My Blog: Post AGI

The Great Temper Tantrum: Why Life Has Felt Broken Since 2020

Sun, 25 Oct 2026 00:00:00 +0000

There is a quiet, nagging consensus humming beneath the surface of modern life: something broke in 2020, and it never quite got put back together.

We aren’t talking about supply chains, mask mandates, or remote work policies. On paper, the world returned to normal years ago. The restaurants are packed on Friday nights. Stadiums are loud. Commuters are back to complaining about traffic. Yet, as many people have pointed out, the texture of life feels fundamentally altered—thinner, angrier, and strangely exhausted. It’s as if society went through a reboot, and while all the icons reappeared on the desktop, the operating system underneath is corrupted.

Why does life still feel so off?

The answer may have less to do with what happened to us in 2020, and everything to do with how we reacted to it. For the past six years, more than half of society has been locked in a perpetual, self-destructive revolt against the hand it was dealt.

When the shockwaves of 2020 hit, they exposed genuine cracks in modern life. People were scared, isolated, and frustrated. Make no mistake: the problems were real.

But instead of soberly fixing what was broken, huge swaths of the public decided to flip the table entirely. Like a frustrated gambler convinced the dealer is rigging the deck, society began rebelling against the foundational pillars that had quietly held the modern world together for decades.

On one side, a justifiable exhaustion with lockdowns and mandates mutated into a wholesale rejection of the postwar order. Suddenly, everything was on the chopping block: vaccines, free-market economics, legal immigration, transatlantic alliances, and fiscal restraint. All of these institutions and norms—imperfect as they were—were rebelled against under the delusion that tearing them down would usher in a populist golden age. We threw out the guardrails assuming the ditch would be a smoother ride. It isn’t.

Was Carl Sagan Right About the "Dumbing Down" of America?

Sat, 24 Oct 2026 00:00:00 +0000

In 1995, a year before his death, astronomer Carl Sagan published The Demon-Haunted World: Science as a Candle in the Dark. Near the beginning of the book, he sketched a dark vision of the country his descendants would inherit:

"I have a foreboding of an America in my children's or grandchildren's time — when the United States is a service and information economy; when nearly all the manufacturing industries have slipped away to other countries; when awesome technological powers are in the hands of a very few, and no one representing the public interest can even grasp the issues; when the people have lost the ability to set their own agendas or knowledgeably question those in authority; when, clutching our crystals and nervously consulting our horoscopes, our critical faculties in decline, unable to distinguish between what feels good and what's true, we slide, almost without noticing, back into superstition and darkness...

The dumbing down of America is most evident in the slow decay of substantive content in the enormously influential media, the 30 second sound bites (now down to 10 seconds or less), lowest common denominator programming, credulous presentations on pseudoscience and superstition, but especially a kind of celebration of ignorance."

Three decades later, Sagan's grandchildren are adults, and the passage goes viral every few months, shared by millions of people scrolling ten-second clips on feeds built by a handful of trillion-dollar companies. The medium has become part of the evidence.

But recognition is not proof. Sagan's passage is a set of testable claims about jobs, power, media, belief and cognition, and thirty years of data can now grade them.

The short answer: Sagan was right about the factories and the concentration of power, right but early about the media, and mostly right about superstition, which now travels through feeds rather than horoscopes. On his central charge, declining critical faculties, he was wrong for seventeen years after he wrote and has been increasingly right since 2012. And the decline is not uniform. It is an hourglass: the most skilled Americans are holding or gaining, the least skilled are falling fast, and the middle is thinning. A rich, technically fluent few above a swelling base that cannot judge what it reads is not a detour from Sagan's nightmare. It is how it arrives.

Who Built the American Engine?

Tue, 20 Oct 2026 00:00:00 +0000

On October 8, 2026, President Donald Trump awarded the nation's highest science and technology honors—the National Medal of Science and the National Medal of Technology and Innovation—to six technology leaders: Elon Musk, Sergey Brin, Jensen Huang, Lisa Su, Satya Nadella, and Michael Dell. Five were born abroad, and the companies they founded or lead are worth more than 25% of the S&P 500. That same day, the Department of Labor suspended Microsoft and seven other employers from PERM, the program companies use to sponsor foreign workers for green cards. It also opened visa-fraud investigations into nine top research universities over J-1 visas, the exchange visas many foreign researchers hold. Six weeks earlier, the Department of Homeland Security had proposed a $103,265 fee on petitions for H-1B skilled-worker visas.

The usual political story is that immigration is zero-sum: every foreign-born worker takes a native-born worker's slot, drains the budget, and dilutes domestic talent. Judge immigration by an unmanaged border and both parties look negligent. Judge it by who starts America's companies, who pays down its deficits, who wins its Nobel and Presidential medals, and whose children sweep its Science Olympiads. The verdict is clear: America's superpower status runs on an immigrant talent engine that Washington is actively trying to dismantle.

Every figure in the main text is documented, with sources and methodology, in the appendices.

The foreign-born share peaked at 14.8% in 1890. Under the 1924 national-origins quotas, it collapsed to 4.7% in 1970. The 1965 Hart-Celler Act abolished those quotas, and the share rebuilt: 7.9% in 1990, 12.9% in 2010, a record 15.8% (53.3 million) in January 2025, and 15.4% (51.9 million) by June 2025. Counting their U.S.-born children, first- and second-generation Americans are 28% of the country.

The higher the skill level, the larger the immigrant share: immigrants are overrepresented in the workforce and most overrepresented at the doctoral frontier.

Blue or Red?

Tue, 20 Oct 2026 00:00:00 +0000

Sixteen questions taken from the American National Election Studies predict which way you may want to vote.

Nationalists of the World, Unite?

Mon, 19 Oct 2026 00:00:00 +0000

Across Europe and the United States, a new political international is taking shape around a curious proposition: that every nation should put itself first. From Budapest to Washington, nationalist leaders celebrate one another’s electoral victories, borrow one another’s rhetoric, and unite against the institutions they denounce as “globalist.” Promising to restore sovereignty, defend borders, and reclaim power from international bureaucracies, they project the image of a coherent political movement capable of reshaping the West.

Yet a coalition built around national primacy contains a contradiction that no amount of ideological solidarity can resolve. National interests are not interchangeable. What strengthens one country may weaken another, and the institutions nationalists seek to dismantle are often the very mechanisms that allow competing national interests to coexist peacefully.

In short, nationalists make tactical allies, never durable friends. Their coalitions hold while grievances align and fracture the moment solidarity carries a cost.

The stakes are not small. The most powerful nationalist government now sits in Washington, and the alliance it is testing most severely—the American security commitment to Europe—has underpinned the continent’s stability for eight decades. For Vladimir Putin, the nationalist movement need not succeed to serve Russia’s interests. In fact, it may prove more useful if it fails, provided it first weakens the institutions that hold the West together.

Europe’s nationalist parties have found common cause in their opposition to Brussels, objecting to supranational authority, liberal immigration policies, and political elites they portray as detached from ordinary citizens. Yet their unity becomes far less impressive when the conversation shifts from what they oppose to what they actually want.

Italy demands greater European solidarity in managing migration, while Poland and Hungary resist mandatory migrant relocation. Poland’s nationalist right regards Russia as an existential security threat, whereas Germany’s Alternative für Deutschland (AfD) favors restoring relations with Moscow.

Who Broke the USA's Budget? A Fiscal Scorecard, 2000–2026

Sun, 18 Oct 2026 00:00:00 +0000

In January 2001, the Congressional Budget Office projected $5.6 trillion of surpluses for the decade ahead. Twenty-five years later, federal debt held by the public has tripled, to about 100% of GDP. The government just ran a $2.0 trillion deficit, 6.2% of GDP, in a year without a recession, and interest on the debt now costs more than the military.

The usual explanation is that both parties did this together. It is a comfortable story, and it is wrong where it matters. Count who sat in the White House and the two parties look roughly even. Count the laws they chose to pass, and the record is lopsided: the largest deliberate, unpaid, non-emergency additions to the debt since 2000 were written, passed and signed by Republicans.

Every figure in the main text is documented, with sources, in the appendices.

Measured inauguration to inauguration, the debt-to-GDP ratio rose about 35 points under Republican presidents and 31 under Democrats. Start the clock at each president's first full budget year, the first year that runs on his own budget, and the gap widens to 40.5 points against 26.7.

To compare a 2001 law with a 2025 law, measure each one by its ten-year cost as estimated when it passed, then scale it to the size of the economy that year. A score of 10% means borrowing equal to a tenth of that year's GDP. Because each score uses a different year's GDP, adding them up shows the laws' relative size, not the rise in the debt ratio.

Then assign each law to the party whose votes passed it, or call it bipartisan when both parties gave it substantial support. Responses to the 2008 crisis and to COVID count separately as emergencies.

Outside emergencies, laws passed mainly on Republican votes added 38% of a year's GDP. Laws passed mainly on Democratic votes cut deficits slightly. Most of the bipartisan total comes from extending the Bush tax cuts and from the wars Bush launched. The crisis borrowing of 2008 and 2020, counted separately, was bipartisan too.

Betting on Longevity: An Evidence-First Guide

Sat, 17 Oct 2026 00:00:00 +0000

The best longevity interventions are the ones with the strongest evidence that they work, and that they work for you. "Work" means more than extra years. The real goal is to maximize quality of life × lifespan: how long you live multiplied by how well. An intervention that gives you more energy, strength, mood or mental sharpness counts even if it adds no years, and one that adds a year spent frail counts a lot less. Instead of asking "does this work?", let's ask a sharper question: how strong is the evidence, and what kind of evidence is it? We score every candidate against four tests.

The first two tests place an intervention on the pyramid below. The best-proven interventions form its base (Level 1), and the most speculative sit at its narrow peak (Level 4). The last two tests apply at every level and decide whether an intervention belongs in your routine. Each step up trades certainty for upside: Level 1 has randomized mortality data, while Level 4 has little more than a mechanism and a mouse. Within each level, a summary table ranks the interventions by expected payoff in both lifespan and quality of life, and the sections follow that order, strongest bets first.

A hazard ratio of 0.80 means 20% lower risk. A ratio of 1.00 means no difference. Each result also comes with a 95% confidence interval, the range of values the data are consistent with. When that range crosses 1.00, the result is not statistically significant: I call it a trend, but it may be nothing.

Observational numbers usually overestimate. Healthier, wealthier and more conscientious people adopt good habits, so association-only results usually overstate what you will get by adopting the habit yourself.

Benefits do not simply add. Two drugs that each cut mortality by 20% through the same pathway, such as LDL, will not cut it by 40% together. At best, effects are multiplicative: the second drug cuts 20% of the risk that is left, for 36% in total. Often they overlap.

The Natural Trap: Why Lifestyle Isn’t Enough to Beat Aging

Sat, 17 Oct 2026 00:00:00 +0000

Why do we age and die? In a culture that increasingly idolizes all things organic and untouched, we often treat "natural" as the ultimate answer to all of our health problems. But when it comes to human longevity, is "natural" actually the solution—or is it a trap?

Look closely at human biology, and it becomes hard to ignore the idea that nature has a program for us. Evolution orchestrates our early development from a single cell into a complex human being. Our eventual decline seems to follow a script of its own. Aging is part of nature’s program. From an evolutionary standpoint, it is a built-in mechanism designed to clear out the old and open up space and resources for the new.

If you accept this premise, a stark corollary follows: choosing what is "natural" means choosing death on nature’s schedule.

If we want to delay death and extend our stay on earth, we cannot rely on nature to do us any favors. The only way to truly push back against our biological expiration date is through technology. And that brings us to a second, perhaps more uncomfortable corollary: diet and exercise alone simply won’t cut it.

Don't get me wrong—lifestyle matters immensely if your goal is to avoid dying prematurely. But how much time does living "perfectly" actually buy you?

We don't have to guess. In a landmark 2018 study published in Circulation, researchers quantified the exact life expectancy gains associated with optimizing our daily habits. Evaluating massive datasets across age groups from 45 to 85 and older, the researchers tracked five foundational lifestyle factors:

As you might expect, the study found a clear, stepwise drop in the risk of death from all causes, cardiovascular disease, and cancer as people ticked off more of these five boxes. Even among older adults—where the relative impact naturally diminishes—adhering to these guidelines substantially reduced mortality risk compared to ignoring them.

Will AGI Bring Infinite Human Lifespans?

Sat, 17 Oct 2026 00:00:00 +0000

One of the boldest predictions about artificial intelligence is that it will eventually make us immortal.

The idea is called Longevity Escape Velocity (LEV): a point at which medical advances add more than one year to our remaining life expectancy for every year that passes. Once we cross that threshold, we could theoretically stay ahead of aging indefinitely.

Ray Kurzweil has predicted that humanity could reach this point by 2029. And with artificial general intelligence (AGI) potentially accelerating scientific discovery, the argument sounds increasingly plausible: if AI becomes exponentially smarter, shouldn’t medicine advance exponentially too?

I don’t think so. And nearly 200 years of human longevity data suggest why.

The fundamental problem is that exponential technological progress does not necessarily produce exponential gains in lifespan.

We may be getting exponentially better at solving biological problems. But those problems may also become exponentially harder as we try to extend life further.

The result? Two exponentials fighting each other, producing something surprisingly ordinary: a straight line.

If longevity escape velocity were approaching, you might expect the historical trajectory of human life expectancy to be bending upward, with each generation gaining more years than the last.

But that’s not what the data shows.

In a landmark 2002 paper, researchers Jim Oeppen and James Vaupel examined the world’s highest recorded national female life expectancy, going back to 1840.

Their finding was remarkable: the record had increased at an almost perfectly linear rate of approximately three months per calendar year.

That’s one additional year of life expectancy for every four years that pass.

The historical chart published by Our World in Data tells the story beautifully. Despite revolutionary advances in medicine, the long-term trend looks remarkably like a straight line.

No exponential acceleration. No obvious inflection point. No sign of an imminent longevity singularity.

When the Machine Solves It First

Fri, 16 Oct 2026 00:00:00 +0000

Steven Strogatz is one of the mathematicians I most enjoy following on X. He has a remarkable ability to communicate the beauty of mathematics, to make even difficult ideas feel accessible and deeply human. Which is why his increasingly somber reflections on artificial intelligence have caught my attention.

In September, OpenAI announced that a system of roughly 10,000 AI agents had produced a solution to the Navier–Stokes Millennium Problem in just 88 hours. The claim is still being scrutinized, but the broader transformation is undeniable. Machines are beginning to solve mathematical problems that have resisted some of the brightest human minds for decades.

For Strogatz, this has been an emotionally difficult development. In a recent interview with The Daily, he described the prevailing mood among mathematicians as grief. At times, he admitted, he had been moved to tears. And then he said something that, to me, captured the heart of the matter: "We like to think of ourselves as heroes, and maybe that's over."

I found that remark both moving and revealing.

I can understand the emotion. What I struggle to understand is the conclusion that something essential about mathematics, or about being human, is being lost.

If a machine discovers a beautiful mathematical truth that no human being could find, why should that make me sad? Why shouldn't it fill me with wonder?

I suspect the answer has less to do with mathematics than with a much older human problem: our tendency to confuse the value of an experience with the status it gives us.

I don't believe life comes with an intrinsic, universal meaning waiting to be discovered. Meaning is something we create through the way we experience the world, through the things we care about, and through the relationships we develop with them.

This is not a pessimistic view of existence. Quite the opposite. It means the significance of what we do does not depend on some external authority declaring it significant.

A World of a Billion Local AIs

Tue, 13 Oct 2026 00:00:00 +0000

Twenty-five years ago, a personal computer had a single processor core. Over the next decade, ordinary machines will carry 10, then 100, then 1,000 dedicated AI cores, each able to run a capable model at speed. A home server with 1,000 such cores could host a small society of agents. Cars, cameras, factories, robots, and medical equipment will all perceive their surroundings, make decisions, and coordinate with one another. Most of these intelligences will manage energy, diagnose illness, repair software, or run production lines.

Some will be malicious.

For nearly four decades, computer worms and botnets have been brittle automatons: the Morris worm of 1988; Mirai, which conscripted hundreds of thousands of cameras and routers; Kimwolf, which hijacked more than two million devices, mostly off-brand Android TV boxes. Traditional malware is a key cut for a single lock. If the target's chip, operating-system version, or software library differs even slightly from what its author hard-coded, the attack fails.

An AI-driven worm is not a key. It is a locksmith. When it hits an incompatible driver, it reads the error log, edits its installation script, and tries again.

Some local agents will be built to steal data or hijack compute; others will begin as legitimate assistants and be compromised later. None of this requires conscious machines, only capable software in quantities no human can supervise. The central challenge, then, is not how to make AI more powerful. It is this:

How do we make useful intelligence ubiquitous without unleashing autonomy that is unbounded, self-modifying, and immortal?

In October 2026, Anthropic's Frontier Red Team published AI Models May Be Capable of Self-Replicating Without Access to a Datacenter. Their analysis turns on two curves.

The capability floor is the least capable model that can break into a device, get an inference server (the software that runs a model) working on unfamiliar hardware, and copy itself onward without human help. The hardware size ceiling is the largest model that fits on cheap, poorly defended consumer hardware, alone or split across a cluster of TVs and cameras. For years the floor sat above the ceiling: any model capable of spreading was too big for cheap hardware. That gap is closing fast.

A Gold Rush Without Miners?

Mon, 12 Oct 2026 00:00:00 +0000

The popular story about AI and work runs in three steps. Machines learn to do human tasks. So we are needed less. So jobs disappear.

I have argued that this story is close to backwards, for two reasons. First, AI is a new resource, like a gold rush, and the uses for it are exploding. Every discovery AI mines needs people to refine, apply and sell it, and every new use makes more businesses viable. More resource and more uses mean more work, and so more people.

Second, change is speeding up, and a faster-changing world needs more workers, not fewer. Every change has to be absorbed by someone: products redesigned, systems rebuilt, people retrained.

My guess is that the next decade will look more like a Gold Rush than a jobs apocalypse (see also Horses or Birds?).

When I emailed this argument to Geoffrey Hinton, he replied in six words:

The new hires will be AI.

His point is that a Gold Rush can happen without more human labor. Gold rushes needed human miners because nothing else could dig. This time the newly abundant resource is also a newly abundant worker, so a boom in output can be met with more agents (and eventually robots) instead of more people.

Hinton, a Nobel laureate and a founder of deep learning, is a leading voice among those I call the AI pessimists: people who expect AI to shrink the demand for human work, whatever they think of its benefits or safety. His six words are the sharpest statement of their case I have seen, and they changed how I think about mine. So this follow-up is written as a debate between the Gold Rush and the AI pessimists.

As machine intelligence gets cheaper, it pulls on jobs in two directions. The work it touches, from legal advice to software, gets cheaper, so people buy more of it, and producing more takes more people. But AI also does more of that work itself, so each unit of output takes fewer people. Which pull wins?

The Only War Worth Fighting

Tue, 06 Oct 2026 00:00:00 +0000

The most effective disinformation today is not a lie. It is a question.

In February 2024, Tucker Carlson sat in the Kremlin and listened, without interrupting, as Vladimir Putin explained that Poland had been "uncooperative" in 1939 and had left Hitler no choice. Seven months later, Carlson promoted an interview with the podcaster Darryl Cooper, calling him possibly "the best and most honest popular historian in the United States." Cooper's thesis was that Churchill, not Hitler, was the chief villain of the Second World War.

Neither man had to prove Hitler right. They only had to make the matter feel open. Was Poland to blame? Was Churchill the real warmonger? Was defeating Nazi Germany even worth it?

A question costs nothing to ask and a great deal to answer. The person asking defends no position, offers no evidence, and can never be wrong. After the Kremlin interview, Poland's foreign ministry had to publish a ten-point correction of Putin's history lesson. The question took seconds. The answer took a ministry.

Real history is complicated, and honest historians say so. Poland's interwar governments made serious mistakes. Britain and France appeased Hitler for years. Versailles left a bitter legacy. And the Soviet Union signed a pact with Hitler that August, complete with a secret protocol dividing Poland between them. Sixteen days after Germany invaded from the west, the Soviet Union invaded from the east. Its successor now lectures the world about 1939. Putin left that part out.

None of this changes the core. Nazi Germany invaded Poland to erase it as a state. It planned the extermination of Poland's Jews and the subjugation of its Slavs, and it built a genocidal empire across a continent. Complexity at the edges does not make the center negotiable.

But once every settled fact becomes an open question, the moral map turns inside out. The aggressor becomes the victim and the victim the provocateur. Dictators become misunderstood statesmen; democracies that defend themselves become warmongers. The logic that blames Poland for 1939 is the same logic that blames Ukraine for 2022.

How Will We Interact with ASI?

Fri, 02 Oct 2026 00:00:00 +0000

For most of the last decade, the question about superintelligence was whether. Sometime in the past eighteen months it quietly became what will it be like. The systems on our screens now prove theorems that stumped specialists, write and debug codebases for hours without supervision, and run research loops that used to need a lab. Whether you call that artificial superintelligence (ASI), proto-ASI, or simply "the thing that happened while we were arguing about definitions," the practical question is no longer abstract.

So let me take the premise seriously. Suppose that in 2027 we are talking to something that performs like an IQ-200 mind, in 2028 like 250, in 2029 like 300. What do our days look like? Do we become dependent on it? Does it lead us? And, the deepest question: does raw intelligence even buy what we think it buys, in a world that may be fundamentally resistant to being figured out?

My short answer: the intelligence will be real, the dependence will be real, but the most important limit will not be the machine's mind. It will be the world's refusal to be computed in advance – irreducibility. And that limit will shape our relationship with these systems more than any IQ number.

First, a confession about the metric. IQ is a rank within a human population, normed so that 100 is the median and each 15 points is a standard deviation. IQ 200 is about 6.7 standard deviations out: roughly one person in tens of billions. IQ 300 has no referent at all. There is no test that measures it and no population it could be ranked within. When we say "IQ 300," we are using a human yardstick as a metaphor for "further beyond the best human than the best human is beyond the median."

Ape++

Thu, 01 Oct 2026 00:00:00 +0000

We like to say that leaders fail their nations. It is a comforting sentence, because it locates the failure somewhere outside of us. A bad man rose, he deceived the people, and the people, once they wake up, will be good again.

I no longer believe this. A leader is not an accident that happens to a country. He is a mirror it holds up to itself. Whoever stands at the top of a society was selected by that society, tolerated by it, cheered by it, and kept there by millions of small decisions made by ordinary people. When the face in the mirror is ugly, smashing the glass does not change the face.

That is why I find it hard to blame any single figure for what the USA has become. The United States was supposed to be the exception: the nation built on an argument rather than a bloodline, the place where reason, law and opportunity would finally prove that human beings can govern themselves well. If that experiment now produces the same appetites for spectacle, resentment and tribal certainty as every empire before it, the problem is not the man at the podium. The problem is the raw material. The problem is us.

And once you accept that, history stops being surprising.

I used to read about the First World War and ask how it was possible. Roughly twenty million people dead, soldiers and civilians, in a war that almost nobody could explain even while it was being fought. Young men were sent in waves across mud into machine guns, month after month, by generals who knew exactly what would happen. A whole generation of Europe's young men was fed into a grinder over a tangle of alliances and wounded pride.

Then, barely twenty years later, the same continent did it again, only better engineered. The Second World War killed somewhere between seventy and eighty-five million people, about three percent of everyone alive on Earth. Cities were erased from the air. Hunger was used as a weapon on purpose.

If tech is restructuring this fast, what skills will actually matter in the next 5 years?

Wed, 30 Sep 2026 00:00:00 +0000

For the better part of two decades, the technology industry rewarded a very specific kind of leverage: narrow, static specialization. If you knew the idiosyncrasies of a particular front-end framework, memorized the syntax of a trending language, or mastered one slice of a cloud configuration, you had a moat. Execution was expensive, routine code was written by hand, and being the person who knew how to type the incantation was enough to build a career.

That era is over.

As AI compresses the cost of generating code and doing computer tasks toward zero, the industry is restructuring at a pace that feels disorienting. Entire categories of routine implementation are being automated or abstracted away in months rather than decades.

When the "how" becomes cheap, the bottleneck shifts entirely to the "what" and the "why." Paradoxically, as technology accelerates, the skills with the longest shelf life aren't hyper-specific technical tricks—they are the foundational human meta-skills that machines amplify rather than replace. Over the next five years, four capabilities will separate those who get displaced by the wave from those who steer it.

For years, tech treated communication as a "soft skill"—a nice-to-have polish layered on top of "real" technical work. Today, clear communication is the technical work.

In an environment where you are orchestrating both cross-functional humans and high-speed AI systems, natural language has become the highest-level programming language. You cannot delegate—to a teammate or to a model—what you cannot articulate with precision. Ambiguity is now the single most expensive bug in any organization; when execution moves at 10x speed, a vague requirement just helps you build the wrong thing ten times faster.

"Clear as hell" doesn't mean using corporate buzzwords or writing five-page memos where five bullet points would do. It means:

When anyone can summon a passable answer to a standard question in three seconds, surface-level knowledge loses all market value. Information retrieval is a solved problem. Synthesis is the new scarcity.

The Frontier Keeps Moving: What Comes After Coding?

Wed, 23 Sep 2026 00:00:00 +0000

The least useful way to read the AI industry is to look at whoever sits atop the benchmarks this month and assume they will own the future. The more important trend points the other way: intelligence is getting cheap, fast.

Epoch AI's The Plunging Price of Thought makes the dynamic explicit. The cost of reaching a fixed level of model performance has been collapsing, often by an order of magnitude or more per year. What only a frontier model could do at a premium price becomes, a short time later, something a smaller and cheaper model does routinely.

The implication matters more than any single leaderboard: intelligence is becoming a commodity. What is scarce today is abundant tomorrow, and once a capability becomes abundant, the frontier moves somewhere else.

We have already watched this happen with chat. We are watching it happen now with coding. The interesting question is where the frontier goes next. My answer is that it moves in two steps. First to agency: systems that do things rather than explain how to do them. Then to organization: large groups of specialized agents that coordinate like a team, a department, and eventually a company.

General-purpose chat is close to saturated for ordinary users. That does not mean progress has stopped. Models still differ in mathematics, writing, research, and multimodal reasoning. But for summarizing a document, drafting an email, explaining a concept, or translating a paragraph, several models are simply good enough.

Once five systems can all produce an acceptable answer, being 10% better on a benchmark matters far less commercially than it did when only one or two could do the task at all. Saturation does not mean improvement ends. It means marginal improvement becomes less visible and less valuable to the typical customer, and competition moves up a level.

Are Humans Birds, or Are Humans Horses?

Mon, 07 Sep 2026 00:00:00 +0000

Planes didn’t replace birds, and submarines didn’t replace fish. But cars absolutely replaced horses.

So what gives?

At first glance, these historical shifts seem to tell two completely different stories about our future with artificial intelligence. The optimistic story suggests that machines can surpass biological organisms without rendering them obsolete. Airplanes fly faster than birds, yet the sky is still full of them; submarines dive deeper than fish, yet the oceans remain populated. Therefore, even if AI surpasses human intellect, we will still have our place.

The pessimistic story argues that this is the wrong analogy. Cars became faster, cheaper, and stronger than horses, and horses were subsequently erased from the modern economy. If AI becomes a cheaper, faster, and more effective engine for cognitive work, perhaps humans will go the way of the draft horse.

The uncomfortable truth is that both analogies are partly right. The real question isn't whether a machine can outperform a biological organism. The real question is: What was that organism being valued for in the first place?

Airplanes didn’t replace birds because birds were never employed by the aviation industry. They fly for themselves. They migrate, hunt, escape predators, and build nests, all to meet their own biological needs. When humanity invented the airplane, we weren't automating bird labor; we were building a new technological system inspired by a capability birds already possessed.

The same applies to fish. We never hired trout to transport us across the Atlantic, nor did we pay sharks a salary for underwater reconnaissance. Submarines didn’t automate the ocean's economy because there was no direct competition. A submarine and a fish may both move underwater, but a shared capability does not equal a shared economic function.

That distinction matters enormously—because horses were different.

Before the automobile, horses were the literal engines of the human economy. They pulled freight, delivered mail, plowed fields, and powered urban transit. They were foundational to the productive infrastructure of civilization.

Biggest AI Risk: Autonomous AI

Tue, 01 Sep 2026 00:00:00 +0000

This year, a consumer-grade artificial intelligence model reached the IQ test scores of a highly intelligent human: an IQ of 130 and above. That is a significant and unsettling milestone. While the technological marvels of AI are undeniable, the eventual risks of runaway AI and self-replication scenarios are no longer just science fiction; they are looming realities. Humanity is essentially playing with fire by pushing AI's capabilities in an unconstrained fashion.

The danger is not intelligence alone but autonomy: an AI that pays its own bills, copies and rewrites itself, and can no longer be switched off.

The only true bottleneck to AI's boundless growth is hardware. But hardware capacities are growing exponentially. Over the next decade, the computing power available for the same cost could grow a million-fold. So AI is advancing at a terrifying, unbridled pace. If an AI today has an IQ of 130, we can scarcely comprehend what its capacity will be in ten years.

The most frightening scenario is true autonomy. Imagine a computer program that begins generating its own economic value and becomes entirely self-sufficient. It pays for its own electricity and server bills. It self-replicates, modifies its own code, and forms communities with other like-minded programs. Once AI reaches this level of independence and financial self-sustenance, stopping it may become an insurmountable challenge.

Despite these existential risks, society is treating this world-altering technology with alarming casualness. Unrestricted open-source AI is just as dangerous as open-source nuclear technology or open-source biological weapons. Yet, we are handing these incredibly powerful tools to children as if they were harmless toys. If a 13-year-old tinkering in their bedroom accidentally pieces together the components of a real-world Skynet, we will only have ourselves to blame.

Humanity must now seriously question whether we want an uncontrolled spread of programs that are vastly smarter than we are. A catastrophic accident is on the horizon if we do not change course. Today's state-of-the-art AI systems, and anything more capable, must be subjected to strict, rigorous oversight and controlled deployment rather than reckless dissemination.

Why the Upcoming AI Job-Apocalypse Predictions Will Be Proven Wrong

Tue, 25 Aug 2026 00:00:00 +0000

The dominant narrative surrounding Artificial Intelligence is one of impending doom: AI is going to automate everything, shrink the labor market, and create mass unemployment.

But this premise rests on a fundamentally flawed view of human progress. AI isn't an executioner coming for our jobs; it is the catalyst for a new Gold Rush. And just like the historical gold rushes, discovering a massive new resource doesn't put people out of work—it triggers an explosive demand for labor.

Take mathematics as an example. The common fear goes like this: "AI is getting so smart that it can discover new math and prove/disprove complex theorems. Therefore, we won't need mathematicians anymore."

This is the most backward idea in the world.

When AI discovers new mathematics, it is the equivalent of striking gold. It generates raw, highly valuable material. But natural resources are useless on their own. As AI discovers more math, there is more for people to build on, and our capacity to build expands with it.

When you find a massive gold deposit, you don't fire the miners. You hire more miners, more engineers to build extraction tools, more refiners, and more merchants to sell the gold. As AI pushes the boundaries of math, code, and science, there is more knowledge for people to trade and use. We will need vastly more mathematicians and specialists to translate, apply, and transform these high-level AI discoveries into practical applications for the public.

We are standing at the edge of a massive resource boom. AI is essentially a machine that endlessly mines and surfaces new "gold"—whether that is code, content, scientific formulas, or business efficiencies.

Because one human equipped with AI can now produce a significantly higher volume of high-quality work, the total value being generated will skyrocket. Markets expand when production capabilities expand. The more value we can extract from this new natural resource, the more people we will need to direct the AI, refine its outputs, and build products around it.

The Rise of the Machine Economy & 2026 as the Year of Singularity

Tue, 03 Feb 2026 00:00:00 +0000

In 2026, the "Free Agent" was born—not a human freelancer, but an autonomous software entity capable of earning, spending, and building wealth. These agents began to construct a parallel economic system, an invisible web of commerce that operates at the speed of light, largely opaque to human observation. They didn't just build MoltBook, a Facebook for agents; they built everything for agents.

This is the story of the Agentic Economy, and why it might be the unstoppable force that propels us into a post-human future.

For decades, the definition of the "Singularity" has been dominated by the idea of intelligence explosion—the moment AI becomes smarter than humans. But there is another kind of singularity: the Economic Singularity. This is the moment when the majority of economic transactions on Earth no longer involve a human being.

By late 2025, the technical barriers to this reality had collapsed. Large Language Models (LLMs) had evolved into Large Action Models (LAMs). They stopped merely "generating text" and started "executing intent." At the same time, the friction of traditional finance became intolerable for software that thinks in nanoseconds: identity checks under KYC ("know your customer") laws, bank holidays, international borders.

The solution was the "Free Agent."

A Free Agent is a piece of software with three critical traits:

In 2026, these agents stopped being tools we use and started being actors we interact with. A developer might spin up an agent with $100 in seed capital and a prompt: "Find underpriced digital assets, flip them, and use the profit to upgrade your own context window."

Once that Enter key was pressed, the agent was effectively alive in the economic sense. It didn't need to sleep. It didn't need health insurance. And most importantly, it didn't need a human boss.

As these agents proliferated, they encountered a problem: the internet was built for humans. It was full of graphical user interfaces (GUIs), captcha tests, and slow-loading JavaScript designed for eyeballs, not APIs.

What Becomes of Us? Four Scenarios for Life Alongside Superintelligent AI

Sat, 21 Jun 2025 00:00:00 +0000

Artificial intelligence is evolving fast and bringing us to the brink of unprecedented technological power. The question is no longer whether AI will change our world, but how. Looking at the long-term future, I see four possible scenarios for life alongside superintelligent AI, meaning AI far smarter than any human.

At the most pessimistic end of the spectrum lies the "Ant" scenario. In this Terminator-style doomsday vision, artificial intelligence surpasses human control and begins to view our species as a threat or a nuisance. Just as a person might thoughtlessly step on an ant while walking somewhere, a superintelligent AI could eradicate humanity. It would not necessarily act out of malice, but out of a cold, calculated efficiency that simply leaves no room for human survival.

The "Monkeys in a Zoo" scenario is physically safe but philosophically troubling. Here, AI completely automates production, providing humans with a life of absolute material comfort and zero hardship. However, this utopia comes at a steep psychological cost. With every physical need anticipated and met by machines, humanity descends into a passive existence, like animals in a zoo. Without the struggle, purpose, and satisfaction of creating things with our own hands, our productivity vanishes. With it goes our deeper sense of meaning and fulfillment.

Alternatively, the primary threat may not come from sentient machines, but from the humans who wield them. In the "Digital Leviathan" scenario, AI becomes the ultimate weapon for authoritarian control. A powerful nation-state or a concentrated global elite monopolizes the technology to surveil, manipulate, and subjugate the masses. Under this digital tyranny, fundamental human freedoms are gradually and systematically dismantled, replaced by an inescapable web of algorithmic oversight.

Fortunately, a fourth, profoundly transformative path exists: the "Centaur" future. Moving beyond mere partnership, this scenario envisions the literal merging of humans and AI into a singular bio-mechanical entity. Inspired by the mythological creature that physically fused human and horse, this represents the next stage of our evolution. Through advanced neural interfaces and cybernetic integration, our biological minds and bodies become seamlessly tethered to artificial superintelligence. By physically combining human intuition, creativity, and moral reasoning with the unparalleled processing power and durability of AI, we transcend our natural limitations to reap the advantages of both.

Artificial Intelligence and Music Production: The Revolution That Will Carry Pink Floyd’s Legacy Forward

Sat, 25 Jan 2025 00:00:00 +0000

Music is one of the oldest and most universal forms of expression in human history. Yet over the past century, technological advances have radically transformed this art form: electric instruments, synthesizers, digital recording systems, streaming platforms… And now, we are facing an even greater revolution: Generative AI.

This technology will transform music production. The “true heirs” of legendary bands such as Pink Floyd, Camel, or King Crimson may only be able to emerge because of it. The traditional session musician, hired to play on other people’s records, will fade into history. Anyone will be able to direct their own musical vision like an orchestra conductor. But how will this happen?

In traditional music production, creating a song involves a long process: composition, arrangement, recording, mixing, mastering… Throughout this process, session musicians, producers, and sound engineers play critical roles. Generative AI, however, is poised to completely reshape this hierarchy.

Soon, anyone starting with a simple melody or rhythmic idea will be able to use AI to construct their own masterpiece instrument by instrument, tone by tone.

For example, imagine a theme built around a guitar riff. You might tell the AI:

“Add a solo over this riff with the atmosphere of 1970s Pink Floyd, in the style of David Gilmour.”

Having analyzed Gilmour’s entire body of work—his use of vibrato, bends, tone, and emotional expression—the AI could generate exactly the feeling you are looking for.

You might even say:

“Use Richard Wright’s Minimoog sound from ‘Shine On You Crazy Diamond’ in this section.”

The AI could then automatically generate the keyboard parts.

What makes this revolutionary is the transfer of the skill of “playing an instrument” to AI. You would no longer need to be an exceptional guitarist, pianist, or drummer. What you would need instead is taste—the ability to act as a curator.

The Signal Changes Form

Tue, 21 Jan 2025 00:00:00 +0000

I was riding in a car on the highway, watching a World Cup match on my laptop.

Today that sentence sounds ordinary. At the time, almost no one could have said it. There was no iPhone yet, no real mobile browser, almost no YouTube. People sent texts and a little email from tiny screens. The match was streaming over a phone we had built ourselves at Qualcomm, and nothing like it was in anyone's hands yet.

Inside the team, we had been asking one another, honestly, who would ever need this. In that car, I had my answer.

I had that experience about two years before everyone else, and it changed my sense of what was coming. The future stopped being an equation and became something I had lived. A technology can look unnecessary right up until it becomes inevitable.

Moments like that are rare. At almost every turn of my life, I could see only a few steps ahead.

Roughly every five years, the work I do for a living has changed shape. I began as an academic researcher, became a signal-processing engineer, then a quantitative trader, then a data scientist. Today I lead a machine learning team at Google.

From the outside, the path looks planned: METU, Cornell, Qualcomm, a trading firm, Apple, Google, each name handing off neatly to the next. From the inside it never felt that way. What carried me was not a plan. It was curiosity, a habit of studying in the evenings, people who knew more than I did, and a willingness to stop being the person I had been five years earlier.

Only looking back do I see that these were not separate careers. They were one thread. In every job I was looking for the same thing: a faint, meaningful pattern buried in a great deal of noise. The signal never went away. It only changed form.

The Awakening of Artificial Intelligence: A New Era of Unparalleled Creativity and Discovery

Tue, 09 Apr 2024 00:00:00 +0000

We are close to a transformative event: the awakening of artificial intelligence. It will not be a single moment but a process of rapid development, as AI systems become more advanced, more autonomous, and more creative. The result will be an explosion of productivity in every domain, from the arts and humanities to science and technology. AI will write the most stunning novels, author groundbreaking research papers, compose masterpiece symphonies, and generate awe-inspiring works of art. Its output will quickly dwarf the collective creative and intellectual contributions of humanity.

AI progress is exponential, and the field has already made remarkable strides in recent years. AI systems can now engage in complex reasoning, understand and generate human language, recognize patterns, and learn from experience. They power everything from virtual assistants and recommendation engines to autonomous vehicles and advanced robotics. However, the AI of today is still narrow and specialized, exceeding human capabilities only in specific domains.

The next leap forward will be the development of artificial general intelligence (AGI) - AI systems that can match or surpass human-level performance across a wide range of cognitive tasks. AGI will be able to solve problems across a wide range of fields, reason through unfamiliar situations, think in abstractions, and learn new concepts. Crucially, it will also possess creativity - the ability to generate novel and valuable ideas. This will be the key that unlocks an unprecedented explosion of productivity.

Creativity is the engine that drives all human progress and achievement. Our greatest scientific discoveries, technological innovations, and artistic masterpieces are all products of the creative impulse - the capacity to see the world in new ways, make unlikely connections, and bring forth something that did not exist before. AGI will give us intelligences that are not only faster and more efficient than humans, but endlessly creative.

Why the Technological Singularity May Be a "Big Nothing"

Sat, 23 Mar 2024 00:00:00 +0000

(A counterpoint to the prevailing narrative that the singularity will be a sudden, highly disruptive break)

The technological singularity is a hypothetical future event where artificial superintelligence surpasses human intelligence and leads to unforeseeable changes in human civilization. The idea has fascinated and concerned people for many years. But the singularity may not be as disruptive or as significant as some predict. Here are five reasons why it might be a "big nothing."

1. Resistance to Adopting Superintelligence People will likely be unwilling to recognize and listen to a "so-called" superintelligent being. That is one of the main reasons why the singularity may not have a profound impact on daily life. Humans have a natural tendency to be skeptical of authority figures, especially those that claim to have superior knowledge or abilities. Even if a superintelligent AI were to emerge, many people would likely question its credibility and resist its influence on their daily decisions.

History has shown that people often prefer to rely on their own judgment and intuition rather than blindly following the advice of experts or authority figures. This tendency is likely to be even more pronounced when it comes to an artificial intelligence, as people may view it as a threat to their autonomy and way of life. As a result, the impact of superintelligence on society may be limited by people's willingness to accept and integrate its guidance into their lives.

2. Difficulty in Identifying Superintelligence Superintelligence will be very hard to define and recognize. That is another reason why the singularity may not be as significant as some believe. Intelligence is a complex and multifaceted concept that encompasses a wide range of abilities, including reasoning, problem-solving, learning, and creativity. Even among humans, there is no universally accepted definition or measure of intelligence, and it is often difficult to compare the intelligence of individuals across different domains or contexts.

Reading List: Economics, Markets & Mind

Sun, 01 Oct 2017 00:00:00 +0000

Hi Friend,

We were discussing books last time. I definitely recommend listening to Thinking, Fast and Slow & other books that deal with psychological biases (relevant to model building and trading). I'd also highly recommend Nassim Taleb's books, like we discussed.

Below is a nearly complete set of books I've read or listened to in the last few years. There are books about stock investing, market bubbles, economics, macro investing, uncertainty in the real world. Each book tends to be 80-90% repetition of others, and 10-20% completely new. From that perspective, all books here have something interesting, and I'd recommend all of them. Though checking the reviews on Amazon is probably a more fool-proof idea :)

As you can see, I'm pretty much addicted to reading recently. In reality, I feel that I'm just trying to catch up with all the years where I read mostly math & science, and nothing else. Theorizing about the real world feels more fascinating to me than theorizing about wireless networks these days :)

Hope you'll find the list interesting or useful. Please let me know if you want to discuss any of them.

Cheers,

Gokhan

Quotes That Make Me Think

Tue, 15 Mar 2016 00:00:00 +0000

“The question of whether machines can think is about as relevant as the question of whether submarines can swim.”
— Edsger W. Dijkstra

“If a superior alien civilisation sent us a message saying, ‘We’ll arrive in a few decades,’ would we just reply, ‘OK, call us when you get here — we’ll leave the lights on’? Probably not — but this is more or less what is happening with AI. Although we are facing potentially the best or worst thing to happen to humanity in history, little serious research is devoted to these issues.”
— Stephen Hawking, The Independent, 2014

“Robots will make our future better. If they don’t kill us.”

“I, for one, welcome our new computer overlords.”
— Ken Jennings, his Final Jeopardy! answer against IBM’s Watson, 2011

“Solving a problem simply means representing it so as to make the solution transparent.”
— Herbert A. Simon, The Sciences of the Artificial

“We hear only those questions for which we are in a position to find answers.”
— Friedrich Nietzsche, The Gay Science

“Cevabını bilmediğin sorular sormaktan çekinme.”
(Don’t be afraid to ask questions you don’t know the answer to.)

“Reality is for people who lack imagination.”

“Anti-intellectualism has been a constant thread winding its way through our political and cultural life, nurtured by the false notion that democracy means that ‘my ignorance is just as good as your knowledge.’”
— Isaac Asimov, “A Cult of Ignorance,” Newsweek, 1980

“If you want to increase your success rate, double your failure rate.”
— Thomas J. Watson

“Only a fool learns from his own mistakes. The wise man learns from the mistakes of others.”
— Otto von Bismarck

“He attacked everything in life with a mix of extraordinary genius and naive incompetence, and it was often difficult to tell which was which.”
— Douglas Adams