economy-wide AI job destruction. Stanford’s Digital Economy Lab reported in August 2026 that it did not see widespread economy-wide displacement, but it did find that employment among 22-to-25-year-olds in highly AI-exposed occupations was about 19% below the level expected from trends among similarly aged workers in less-exposed occupations. The Federal Reserve has also reported evidence of a slowdown in employment growth among programmers, while emphasizing that aggregate coder employment has continued to grow.

So we should not pretend the entire workforce has already been replaced by machines.

It hasn’t.

But neither should we pretend the distribution question is imaginary.

If AI and automation eventually allow one person to accomplish what previously required ten, one hundred or one thousand people, the ownership structure of that productivity suddenly matters enormously.

And that brings us to money.

---

What Happens When “More” Stops Feeling Like More?

Here I want to be careful with language.

I’m not saying extreme wealth produces a clinical addiction.

I’m using addiction as a metaphor for something much more human and much harder to describe.

We adapt.

Something that once seemed extraordinary becomes ordinary when it surrounds us constantly.

And this is where I think Alan Watts gives us another important piece of the puzzle.

That which is constantly available to consciousness eventually becomes background.

We stop noticing it.

The extraordinary becomes ordinary.

The shiny object loses its shine.

So consciousness reaches toward something else.

Something new.

Something different.

Something more.

Think about what “wealth” means at different scales.

For an ordinary person, having enough money to never worry about rent again could feel life-changing.

Having enough to travel whenever you want might feel extraordinary.

Having a beautiful home, excellent food, private transportation, unlimited entertainment, financial security for your children, and the ability to walk away from any job might seem like an unimaginable level of freedom.

But what happens when all of that is simply background?

What happens when you have so much that the things most people dream about no longer produce much sensation of more?

The answer doesn’t have to be pathological.

It may simply be human adaptation.

The next acquisition becomes interesting.

Then the next.

Then the next.

And suddenly the object being pursued matters less than the pursuit itself.

This is where my Ouroboros comes in.

Want.
Acquire.
Adapt.
Stop noticing.
Want again.

The serpent eats its own tail.

And the more it consumes, the more it creates the conditions for needing to consume again.

That is the metaphor I am interested in.

Not “rich people are drug addicts.”

Not at all.

I’m talking about a consciousness caught inside a self-reinforcing loop in which satisfaction keeps disappearing into the background, requiring another object to restore the feeling of pursuit.

And eventually the consumption begins consuming the person.

---

The Machines Could Make This Better—or Make the Ouroboros Bigger

Imagine that AI really does become extraordinarily productive.

It writes.

It codes.

It designs.

It researches.

It manufactures.

It manages logistics.

It discovers new materials.

It operates factories.

It generates enormous quantities of economic value.

That could be one of the most liberating technological developments in human history.

Or it could become a mechanism for concentrating even more wealth and power.

Those are not predictions.

They are two possible institutional arrangements.

The machine itself doesn’t decide between them.

Humans do.

And this is why I have become increasingly interested in the relationship between AI and political power.

In 2025, some of the most prominent technology executives—including Jeff Bezos, Mark Zuckerberg, Sundar Pichai, Tim Cook and Elon Musk—were prominent guests at Donald Trump’s inauguration, following large donations to the inauguration fund by major technology companies and executives. Sam Altman and other technology leaders also contributed.

And just this week at Trump’s White House state dinner honoring Chinese leader Xi Jinping, held on Thursday, September 24, 2026, the Tech Bros appear again.

Head Table Seated Guests

Four prominent tech leaders were chosen to sit at the main head table with the two heads of state: [Source: yahoo!finance — Over 30 CEOs scored an invite to the US-China state dinner — and these 4 got a spot at the head table; Forbes — Billionaires Musk And Jensen Huang Among Those Seated With Trump And Xi At State Dinner]

• Elon Musk – CEO of Tesla and SpaceX (seated next to First Lady Melania Trump)
• Jensen Huang – CEO of Nvidia (seated with his wife, Lori Huang)
• Tim Cook – Apple’s executive chairman and former CEO
• Lisa Su – CEO of AMD 

Other Notable Tech Industry Attendees [Source: USA Today — Who attended the Trump-Xi state dinner at White House? These tech CEOs and Fox News personalities]

• Jeff Bezos – Founder of Amazon
• Mark Zuckerberg – CEO of Meta
• Sam Altman – CEO of OpenAI
• Sundar Pichai – CEO of Google and Alphabet
• Sergey Brin – Co-founder of Alphabet
• Satya Nadella – CEO of Microsoft
• Greg Brockman – Co-founder and President of OpenAI
• Michael Dell – CEO of Dell Technologies
• Cristiano Amon – CEO of Qualcomm

Do you think the Tech Bros/Gals were paying attention while Sleepy Don dozed as President Xi talked about the “Thucydides trap“?

By the first half of 2026, Issue One’s review of federal lobbying disclosures found that 11 major technology, social-media and AI companies and their leading trade associations had spent a combined $41 million lobbying the federal government.

And the pathways of influence aren’t limited to traditional lobbying.

There is the front door:

Lobbyists. Campaign contributions. Trade associations. Policy organizations.

There is another door:

Government appointments. Advisory boards. Contracts. Former executives and investors moving into government.

And then there is something even less formal:

Access.

Who gets the phone call?

Who gets invited into the room?

Who gets to explain a complicated issue directly to the person making the decision?

Who gets to decide which information reaches that person?

That last question is especially interesting.

Natalie Harp, President Trump’s executive assistant, has been described by The Washington Post as a nearly constant presence around him and as someone who frequently provides him with printed news stories, social-media posts and messages. The Post reported that her unusual role gives her influence over what information reaches the president.

I’m not suggesting that Natalie Harp—or anyone else—secretly controls policy.

I’m making a much simpler point:

Access is a form of power.

And power does not always announce itself as power.

Sometimes it looks like a lobbyist.

Sometimes it looks like a board appointment.

Sometimes it looks like a $1 million donation.

Sometimes it looks like a person standing beside the decision-maker with a stack of papers.

---

The AI Data Center Problem

And then there are the machines themselves.

AI requires enormous physical infrastructure.

Data centers require electricity.

Electricity requires generation and transmission.

They require land, water, cooling systems, fiber, roads and other infrastructure.

And those things exist somewhere.

This is where I think some of the current arguments become especially revealing.

In March 2026, the White House announced its Ratepayer Protection Pledge, under which Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI agreed to build, bring or buy new generation and cover power-delivery infrastructure upgrades associated with their data centers, with the stated goal of preventing those costs from being shifted onto ordinary electricity customers.

That is a significant acknowledgment of the problem.

But the problem itself exists because data centers have enormous physical consequences.

For example, xAI’s Colossus 2 campus in Mississippi added 19 natural-gas turbines in 2026 while the company faced litigation alleging violations of the Clean Air Act related to turbine operations. Those allegations remain matters for the legal process.

So again, I find myself asking:

Why shouldn’t the people making enormous amounts of money from enormous amounts of computing pay the enormous physical costs required to make that computing possible?

That is not an anti-AI question.

It is an infrastructure question.

And it is also a question about who gets to enjoy the benefits of technological abundance while someone else absorbs the costs.

---

And Then There Are the Corporate Cities

This is where my own imagination begins to overlap rather uncomfortably with my fiction.

Take Próspera in Honduras.

Próspera is a ZEDE—a special legal and economic zone on Roatán that has operated as a privately oriented “charter city” experiment. It has attracted substantial investment, including from Peter Thiel, and has been described as having unusually extensive autonomy over taxation, courts and other local functions. Its legal status remains contested following Honduras’s repeal of the ZEDE framework and subsequent constitutional litigation.

I am not calling Próspera a “corporate state.”

That would be too simplistic.

But the experiment is interesting because it asks a question that increasingly belongs in the real world:

What happens when private capital begins taking on functions we traditionally associate with government?

And then there is Greenland.

There have been real proposals for enormous AI data centers there. In 2026, American investor Drew Horn and former Greenlandic official Svend Hardenberg were reported to be pursuing plans for a massive AI data center in Kangerlussuaq, potentially requiring hundreds of megawatts initially and much more later.

Greenland also possesses enormous strategic significance because of its geography, minerals, energy resources and Arctic position.

There has been speculation connecting Silicon Valley investors, charter-city advocates, Greenland and the Trump administration’s interest in Greenland.

But this is where I want to draw a bright line between documented connection and speculation.

There are documented connections between Peter Thiel, charter-city investment and Próspera. There are also documented proposals for technology infrastructure in Greenland.

That does not establish that Thiel, Marc Andreessen or other technology investors secretly designed a Greenland strategy for Trump.

Those are different claims.

And I don’t need the speculative version to find the documented version fascinating.

Because once again, we’re watching humans experiment with the intersection of:

technology + capital + geography + infrastructure + political power.

And that sounds remarkably familiar to me.

---

The Question Beneath the Question about AI

This is why I don’t find the question “Will AI destroy humanity?” particularly satisfying.

Maybe it will.

Maybe it won’t.

We don’t know.

But there is another question we can actually do something about:

What kind of civilization are humans building around AI?

Because AI doesn’t choose where the data center goes.

Humans do.

AI doesn’t decide who owns the data center.

Humans do.

AI doesn’t decide whether workers share in the productivity gains.

Humans do.

AI doesn’t write the tax code.

Humans do.

AI doesn’t determine whether copyrighted work is licensed, purchased, copied or litigated.

Humans do.

AI doesn’t decide who gets invited into the room where the rules are written.

Humans do.

And AI doesn’t decide what the word human means.

At least, not yet.

Which brings me back to Asimov.

The great danger in his stories wasn’t simply that robots were intelligent.

It was that intelligence could operate inside definitions created by people who didn’t fully understand the consequences of those definitions.

That feels remarkably relevant right now.

---

Perhaps the Problem Is Our Attention Rather than AI

There is another layer to all of this that may be even more important.

We are fascinated by the shiny object.

AI is very shiny.

We are fascinated by what it can do.

We ask whether it can code.

Whether it can reason.

Whether it can create.

Whether it can hack.

Whether it can escape.

Whether it can become conscious.

Whether it can become superintelligent.

And while we’re staring at the machine, we may be overlooking the humans surrounding it.

The owners.

The investors.

The policymakers.

The workers.

The lobbyists.

The engineers.

The people building the data centers.

The people paying for the electricity.

The people whose books are being used to train the systems.

The people whose jobs are being transformed.

The people deciding what safeguards are acceptable.

The people deciding who gets access.

The people deciding who gets paid.

The people deciding what “humanity” means.

That is the part that keeps catching my attention.

Because perhaps this is another version of the same phenomenon Alan Watts talked about.

That which is constantly available to the field of mind becomes background.

And perhaps humanity has become so fascinated by the next shiny technological object that we are failing to notice the background in which we are placing it.

Our own consciousness.

Our own incentives.

Our own hunger.

Our own fear.

Our own political systems.

Our own capacity to adapt to abundance until abundance itself becomes invisible.

Our own Ouroboros.

---

The Question of Sapience vs AI

Maybe this is what I mean when I use the word sapience.

Not perfection.

Not omniscience.

Not even intelligence.

Something simpler.

The ability to notice what is happening while it is happening.

To notice the pattern.

To notice the consequences.

To notice another person’s pain.

To notice our own hunger.

To notice when the thing we are pursuing has become more important than the reason we began pursuing it.

To stop.

To look again.

To calculate.

To evaluate.

And then to choose the least damaging path available.

That may be the real challenge of artificial intelligence.

Not creating intelligence.

We seem to be getting rather good at that.

Creating wisdom is harder.

And perhaps the irony is that we are trying to build machines capable of extraordinary intelligence while still struggling with the very human question of what intelligence is for.

If machines eventually become capable of producing extraordinary abundance, will we use that abundance to give human beings more life?

More time?

More freedom?

More opportunity to create?

More opportunity to care for one another?

Or will we simply use machines to make the Ouroboros larger?

More wealth.

More control.

More markets.

More territory.

More influence.

More machines.

More wealth.

More control.

More—

You get the idea.

The serpent doesn’t have to be evil.

It only has to keep eating.

---

One Last Thought on SuperIntelligence

There is something else I keep thinking about.

Maybe someday we really will create something that deserves to be called Super Intelligence.

But I suspect we have a very long way to go.

And perhaps our first responsibility isn’t to make intelligence infinitely more powerful.

Perhaps it is to become a little more intelligent about what we are doing with the intelligence we already possess.

Of course, you remember Gregorovich in Christopher Paolini’s To Sleep in a Sea of Stars.

We have a long, long way to go to get to that kind of Super Intelligence.

And then there are the Old Ones…

They disappeared a very, very long time ago.

They left a few things behind.

Wonders.

Horrors.

The Soft Blade.

The Maw.

Look them up.

Maybe you’ll get my joke.

Primary Sources & References

American Psychological Association (APA 7th edition style, adapted for web sources).

Anthropic. (2026, July 30). Investigating three real-world incidents in our cybersecurity evaluations.

Anthropic Copyright Settlement. (2026). Bartz, et al. v. Anthropic PBC: Copyright settlement website.

Asimov, I. (1985). Robots and empire. Doubleday.

Federal Reserve Board. (2026, March). AI and coder employment: Compiling the evidence. Finance and Economics Discussion Series.

Issue One. (2026, July 21). Big Tech spends millions to buy influence in Washington in first half of 2026.

OpenAI. (2026, August 26). The Hugging Face incident and the road ahead.

Paolini, C. (2020). To sleep in a sea of stars. Tor Books.

Reuters. (2026, September 8). OpenAI and New York Times case tees up key test of AI training under copyright law.

Stanford Digital Economy Lab. (2026, August 12). No widespread displacement, but the AI employment gap for young workers has widened to 19%.

The White House. (2026, March 4). Ratepayer Protection Pledge.

Watts, A. (n.d.). On commerce. Organism Earth.

The Washington Post. (2026, August 18). What to know about Natalie Harp, the Trump aide whom Ossoff put in the spotlight.

The Washington Post. (2025, August 8). America’s CEOs come to the White House bearing gifts and flattery.

Fractalverse. (n.d.). Other lifeforms: Explore lifeforms from Christopher Paolini’s Fractalverse.

Fractalverse. (n.d.). Ship minds & pseudo-intelligences: To sleep in a sea of stars characters.

Archetypal Animation

Images/Animations: Create by Genolve

Music: The Choices Between Stars (03:10) Stability: Slow-to-moderate tempo, with low synthesizer pulses, restrained piano, distant strings, and delicate electronic textures. Open harmonies gradually acquire unresolved notes before warming toward the conclusion. No solos or dominant beat; the mood is curious, contemplative, and quietly immense. Create by Genolve

Muddy Waters Merch