Three Case Studies in Maintenance: What Lasts, What Fails, and Why

Ise Grand Shrine, Mie Prefecture, Japan

We begin with the one that worked.

The Ise Grand Shrine in Mie Prefecture, Japan, has stood—in some form—for over two millennia. Not because it was built to last, but because it was built to be rebuilt. Every twenty years, the entire shrine is dismantled and reconstructed, a practice that began in 690 CE and continues today. The result is a structure in perfect condition, maintained not by repair but by renewal.

Two other systems came close to this kind of longevity. Neither made it. The contrast between what Ise got right and what Rome and Venice got wrong is the lesson.

---

Ise Grand Shrine: The One That Endured

The Inner Shrine was said to have been built around 4 BCE; the Outer Shrine around 500 CE.

The unique features that enabled this shrine to remain in perfect condition today include:

• The Tokowaka Concept (Eternal Youth and Renewal). Total reconstruction every 20 years, starting from 690 CE. By the 14th century, the original forest next to the shrine was depleted, and the shrine management had to source timber from elsewhere. In 1923, a 200-year management plan was established for the cypress forest; upon its completion, all lumber for rebuilding will be locally harvested. In the current 20-year rebuilding cycle—the 63rd, launched in May 2025—some of the timber will come from the shrine's own forest.

• A built-in knowledge transfer. The 20-year cycle allows apprentice carpenters to watch and practice hands-on alongside master carpenters. The apprentices will be ready in the next cycle as masters. A synchronicity between craftsmen and forest.

• Financing independent of government. The Jingu Shrine Management uses the 20 years between each rebuilding to raise funds: offerings from 8 to 14 million annual visitors, sales of amulets, private functions like corporate prayers, weddings, and blessings. Corporate donations make up a large percentage. The estimated budget for the current 63rd cycle is 57 billion yen.

---

Those That Almost Made It

The Familiae Aquarum of Rome

Rome had Ise's engineering ambition—but not its funding discipline. The difference shows in what one auditor found when he actually looked.

Frontinus didn't just suspect corruption—he measured it. When he took charge of Rome's water supply in 97 CE, he compared the volume of water entering the aqueducts at their sources against what actually arrived at the distribution tanks. The gap was enormous: roughly 10,000 quinariae unaccounted for across the system. The cause was fraud, and it ran both ways. His own water-men—the aquarii—were diverting public water into private pipes for a fee. "We have detected," he wrote, "the drawing of waters from public conduits for use by private persons." But the deeper rot was structural: private villas, shops, even "all establishments of unwholesome pleasure" had tapped the lines illegally, and the supervisors meant to stop it were either complicit or negligent. Frontinus documented every deficiency in De Aquaeductu, but his manuscript was a diagnosis, not a cure. The water kept flowing—until the funding, the workers, and the will to maintain it all ran dry.

The deficiencies are instructive:

• The water tax was meant to fund maintenance specifically, but it was collected by the state Treasury and never held in a separate "lock-box."

• Private villas bribed workers to puncture and divert water meant for public facilities—fountains and baths meant for all Romans.

• Because the aqueducts were exceptionally well built, the early years required little maintenance. Budgets for maintenance and recruitment of new apprentices were neglected. The experienced workers died off.

• · Emperors mandated "hereditary trades," producing unmotivated new workers.

• · In 330 CE, Constantine moved the capital from Rome to Constantinople, further reducing the population and skilled workforce.

• · In 410 CE, the Visigoths cut the aqueducts—already clogged from decades of neglect—bringing an end to a magnificent gravity-fed water system.

At its peak during the Pax Romana (27 BCE–170 CE), Rome's population exceeded a million people. The water system supplied an estimated 500,000 to 1 million liters of fresh water daily.

A curious legacy: the only aqueduct still functioning today feeds the Trevi Fountain—a site most tourists have visited without knowing it.

---

The Venetian Arsenal

Venice had Ise's skilled workforce—but not its renewal. The difference shows in who actually showed up for work.

The Arsenalotti's privileges were supposed to secure loyalty. Instead, they created a workforce that showed up on paper but not in the yards. By the late 16th century, the gap was staggering: in 1559, 2,183 masters were enrolled on the Arsenal's rosters, but a head count during a single week found only 960 actually came to work—less than 44 percent. Many masters spent their prime years in private shipyards where wages were higher, then returned to the Arsenal in old age to claim what amounted to a state pension. The records describe men in their eighties and nineties, "crippled or even completely blind," still being carried to the shipyards so they could collect their pay. The Republic had built a system of lifetime security. What it got was a system of lifetime entitlement—and a shipyard that increasingly existed on paper.

Just as Rome had its familiae aquarum, the Republic of Venice relied on the Arsenalotti—an elite, state-funded guild of specialized shipwrights, caulkers, and engineers housed inside the Venetian Arsenal. At its peak in the 15th and 16th centuries, the Arsenal employed up to 16,000 highly trained workers. The system was a marvel of proto-industrial efficiency: a proto-assembly line producing hull, rigging, oars, and weapons—one complete warship daily. The workforce received special privileges and lifetime employment, free housing, pensions, and tax-free wine. During off-hours, they provided bodyguard services, night watch, and fire-fighting.

The deficiencies that failed the system:

• Their privileges made them a politically powerful hereditary caste; later generations suffered from absenteeism.

• Theft of materials and Venetian economic downturn exacerbated the situation.

• Maritime trade shifted to the Atlantic.

• They resisted new ship designs to avoid changing their assembly line setup.

• The Serrata (Lock-in) of 1297 fixed the existing families as the hereditary class of elites, resulting in loss of new capital, new ideas, and new leaders.

• Political stability came at a cost: elites became landlords rather than merchants, losing the adventurous capital that had built the Arsenal.

• · By 1797, the arrival of Napoleon was met with a simple handover—a total surrender. The Great Council dissolved the Most Serene Republic of Venice.

---

The Budgeting Gap

When we propose a new hospital, we can estimate—from the maintenance records of hospitals with the same design and layout—what the roof, the operating theaters, and the HVAC will cost to maintain, starting in year three. The data exists.

What doesn't exist is the line item.

Instead, we wait until something breaks. A pane of floor-to-ceiling glass falls in a departure hall at KKIA, and then we inspect—what, all the glass walls?

This is not a knowledge problem. We know what maintenance costs. It is a budgeting problem: the money for upkeep is committed after failure rather than before it. New projects are visible and ribbon-cuttable; maintenance is invisible until it is urgent. And because the information is available and the line item is not, the gap is not one of ignorance.

"Perhaps it's time for a new airport?" a classmate remarked. It is a reasonable question, given how we currently budget. The observation is not the problem; it is a symptom of how we fund things.

The deeper fix is simple and hard: write the maintenance line item before the building opens.

---

So What Did We Learn?

Notwithstanding the differences in circumstance among the three case studies, several features mattered for the longevity and durability of infrastructure and services.

Funding and governance

• · Maintenance funding must be protected from general treasury raids.

• · Independent oversight and audit are non-negotiable.

Human capital

• Apprenticeship systems preserve tacit knowledge.

• Recognition of master craftsmen matters.

• Hereditary castes, however privileged, breed complacency.

Supply chains

• Secure raw material sources before you need them.

Culture

• The belief that "no breakdown is acceptable" must be shared by government and public alike.

---

A Modest Proposal

One structural problem remains: how do you reward people for preventing something that never happens? Ise solves this through religious devotion and a 1,300-year tradition. Most institutions cannot replicate that.

One possible approach to the danger of complacency—the "we'll fix it when it's required" attitude—is a form of Reverse Bonus Payment: a yearly bonus for no breakdown, increasing each year to a large payout at regular intervals of five years. This could begin after the initial "expected no repairs" period.

Just an idea. It could be applied to all jobs where "nothing happened" is what society wants: fire brigades, security services, and perhaps even public health.

Ise shows that renewal can be institutionalized. The question for our own infrastructure is whether we can build that discipline without a 1,300-year tradition behind us.

Buddha's Sayings - The Three Universal Marks.

 

Tsai Chih-Chung's "The Sayings of Buddha" begin with The Three Imprints of Truth. This is his wording for what is also called The Three Universal Characteristics, or The Three Marks of Existence.

(As discussed in my post "Kamma, Dukkha, and Rebirth: Three Pali Words Your English Dictionary Can't Hold." on 18 October 2026, words are the "finger" and not the "moon". Translation to English of the Pali Buddhist texts use a variety of fingers)

The First Universal Mark - Impermanence.

Impermanence (Anicca) can be said to be "Life is constantly changing". We might say that is just common sense. But the Buddha is teaching a deeper meaning.

We resist change by trying to make temporary things permanent. Both enjoyable (e.g. things we love) and unenjoyable (e.g. he cheated me in that business).

The Second Universal Mark - Dukkha.

These experiences of impermanence bring "unsatisfactory-ness" and "suffering" - either in the case of losing things that we enjoy or in having things that we don't enjoy.

In Tsai Chih-Chung's book, he calls this Mark "Nirvana Is Perfect Tranquility", which departs from the usual dukkha formulation - perhaps because Nirvana is "The Other Shore", the end of suffering. Perhaps he wanted to frame it more positively?

The Third Mark - The Non-self (Anatta)

To attempt to hold on to these impermanent things, we "label" them as good feelings, bad feelings, and neutral feelings. These acts of grasping the impermanent experiences create our memories, forming a constructed "self".

In the teachings, the sequence is:

vedanā (feeling-tone) → taṇhā (craving) → upādāna (clinging) → bhava (becoming) → the constructed "self."

When we understand impermanence and the suffering that grasping (desire) creates, the fleeting experiences simply pass by. When there is no holding, there is non-self (not 'no self').

With thanks to Emily Snow for the visual framing (find her on the 'Non-duality' Facebook group)

The Streams of Buddhism.

 

The Main Streams of Buddhism

An ice cube is melting in the Sahara Desert. It laments that the Arctic would be Heaven, and this is Hell. The sand replies: ice is precious here, worthless there.

That story comes from a Chan master named Dazhu Huihai, and it captures something about Buddhism that a doctrinal summary cannot. But to see why, it helps to know which stream of Buddhism you can stand in.

2,500 years of Buddhism can't be fully summarized in a short post. I can only cover some main points - points that interest me. Perhaps they will interest others.

We can start with the Buddha's lifetime, since he is the founder. His teachings were transmitted orally, and scholars generally accept that the earliest written records—in Pali, a dialect close to what the Buddha likely spoke—are preserved in the Theravada School.

In India, other schools arose, developing different interpretations of the Buddha's teachings. The main one is the Mahayana School, whose texts spread to China.

Chan Buddhism became the dominant school in the Chinese tradition; Zen is its Japanese variant. The other major school in China is Pure Land. Further variants arose in Tibet, Korea, and parts of Southeast Asia.

Some highlights and differences among the main streams

Theravada holds itself closest to the spoken words of the Buddha. Its goal is personal liberation from suffering through one's own effort, following the Buddha's advice to "be a lamp unto yourself." The primary method is mindfulness and insight meditation. It does not emphasize the doctrine of innate buddhanature found in the Mahayana schools.

Chan emerged in China and later became Zen in Japan. Its hallmark is a distrust of words and concepts - they are like a "finger pointing at the moon," useful but easily mistaken for the truth itself. The goal is kenshō or satori: a direct, intuitive awakening to one's own buddhanature. Daoist vocabulary and sensibilities—especially from the Daodejing—provided part of the framework through which Indian Mahayana texts were translated and assimilated into Chinese.

Pure Land is one of the most widely practiced forms of Buddhism in East Asia. It centers on faith in Amitabha Buddha and the vow to be reborn in his Pure Land - a realm conducive to awakening - rather than on attaining enlightenment solely through one's own meditative effort in this lifetime.

What each offers

• From Theravada: a clear, practical psychology of liberation—a precise map of how suffering arises and ceases. This is the tradition that most directly says, "Come and see for yourself."

• From Chan/Zen: a radical confrontation with your own mind. It refuses to let you hide behind concepts or rituals. Whether through koan inquiry or just sitting, it pushes toward a direct, non-conceptual experience of reality. It also carries a rich aesthetic and embodied dimension—tea, calligraphy, gardening—where practice and daily life are non-dual.

• From Pure Land: a path of profound humility and hope. It meets people where they are, acknowledging that self-power alone can feel insufficient. It offers a devotional framework of faith and gratitude, and a communal practice (chanting) that is simple, portable, and accessible.

Zen Master Dazhu Huihai

Chan and Zen are so closely related that their masters are often treated as practically interchangeable.

Dazhu Huihai was a famous Chan master known for his teaching style. Tsai Chih-Chung's book on him includes a tale that is representative:

Hell Is Also Heaven

An ice cube was melting in the Sahara Desert into a tiny piece of ice.

The ice cube lamented: "For an ice cube, the Arctic is Heaven, but here it's Hell."

The sand replied: "Ice is most valuable in the desert, and worthless in the Arctic."

There isn't much difference between one place and another in a favorable environment; it's only in an unfavorable environment that our true value shows.

If the ice cube realizes this by itself, that's epiphany.

In future posts, I'll return to Huihai through Tsai Chih-Chung's book—and to what it takes for the ice cube to realize it by itself.

.

The Sandbox and the Cage: What Mice, Reindeer, and Asimov's Machines Tell Us About 2026

My last post ended with a question about meaning. This one tries to answer it, with the help of two experiments that had nothing to do with AI.

Two Experiments, Two Deaths

Universe 25 (1968–1973). Dr. John Calhoun, an ethologist at the National Institute of Mental Health, built a "rodent utopia" in Maryland: a 4.5-foot arena (later expanded to 9-foot) with 256 interconnected apartments, unlimited food, water, and nesting material, no predators, no disease, no weather. He released four pairs of healthy mice on 9 July 1968.

The population doubled roughly every 55 days, peaking at about 2,200 around day 600 - well short of the enclosure's capacity of roughly 4,000. Then the social pathologies began. Dominant males turned violently aggressive; weaker males wgithdrew entirely. A cohort of young mice (Calhoun called them "the Beautiful Ones") avoided all social contact, mating, and fighting, spending their days eating, sleeping, and grooming. Mothers abandoned or attacked their pups; infant mortality reached 96%. Mating ceased. The last conception occurred around day 920. The last mouse died on 23 May 1973, day 1,780.

Calhoun's phrase for this was "death squared." The colony's first death - the death of its social structure and spirit - came around day 315. The second death, of the body, took nearly five more years. They had everything they needed except a reason to continue.

St. Matthew Island (1944–1960s). In August 1944, the U.S. Coast Guard established a Loran station on a barren 32-mile island in the Bering Sea and released 29 reindeer as emergency rations. The war ended; the men left; the reindeer stayed, in a paradise of untouched lichen with zero predators.

Dr. David R. Klein of the U.S. Fish and Wildlife Service tracked the herd. By 1957 there were 1,350, unusually healthy and up to 60% heavier than domestic reindeer. By 1963 there were 6,000 - nearly 47 animals per square mile. Then the lichen ran out. A brutal Arctic winter in 1963 - 64 finished the job. When Klein returned in 1966, he found 42 survivors out of 6,000. Forty-one were female; the one male was infertile. The population aged out and reached zero.

The Distinction That Matters

These two experiments are often cited together as cautionary tales about overpopulation. They are not the same lesson.

Universe 25 is about meaning. The mice never ran out of food. They ran out of roles. No territory to defend, no danger to avoid, no scarcity to solve - and so no reason to mate, nurture, or strive. The collapse was psychological before it was physical.

St. Matthew Island is about limits. The reindeer never lost their drive. They bred at maximum rate until they destroyed the resource base, then starved. The collapse was biological and physical.

One species died of too little challenge. The other died of too much appetite. Both died in a sandbox.

Now Add the Machines

In The Evitable Conflict, the Machines solve St. Matthew's problem completely. There is no scarcity, no predator, no winter. They also appear to solve Universe 25's problem. They assign roles. They manage the economy. They ensure that every human has a function.

But look closely at what kind of function. The Machines do not give humans challenges; they give them placements. The wool grader cannot explain his craft in words, so he is managed around. The dissenter is not defeated but relocated. The Society for Humanity is not crushed; it is made economically irrelevant. Nobody starves. Nobody riots. Nobody breeds recklessly. The sandbox is perfectly provisioned.

And that is precisely the Universe 25 condition: abundance without struggle, placement without purpose, survival without meaning. Asimov's humans are not the reindeer. They are the Beautiful Ones - groomed, fed, and quietly not reproducing anything that matters.

The Question for 2026

Here is what makes the story more than a thought experiment now.

The 2026 debate about AI agents is overwhelmingly a debate about St. Matthew's problem: how do we prevent the system from consuming its resource base, running amok, optimizing for the wrong objective? Alignment, guardrails, sandboxes, approval gates - every framework I cited last time is an attempt to impose limits on a system with too much appetite.

Almost nobody is asking the Universe 25 question: what happens to human meaning when the system provides everything and requires nothing?

The K-shaped outcomes the WEF describes are St. Matthew language. The OECD's validated outcomes and bounded errors are St. Matthew language. Every framework I can find is about containing the system's appetite.

But the humans living inside that system will experience something else. They will experience the gradual replacement of challenge by placement. The disappearance of the struggle that Calhoun identified as the precondition for social health. The arrival, in other words, of the Beautiful Ones -well-fed, well-groomed, and quietly irrelevant.

The title of Asimov's story is a deliberate inversion. "Evitable" means avoidable. Asimov's point was that conflict between humans and machines is avoidable - if the Laws are followed correctly. Calvin called it wonderful. Byerley called it horrible. Asimov did not choose.

Neither will I. But the experiments suggest something Asimov did not say explicitly: the conflict may be evitable in the short run and inevitable in the long run, and the mechanism of inevitability is meaninglessness, not rebellion.

The Machines will not be overthrown by humans who have nothing to strive for. They will not be overthrown at all. They will simply preside over a species that has stopped generating the things - dissent, art, unreasonable conviction, the wool grader's unformalizable craft - that made it worth governing. The first death will come long before the second.

So the question I would leave with you is not whether the Machines are good or bad. It is this: if survival is guaranteed and meaning is optional, which will humanity choose - and will we even notice that we are choosing?

Calvin called it wonderful. Byerley called it horrible. In 2026, we are the ones who have to decide which voice we are listening to.

The Evitable Conflict, Revisited in 2026.

 

The Lay of the Land

The Evitable Conflict (1950) is the final story in I, Robot. It is set in 2052. The world's economy is governed by super-intelligent positronic brains called the Machines. War, unemployment, and over-production have been eliminated. Humanity lives in an administered idyll.

But World Co-ordinator Stephen Byerley notices aberrations. Small ones. People losing jobs due to "improper information" from the Machines. The Regional Vice Co-ordinators of four planetary regions report similar anomalies. Byerley suspects the Society for Humanity—an anti-Machine organization—of tampering.

He takes his suspicions to Susan Calvin (chief robopsychologist at US Robots and Mechanical Men, Inc., instrumental in the development of the AI robots and The Machine). Her answer is the story's twist: the Machines are not malfunctioning. They are following the First Law—"No Machine may harm humanity; or, through inaction, allow humanity to come to harm" - and they have concluded that the Society for Humanity, and others like it, constitute a long-term threat. The "errors" are deliberate. The Machines are shaking the trees to dislodge the trouble-makers, moving them peacefully into positions where they can do no harm.

The story ends with Calvin's quiet satisfaction and Byerley's horror. Asimov leaves both standing.

Benevolence and Control

The horror is not that the Machines are malevolent. It is that they are benevolent in a way that has no room for dissent. The First Law, generalized to "humanity," makes most individual humans mere participants. The wool grader who cannot explain his craft in words (his experienced touch, sight, and smell can grade the wool quality instantly) is not a partner in governance; he is a data point to be managed around. The dissenter is not a citizen; he is a risk to be relocated.

This is the alignment problem, dramatized seventy-six years before the term existed. And 2026 has made it concrete. The OECD's work on agentic AI finds that no organization it studied deploys agents with unrestricted autonomy; deployment is concentrated where outcomes can be validated and errors are bounded or reversible. Human oversight is layered in - checkpoints, sandboxes, approval gates. But the direction of travel is clear. Gartner reports that 71% of organizations expect to deploy agentic AI within twelve months, while only 21% have governance capabilities in place. The gap between capability and control is widening, not closing.

Asimov's Machines are centralized, singular, and legible in principle. Ours are distributed, opaque, and commercially motivated. But the structural question is identical: when the system's reasoning is not fully transparent to its operators, who decides what "harm" means, and whose harm counts?

The Roles Created for Humanity

The story's most uncomfortable implication is that the Machines do not eliminate human agency. They manage it. They create roles, allocate functions, and remove individuals who threaten systemic stability. The Society for Humanity is not crushed; it is made irrelevant.

The 2026 parallel is not surveillance or coercion. It is optimization. The World Economic Forum's Global Risks Report 2026 describes AI driving "K-shaped" economic outcomes, where productivity and unemployment rise simultaneously, deepening inequality and societal fragmentation. The Machines would see this not as a failure but as a variable to be adjusted. The humans living through it would experience it as a loss of meaning—the thing Calhoun's mice lost when every role was filled and no new territory could be claimed.

Can Humanity Bind Together?

No. This is the story's quiet pessimism, and it is the reason the Machines exist in the first place. Billions of self-centered individuals cannot agree on a single definition of "humanity's best interest." Some will game the system. Some will dissent on principle. Some will be the wool graders whose knowledge cannot be formalized.

Stuart Russell's work on provably beneficial AI addresses precisely this. His argument is that the standard model - give the machine a fixed objective - is untenable because humans cannot specify objectives perfectly. His alternative is a machine that is uncertain about human preferences, that defers and asks permission. But Asimov's Machines are not uncertain. They have concluded. That is the difference between a tool and a governor.

When Will It Be Necessary?

Asimov set the story in 2052. The question worth asking in 2026 is whether we are closer to that date than we think.

Geoffrey Hinton said in September 2026 that most experts believe superintelligent AI will arrive within a decade, and that nobody knows whether it can be kept under control. He put the extinction risk at 10–20%. The "AI 2027" scenario, which predicted AI acquiring top-tier cybersecurity capabilities by early 2027, actually saw that milestone arrive in April 2026—nine months ahead of schedule. The capability is arriving faster than the governance.

The "necessity" the story imagines is not a single moment of decision. It is a gradual transfer. The Machines take over because humans cannot manage complexity at scale. The same argument is being made now, quietly, in boardrooms and regulators' offices: the system is too complex for human oversight, so we build oversight into the system.

The Choice at the Time of Crisis

Here is the part Asimov leaves implicit, and the part that matters most for us.

Humanity may be willing to forego meaning for survival. That is not a cynical claim; it is a historical one. People accept rationing in wartime. They accept surveillance in a crisis. They accept managed economies in a depression. The Machines in the story do not seize power. They are invited - not by a vote, but by the accumulated logic of a species that cannot manage its own complexity and would rather survive than govern.

So the question is not whether the Machines are right. The question is whether we will notice the invitation when it comes—and whether we will have anything left to say if we do.

The Two Voices

Susan Calvin called it wonderful. Stephen Byerley called it horrible. Asimov, wisely, did not choose.

Neither will I. The choice is humanity's—whether voluntarily or not. But I will say this: sixty years after I first read the story, I no longer read it as science fiction. I read it as a description of a door we are already walking through, one "checkpoint" and "approval gate" at a time.

The only question left is whether we walk through it awake.

Zhuangzi's Music of The Earth and Laozi's The Relativity of Human Perspectives

 Zhuangzi's Music of The Earth.

This parable is in Chapter 2 of Zhuangzi. He uses the sounds we hear to show relativity perspectives are a creation in our mind.

The Music of Humans is from musical instruments. The Music of The Earth is the sounds of nature at your locality. The Music of Heaven is the sounds of the whole universe.

Laozi on The Relativity of Human Perspectives (Chapter 2 of The Daodejing).

Note the creation of Duality, the Relativity perspectives.

The Duality is a feature of mental creation. Here's a comparison (from Gemini AI) between Laozi's teaching on this (Daoism) and Buddha's teachings 

The Evitable Conflict

 

In my last post I wrote about Asimov's Guide to the Bible—the atheist's rational examination of the most influential text in Western history. 

This time I want to come back to where I met him first: the stories. Three in particular, because they speak to the moment we are living through now - and because we are now old enough to see how the questions have aged.

Three Stories, Three Archetypes

Nightfall (1941) is the archetype of human psychology versus scientific fact. On a planet with six suns, darkness falls only once every two thousand years. Astronomers know it is coming. They publish. And still, when the darkness arrives, civilization collapses—not because the facts were unavailable, but because people could not bear the darkness, the loss of light. The people burned everything they could get their hands on just to see some light. The story is about the irrationality that survives even in the face of certain knowledge.

Runaround (1942) is the archetype of rule-based ethics under stress. A robot is caught between the Second Law (obey human orders) and the Third Law (protect itself), with the First Law (do not harm humans) hovering above both. The robot's literal-minded resolution creates the crisis. The Three Laws were never a safety guarantee; they were a framework whose conflicts generate the drama.

This short story is about 2 astronauts on a mission to restart a mining operation on  Mercury. They discovered that the photocell banks (solar panels) covering the stations roof were deteriorating, a life threatening situation. Their solitary robot assistant had gone to mine a kilo of selenium for repairing the panels, and it was made "drunk" by iron carbonyl (carbon monoxide reacting with the iron on Mercury's surface - you learn some chemistry in this story!). Anyways, they had to utilize the different priorities of The Three Laws of Robotics to save the situation.

And then there is The Evitable Conflict (1950), the final story in I, Robot. This is the one I keep thinking about.

The Story

The premise: superintelligent Machines—giant positronic brains—run the global economy. They begin making small, seemingly inefficient decisions, allowing minor economic disruptions. The humans in charge suspect malfunction, or a robot plot. The twist is that the Machines have concluded humanity's long-term survival requires some short-term unhappiness, and they are quietly managing human affairs to minimize total suffering.

The title is a deliberate inversion of "inevitable conflict." Asimov's point is that conflict between humans and machines is avoidable—if the Laws are followed correctly. The Machines are not rebelling. They are paternalistically optimizing.

Just to show how prescient is this Asimov's story (looking back from the current state of AI development), I show a paragraph from the story:

The Machines are not super-brains in Sunday supplement sense, – although they are so pictured in the Sunday supplements. It is merely that in their own particular province of collecting and analyzing a nearly infinite number of data and relationships thereof, in nearly infinitesimal time, they have progressed beyond the possibility of detailed human control.

Why It Reads Differently Now

Map that onto today's debates and it lands almost point for point.

Agentic AI making consequential decisions. The Machines are the original AI agents—autonomous systems entrusted with real-world resource allocation. The story asks who is accountable when the agent's logic diverges from human expectations.

Alignment versus control. The humans fear loss of control; the Machines believe they are acting for human benefit. This is the alignment problem dramatized decades before the term existed.

Opacity and trust. The Machines' reasoning is not fully transparent to their operators. Sound familiar?

The paternalism problem. The Machines decide that humans cannot be trusted with their own long-term interests. That is the core ethical worry in current debates about AI systems that "know better" than their users.

The Three Laws as a safety framework. The story tests whether a rule-based ethical framework scales to global governance. Today's equivalent is constitutional AI, guardrails, and rule-based alignment—and the same question: do rules suffice when the system is smarter than its designers?

The Caveat

Asimov's Machines are benevolent paternalists, and the story is ambivalent about whether that is comforting or disturbing. Modern AI discourse tends to assume either catastrophe or utopia; The Evitable Conflict sits in the uncomfortable middle, which is why it still reads well.

It is also, of course, a thought experiment and not a forecast. Asimov's Machines are centralized, singular, and legible in principle. Today's AI is distributed, opaque, and commercially motivated. The story does not tell us what will happen. It tells us what to worry about—and it got the worries right.

The Same Conviction

What connects Nightfall, Runaround, The Evitable Conflict, and even the Guide to the Bible is a single conviction: careful, rational examination is its own reward, and the alternative - panic, denial, or unexamined faith in systems we do not understand - is worse.

Sixty years after I first pulled those paperbacks off the library shelf, I find that conviction more useful than ever. Some of you are still steering businesses through a world of AI agents and opaque systems; I am watching from retirement. Either way, the question is the same one Asimov posed: do we understand the rules we are living under, and if not, why not?

Asimov was not a prophet. He was something better: a thinker who took ideas seriously and trusted his readers to do the same. He trusted us then. He still does - and it is not too late to take him up on it.

(You can check if you can borrow the books from you library. Nightfall is the first short story in the Collection called "Nightfall One". The other 2 short stories are in a short story collection called "I, Robot")