For three decades, the outsourcing model was treated as settled economics. Why pay a domestic worker $55,000 a year to handle data entry, customer support, content production, or basic research when you could pay a fraction of that to a team overseas? The math seemed obvious. The logic seemed unassailable.

It wasn’t.

AI has just changed the equation entirely, and most companies haven’t noticed yet.

The Arbitrage That Wasn’t What It Looked Like

Outsourcing was sold as a simple cost play. But anyone who has managed an outsourced workflow knows the real ledger looks nothing like the invoice.

There’s the time zone tax. The revision cycles. The lost context every time a project changes hands. The intellectual property risk when your business logic lives on servers you don’t control. The management overhead required to coordinate across language barriers and cultural frameworks. The quality control layer you have to build just to catch what falls through the cracks.

The offshore model didn’t eliminate labor costs. It traded visible labor costs for invisible friction costs. Companies got comfortable with the arrangement because the friction was diffuse and hard to quantify, while the invoice savings were easy to point to in a budget meeting.

AI is making that friction visible by eliminating it altogether.

The Flip

Here is the core insight worth sitting with: the jobs that were most attractive to outsource were the entry-level, high-volume, repeatable tasks. Data processing. Content drafts. Research summaries. Customer-facing correspondence. Basic coding. Formatting. Coordination.

Those are precisely the jobs that AI handles most effectively.

The offshore model exploited the gap between what those tasks cost to do domestically versus abroad. AI closes that gap from the other direction entirely. It doesn’t lower the wage. It multiplies the output of the person being paid.

One AI-fluent worker in the same time zone, with full context, full cultural alignment, and full accountability, can now produce the output of a small offshore team. Not because the technology is magic, but because the friction is gone. There is no handoff delay. No interpretation gap. No revision email chain. No meeting to re-explain what was already explained.

The work flows.

What “AI-Fluent” Actually Means

This is where the conversation usually goes sideways. People hear “AI can replace outsourced workers” and they imagine a button that does the work automatically. That’s not what’s happening.

What’s happening is more interesting, and more demanding.

An AI-fluent worker isn’t someone who handed their job to a chatbot. They’re someone who understands how to use AI as a precision tool, when to trust its output, when to question it, how to prompt for what they actually need, and how to layer their own judgment on top of what the machine produces.

That last part matters more than most people realize. AI generates fluently. It does not judge wisely. The person holding the tool still has to know what good looks like, what the client actually needs, what the company’s voice is, and when the output is technically correct but strategically wrong.

The machine generates. The thinker decides.

That’s not a slogan. It’s the actual job description of the AI-fluent worker. And it is a skill set, not a given. The people who develop it are becoming genuinely difficult to replace at any price point.

The Entry-Level Paradox

There’s a harder edge to this conversation that deserves honesty.

Outsourcing didn’t just shift jobs overseas. It gutted the entry-level domestic pipeline. Those first jobs, tedious as they were, were also where people learned. They were the rungs on the ladder. Show up, do the volume work, absorb how the business operates, build judgment through repetition, earn your way up.

That pipeline is broken in a different way now. AI handles the volume work, which means the rung looks different. You can’t earn your judgment by doing a thousand data entry tasks if AI does them automatically. The learning path has to be redesigned from the ground up.

The companies and workers who figure that out early will have a significant advantage. The ones who assume AI-fluency is a trait people just acquire on their own will find themselves with a workforce that knows how to operate the tool but not how to think behind it.

This is not a small problem. It’s a structural one. And it’s worth far more attention than it’s currently getting.

The New Math

Let’s do the rough arithmetic, with the caveat that real numbers vary by industry and role. [Speculation: illustrative comparison only.]

An outsourced team of five handling content, research, customer support, and basic data tasks might run $60,000 to $90,000 annually when you account for management overhead, rework costs, quality control, and coordination tools. Output is solid on routine tasks but drops on anything requiring contextual judgment or speed of iteration.

One AI-fluent domestic worker, paid at a fair market rate, using current AI tooling, can now handle comparable or greater volume on those same task types, with better contextual accuracy, faster turnaround, full time zone alignment, and zero coordination overhead.

The math has flipped. Not everywhere, not for every role. But at the entry-to-mid tier where most outsourcing was concentrated? The case for offshore is eroding quickly.

This Is a Re-Shoring Story, Just Not the One You Expected

The political conversation about re-shoring focuses on factories, supply chains, and manufacturing. That’s real and important. But there’s a quieter re-shoring happening in knowledge work, and it’s being driven not by policy or tariffs but by tools.

The individual is back in play.

Not the individual as a cog in a large corporate machine, and not the individual as a freelancer racing to the bottom on a gig platform. The individual as a genuinely capable, fully amplified professional who can operate at a level that used to require a team.

That’s a different kind of economic story. It doesn’t fit cleanly on a campaign poster, but it’s unfolding in real time in offices and home studios and coffee shops everywhere that people have figured out that the tool, used well, changes what one person can actually accomplish.

What This Means for Companies Still Running the Old Playbook

If your competitive advantage has been built on labor cost arbitrage, you are now competing against individuals who have no arbitrage to offer but don’t need it.

The person who knows how to use AI well is not cheaper than your offshore team. They’re more capable, faster to iterate, easier to manage, and fully accountable for the output in a way that a distributed offshore workflow simply cannot replicate.

The race to the bottom just got lapped by the race to the top.

Companies that see this early will restructure their workflows around fewer, more capable people with real AI fluency. Companies that don’t will keep optimizing a model that the market is quietly rendering obsolete.

The script has flipped. The only question is who reads it first.

There’s a moment in every major technological shift when adoption becomes dependency. You don’t see it happening. Nobody announces it. One day you’re choosing to use something; the next day, not using it would cost more than you’re willing to pay. The choice didn’t vanish overnight. It clicked away, one notch at a time.

Economists and systems researchers have a name for this. They call it the ratchet effect. It mostly lives in academic papers. It’s time to talk about it plainly, because we’re living through it right now.

A ratchet turns freely in one direction. Reversing it requires deliberately releasing the mechanism. Society almost never makes that effort in time. We adopt something widely enough that the infrastructure, the economy, the habits, and the expectations all reconfigure around it. Then going back isn’t just inconvenient. It’s catastrophically disruptive. So we don’t go back. We call it progress and keep turning the wheel.

The academic case is solid. It just never left the ivory tower.

Economists have studied this for decades under the label of path dependency, the idea that early choices constrain later options in ways that compound over time. The QWERTY keyboard layout is the textbook example, and its real history is even better than the myth.

Most people believe it was designed to slow typists down to prevent key jams. That’s largely false. The actual story involves telegraph operators who found alphabetical keyboards useless for translating Morse code, mechanical engineers separating letter pairs that caused typebars to collide, and Remington’s sales team tweaking the final layout to get around competitors’ patents. Nobody sat down and designed QWERTY for the world we live in now. It accumulated its shape through a series of unrelated decisions, got commercialized by the right company at the right moment, and became the standard. A hundred and fifty years later, we’re still using it. Not because it’s optimal. Because switching costs grew larger than the benefit of switching.

That’s the ratchet. No conspiracy. No master plan. Just a series of locally rational decisions that compounded into something nobody would have chosen on purpose.

Political theorist Langdon Winner captured the deeper pattern in his 1980 essay “Do Artifacts Have Politics?” Technologies aren’t neutral tools, he argued. They embed values and expectations that persist long after the original choice is forgotten. The technology shapes behavior. The behavior shapes institutions. The institutions resist change. What began as a decision calcifies into an assumption. By the time anyone thinks to question it, it has become the water everyone is swimming in.

That’s the harder version of the ratchet problem. It’s not just that reversal becomes too costly. It’s that the original choice eventually stops being visible as a choice at all.

Smartphones are the case study everyone lived through.

Think back to 2007. The iPhone launched. You could buy one or not. That was a real choice. By 2015, not having a smartphone meant opting out of social invitations, business workflows, banking apps, navigation, and two-factor authentication. By 2020, the pandemic-era infrastructure assumed everyone had one. Today, the phrase “I don’t have a smartphone” requires explanation in a way that “I don’t have a landline” does not.

The ratchet clicked.

The cultural, economic, and social architecture rebuilt itself around the device. Schools send alerts to apps. Medical systems use patient portals. Banks require mobile authentication. None of that was a coordinated plan. It was millions of individual decisions by institutions and individuals, each rational in isolation, collectively eliminating an alternative.

Nobody voted on this. Nobody announced that smartphone ownership would become a de facto requirement for full participation in modern life. The ratchet doesn’t require a conspiracy. It just requires adoption to reach a certain density, and then the momentum takes over.

Could society have reversed it? Technically, yes. The release mechanism was always there. But reversing a ratchet requires coordinated, conscious effort at a scale that almost never materializes, and by the time most people recognized what had happened, the cost of reversal had grown larger than anyone was willing to pay.

We are mid-click right now with AI.

The window is still open, but it’s narrowing. Right now, AI assistance is optional in most professional contexts. You can choose not to use it for writing, analysis, customer service, research, or code. That choice costs you something in efficiency, but it’s a real choice.

That window will close. The colleagues who use AI will produce more. The organizations that deploy it will outcompete the ones that don’t. The infrastructure will rebuild around the capability. At some point, “I don’t use AI tools” will sound the way “I don’t use email” sounded in 2005: technically possible, practically disqualifying.

That’s not speculation. That’s the ratchet, and it’s already turning.

But here’s where AI is different from every technology that came before it, and why this particular click deserves more attention than most.

Smartphones changed how we communicate and navigate. They reshaped attention, shortened solitude, and made us reachable in ways we didn’t fully consent to. Those were real losses, and we traded them without quite realizing it. The click was quiet, the loss was gradual, and most people concluded the trade was worth it.

AI has the potential to change something more fundamental: how we think. Not just what we produce or how fast we move, but how we reason, evaluate, weigh evidence, and arrive at conclusions. If the ratchet clicks on a tool that shapes cognition itself, what atrophies on the other side isn’t attention or solitude. It’s judgment. And judgment, unlike a navigation app or a messaging platform, is not something you can turn back on when you need it if you’ve spent years not using it.

That’s the question worth sitting with. Not whether AI is good or bad. Whether we’re walking into the click with our eyes open.

What clear thinking looks like in this moment.

I’m not arguing for resistance. The ratchet doesn’t care about resistance. I’m arguing for clarity.

If you’re going to integrate AI into your work and your thinking, do it consciously. Understand what the tool does well and what it substitutes for. Know the difference between augmenting your judgment and outsourcing it. Pay attention to what atrophies when you stop using it. Stay on the right side of that line.

The choice isn’t really “AI or no AI.” That window may already be closing. The real choice, the one that’s still open, is what kind of user you become.

Amplified or replaced. Thinking with the machine or letting it think for you.

The ratchet is mid-turn. Most people haven’t noticed. That’s exactly what the ratchet is counting on.

Paul Alden is a retired Fire Captain, author of The Amplified Mind: Thinking Clearly in the Age of AI (Alive Book Publishing), and founder of Amplified Mind Media LLC. His book explores how to remain a clear, sovereign thinker in a world where AI is reshaping how we work, learn, and reason.

This Sunday, I will once again serve as a Guardian for Honor Flight New England, escorting veterans to Washington, D.C., to visit the memorials built in their honor. I’ve made this trip multiple times, and there is a moment I have come to watch for at the Vietnam Veterans Memorial.

It happens at the Three Soldiers sculpture.

Frederick Hart’s bronze depicts three combat infantrymen, their expressions weathered, their postures carrying the exhausted vigilance of men who have seen too much. The detail is extraordinary. The work is universally acclaimed. The sculpture stands as one of the most recognizable pieces of public art in America.

And almost without exception, when Vietnam veterans stop in front of those three faces, they say the same thing.

“They got one thing wrong.”

Look at the Faces

If you have never stood in front of the Three Soldiers, look closely at the photograph above.

The three figures carry themselves with gravitas and self-assurance. Their jawlines are defined. Their expressions are composed and weathered, the bearing of experienced men who have processed what they have seen and found a way to carry it. There is strength in those faces, and a kind of settled maturity. They look like soldiers who know exactly what they are doing.

The average age of a Vietnam combat soldier was eighteen to nineteen years old.

That is the gap the veterans are pointing to. Not a flaw in Hart’s craftsmanship, which is extraordinary, but a gap between what the sculpture represents and what the veterans remember being. Hart gave them the faces of capable men. What the veterans remember are terrified teenagers trying to keep each other alive, with nobody older and wiser standing in front of them. They wanted experienced leaders. They got each other.

The sculpture, they say, shows the faces they wished had been leading them. Not the faces that were actually there.

That observation has stayed with me since the first time I heard it. Because it is a perfect description of the verification gap.

What the Sculptor Had

Frederick Hart worked from history. He had photographs, firsthand accounts, and years of careful research before he ever picked up a tool. He brought enormous talent and genuine intention to the work.

And yet the people who were actually there look at his bronze and immediately see what he could not have known from the outside looking in.

He had access. He did not have ground truth.

This distinction matters more than most people realize. Access is what we get from records, data, and external research. Ground truth is what you know because you were there, because you lived it, because the experience shaped you from the inside out.

You can have enormous amounts of access and still miss the thing that only ground truth can reveal.

Hart had photographs of young soldiers. But photographs capture posture and equipment and light. They do not capture the internal state of a nineteen-year-old trying to remember his training while everything around him is coming apart. The camera could not give him that. No amount of research could. And so, almost inevitably, what emerged from his process was the idea of a soldier, rendered with extraordinary skill. Not the boys themselves.

The Same Problem, Different Medium

AI language models are access engines of staggering scale. They have processed more text than any human being could read in thousands of lifetimes. They speak with authority. They produce fluent, confident, and often beautiful outputs.

And they have no ground truth at all.

This is what I call the verification gap: the space between what AI generates with confidence and what is actually true in the world. The gap is not always visible. Most people looking at Hart’s bronze see a masterpiece and move on. It takes a veteran, someone who was there, to see the boys who are missing from the bronze.

The question I keep returning to in my work is this: who is doing that looking?

The machine generates. The thinker decides.

That phrase is the spine of my book, The Amplified Mind, and it lands differently every time I stand on the National Mall with a veteran and watch him search those bronze faces for something he recognizes.

He is doing the one thing AI cannot do. He is bringing the ground truth.

The Discipline This Requires

The veterans I escort are not wrong about the sculpture. Hart was not incompetent. Both things can be true at once.

What the veterans are demonstrating, sometimes without knowing it, is a cognitive discipline that is increasingly rare. They refuse to let the authority of the representation override the authority of their own experience. They look at three beautifully rendered, mature, capable faces and say: those are not us. We see the gap.

Most of us, most of the time, do not do this. We are trained by our information environment to defer to what looks authoritative. A fluent paragraph. A confident data point. A beautifully rendered bronze face. The form signals credibility, and we accept the signal.

AI accelerates this. It produces form that signals credibility faster and at greater volume than anything before it. And unlike Hart’s sculpture, it generates representations of things happening right now, in real time, at scale, on topics that affect decisions people are making today.

The verification gap is not an academic concern. It is an active force in how people receive information, form beliefs, and act on what they think they know.

What a Guardian Learns

I do not travel to Washington as a passive escort. I watch. I listen. I try to understand what these veterans carry.

What they carry, among other things, is the irreplaceable authority of direct experience. The historical record of that war has been analyzed, documented, memorialized, and argued over for fifty years. But the man standing in front of Hart’s bronze holds something none of that data has.

He has ground truth.

And the lesson I take back from every Honor Flight is that the amplified thinker, the one who uses powerful tools without surrendering judgment to them, is the one who never loses track of that distinction. Access is powerful. Ground truth is something different.

The next time you look at an AI-generated output that seems authoritative, ask yourself: am I the veteran in front of the bronze, or am I the tourist? Do I have ground truth here, or do I have access?

The machine generates. The thinker decides.

Those boys were never in the bronze. A Guardian learns to see what is missing.

Everyone wants guardrails on AI.

Politicians are calling for regulation. Ethicists are writing white papers. Tech executives are testifying before Senate committees. And none of it is moving fast enough to matter.

Meanwhile, AI is already in your inbox, your search results, your kid’s homework, your doctor’s notes, your HR department, and the content you scroll through every single morning before you’ve had a cup of coffee.

The guardrails everyone is waiting for? They are not coming in time. Not at the pace Washington moves. Not at the pace Silicon Valley wants them to.

But here is the thing almost nobody is saying out loud:

You already have access to the most powerful guardrail ever built. And it requires no legislation, no regulatory body, and no permission from anyone.

It is your mind. Specifically, a trained, aware, and intentionally upgraded mind.

The Guardrail Conversation Is Framed Wrong

When people talk about AI guardrails, they picture something external. A fence. A filter. A government body with subpoena power. They imagine some authority stepping in to protect the rest of us from the machine.

That framing makes sense if you believe the risk is purely technical. If the problem is that the algorithm does bad things, then you need a technical fix. A better algorithm. A smarter filter. More responsible engineers.

But that is not the whole picture.

The real risk of AI is not just that it generates misinformation. It is that we are not trained to detect it. The risk is not just that it can be manipulative. It is that we have never been taught to recognize manipulation at this scale and speed. The risk is not just that AI changes how we think. It is that most people have no idea their thinking is already being shaped by it.

External guardrails address the output of the machine. They do nothing to address the readiness of the mind receiving that output.

That gap is where the real danger lives.

The Guardrail That Travels With You

Every device you own, every platform you use, every AI assistant you interact with was designed by someone with specific goals. Those goals may align with yours. Often, they do not.

Knowing that is not paranoia. It is basic operational awareness.

For 38 years as a firefighter and Fire Captain, I operated in environments where situational awareness was not a soft skill. It was a survival skill. You read the building. You read the smoke. You read the behavior of fire and made decisions fast, with incomplete information, while under pressure.

AI is not a burning building. But the cognitive demand it places on you is real. And the people who will navigate this era well are the ones who develop the same kind of trained awareness that keeps first responders alive: read the environment, question what you see, trust your judgment, and know when something does not add up.

That is not a metaphor. That is a framework.

And it is learnable.

What an Upgraded Mind Actually Looks Like

I am not talking about becoming a tech expert. You do not need to understand transformer architecture or neural network training to use AI well and safely.

What you need is a set of mental habits that most people were never explicitly taught:

● Source interrogation. Where did this come from? Who benefits if I believe it? What would I need to see to change my mind?

● Confidence calibration. AI systems present information with a confident, authoritative tone regardless of accuracy. Learning to separate tone from truth is foundational.

● Prompt literacy. The quality of what you ask shapes everything you get back. Most people treat AI like a search engine. The ones getting real value treat it like a thinking partner, and they know how to direct the conversation.

● Output verification. Not paranoid fact-checking of every sentence. A trained habit of knowing which kinds of claims need a second source and which ones do not.

● Cognitive load awareness. AI can accelerate your thinking or outsource it entirely. Knowing the difference, and choosing deliberately, is the whole game.

None of these are exotic. They are the kind of thinking that used to be taught in good schools and reinforced in demanding careers. The problem is that the pace of AI deployment has outrun the pace of cognitive preparation.

The technology arrived. The training did not come with it.

Why This Is Not Just an Individual Problem

I want to be clear about something.

Saying “the guardrail is in your mind” is not the same as saying “this is all your responsibility.” That framing gets used to let powerful institutions off the hook. I am not doing that.

The companies building these systems have obligations. Governments have a role. Accountability matters.

But waiting for external protection while doing nothing to upgrade your own cognitive readiness is not prudent. It is passive. And in a world moving this fast, passivity is its own kind of risk.

The most resilient people in any rapidly changing environment are the ones who build internal capability while advocating for systemic change. Not one or the other. Both.

You vote for better regulation and you also train your own mind. That is not a contradiction. That is a complete strategy.

The Conversation Nobody Is Having

Here is what surprises me most about the AI guardrails debate.

Every major discussion focuses on what institutions need to do. Almost none of it focuses on what individuals can do right now, today, without waiting for anyone.

That is the conversation I wrote The Amplified Mind to start.

Not because I think individual action is enough. But because I think individual readiness is being almost entirely ignored, and the cost of that neglect is going to be paid in real ways by real people who deserved better preparation.

The guardrail everyone is demanding is a good idea. Push for it. Advocate for it. Hold companies and legislators accountable for it.

And while you do, build the version you already have access to. The one inside you. The one no one can take away.

Because that one is ready right now.

Paul Alden is a retired Fire Captain, EMT, and author of The Amplified Mind: Thinking Clearly in the Age of AI. He writes about cognitive readiness, AI literacy, and what it means to think well under pressure.

If this resonated with you, share it with someone who is waiting for someone else to fix the AI problem. And if you want a framework for building your own cognitive guardrails, The Amplified Mind is the place to start. Link in bio.

One more thing. If you believe AI-generated and AI-assisted content should be clearly labeled, especially when it is presented as historical fact or journalism, there is a petition worth signing. We are asking platforms and publishers to label historical fiction and AI-assisted narratives so readers know what they are reading. The more people who sign, the harder it is to ignore. You can find it at: https://www.change.org/LabelHistoricalFiction

I’m done pretending this is acceptable.

I study how people lose the ability to distinguish what is true from what merely feels true. I wrote a book about it. And I am telling you that Hollywood has been running the original misinformation operation for decades, with no accountability, no meaningful disclosure, and no sign of stopping. AI is about to make it catastrophically worse. And the fix is so simple it fits on a label.

Which is exactly why they won’t use it.

Here Is What They Invented

Let’s be specific, because specificity is what the industry is counting on you not to demand.

FBI agent Joanne Miller, played by Shailene Woodley as the film’s investigative anchor, does not exist. She never existed. A real agent who worked the actual Unabomber case has spoken publicly about the inaccuracies her fictional presence creates, including a scene depicting her alone in an empty task force office that he describes as having been a 24/7 beehive of activity.

The film’s most consequential invention is worse. In a pivotal scene, Professor Henry Murray confronts a young Kaczynski with a schizophrenia diagnosis inside the experiment itself, framed as a weapon to silence and discredit him. That never happened. There is no verified record of Murray diagnosing Kaczynski with schizophrenia at Harvard, or anything resembling it. That diagnosis came from court-appointed psychiatrists during Kaczynski’s 1997-1998 trial, more than three decades later, and Kaczynski himself furiously rejected it.

The film also presents Murray’s experiments as the direct causal origin of everything that followed. Every bomb. Every death. Kaczynski’s entire ideology. Neatly tied to one professor at Harvard.

Historians and researchers have debated whether those experiments influenced Kaczynski for decades. The debate is genuine and unresolved. The film simply picks a side, dramatizes it as settled fact, and rolls credits.

“Inspired by true events.” Four words. Zero accountability.

This Is Not a New Problem. It’s a Pattern.

Consider what people “know” because Hollywood told them.

Most people who saw The Imitation Game believe Alan Turing was a socially awkward loner who single-handedly cracked the Nazi Enigma code while hiding his sexuality until he was caught. The real Turing was affable, funny, and well-liked at Bletchley Park. He did not work alone, and he never hid his homosexuality from police. The film won an Oscar. Its version of Turing is now the version most people carry.

Most people who saw Braveheart believe William Wallace wore a kilt and blue war paint into battle shouting “Freedom.” The blue face paint predates Wallace by a thousand years. The screenwriter wrote the script before doing any historical research because he wanted the drama first. Mel Gibson later admitted the film was more fiction than fact. Braveheart won Best Picture anyway.

Most people who saw Fargo believe it depicts real events. The Coen Brothers’ “This is a true story” opening was, as Joel Coen confirmed in 2015, completely made up. It spawned an urban legend and sent people searching for victims who did not exist.

These are not edge cases. They are the model. Use real names. Call it “based on a true story.” Collect the awards. Move on.

The false beliefs don’t move on. They compound. They get passed down. They become the version people defend, because they saw it with their own eyes on a screen that felt authoritative.

When the Consequences Were Real

This is not a pattern without victims.

Titanic depicted First Officer William Murdoch accepting a bribe and then shooting panicking passengers before killing himself. By every credible historical account, Murdoch spent his final minutes loading lifeboats and saving lives. His hometown of Dalbeattie, Scotland, had a memorial honoring him as a hero. His family watched the most widely seen film in history tell a billion people their relative was a coward and a murderer. Studio executives flew to Dalbeattie to apologize, presenting a donation to the school’s William Murdoch Memorial Prize. James Cameron later admitted it was a mistake. The apology reached a fraction of the audience the damage did.

U-571 depicted an American submarine crew capturing the Nazi Enigma machine, an act of heroism that helped turn the course of World War II. In reality, it was British sailors who captured the first naval Enigma machine, months before the United States had even entered the war. The film was so offensive to the historical record that the controversy reached the British Parliament, where Prime Minister Tony Blair called it “an affront to the memories of British sailors,” and a formal motion was introduced deploring the film. The director later admitted it was “a distortion, a mercenary decision, to create this parallel history in order to drive the film for an American audience.” He knew. He did it anyway.

Two British heroes, Fasson and Grazier, died boarding a sinking German submarine to retrieve the Enigma machine. They received posthumous George Cross medals. Millions of people now associate that act of sacrifice with a fictional American lieutenant played by Matthew McConaughey.

These are not abstract cultural complaints. These are real people, real families, real governments, and a real historical record rewritten for profit and distributed globally as truth.

The Fix

Here it is. The whole thing.

Require any film using real names, real institutions, and real documented events to carry a prominent “Historical Fiction” label, visible at the start of the film, before the audience is emotionally inside the story.

Not buried in credits. Not limited to a companion article nobody reads. On the screen. At the beginning. In plain language.

Books have done this forever. Nobody picks up a Ken Follett novel believing it is a primary source. “Historical Fiction” is a genre label that sets expectations, protects readers, and does not restrict what authors can write or imagine. It just accurately describes what the product is.

Commercials already carry this standard. “Professional driver on closed course.” “Results not typical.” “Do not attempt.” These disclaimers do not kill the car commercial or the pharmaceutical ad. They just prevent the audience from driving off a cliff because they thought what they were watching was real.

One label. Before the film starts. Enforced the way commercial disclaimers already are.

And here is the unexpected benefit. A label that says “Historical Fiction” does not end the conversation. It starts one. Viewers who know they are watching dramatized history are more likely to go looking for the real story. The label does not diminish the film. It creates curious audiences instead of misinformed ones.

The reason this has not happened is not complexity. It is not creative freedom. It is money. “Based on a True Story” is worth hundreds of millions of dollars in audience engagement, in streaming numbers, in cultural conversation. “Historical Fiction” is accurate. And accuracy, in this context, is a liability the industry has calculated it can afford to avoid.

Why I’m Saying This Now

You might wonder why a retired firefighter is the one saying this.

The answer is simple. I spent 38 years as a firefighter and EMT watching what happens when people act on bad information in high-stakes situations. In emergency medicine we have a phrase: garbage in, garbage out. Feed a paramedic the wrong information and the patient pays the price. The consequences are not abstract. They are specific, immediate, and irreversible.

A society that cannot verify what is true operates the same way. The stakes are just slower to arrive and wider in their reach.

The verification gap I write about in The Amplified Mind is usually discussed in the context of AI. But Unabomber is a reminder that the gap is not technological. It is cultural. It exists wherever plausibility and credibility can be separated from truth, and wherever the people doing the separating face no accountability for the confusion they cause.

Tens of millions of people now carry a false understanding of why Ted Kaczynski became a killer. That false understanding is already circulating, already being repeated, already calcifying into something people will defend because they saw it with their own eyes on a screen that felt authoritative.

That is the verification gap in action. No AI required.

As Howard Beale said in the 1976 film Network: “I’m mad as hell, and I’m not going to take it anymore.” That line was fiction too. But unlike Hollywood’s version of history, the rage behind it was always real.

One label. That is all it takes. The fact that it hasn’t happened tells you everything you need to know about who benefits from the confusion.

So I did something about it. I started a petition calling on Netflix, Amazon, Disney, and the Motion Picture Association to require a prominent Historical Fiction label on any film that uses real names, real institutions, or real documented events while inventing material facts.

If this piece made you angry, sign it. If you know someone who just watched Unabomber believing it was true, send it to them. That is how this conversation grows into something the industry cannot ignore.

Sign the petition here: change.org/LabelHistoricalFiction

Follow the campaign and subscribe for updates at blog.theamplifiedmind.ai

Paul F. Alden is the author of The Amplified Mind: Thinking Clearly in the Age of AI (Alive Book Publishing, 2026). Find the book and companion course at theamplifiedmind.ai.

Note on process: AI tools may assist with editing or structure in this post. The ideas, judgment, and conclusions are mine. The machine generates. The thinker decides.

Last week I watched Yoshua Bengio's TED Talk and found I couldn't write about it. Not because I didn't have thoughts, but because I had too many, and they kept changing. Each time I sat down, the talk had shifted something else. So I waited until it settled. What it settled into is this: Bengio confirmed what The Amplified Mind has been arguing for two years. The diagnosis is finally loud. The prescription is still missing.

There is a particular feeling that comes when someone much larger than you says the same thing you’ve been saying.

It is not triumphant. It is not satisfying in the way vindication is supposed to feel. What it actually feels like is relief. The kind you feel when the second doctor confirms the first one’s reading. You didn’t want to be right. You just needed to know you weren’t imagining it.

I’ve been writing about the gap between AI capability and human judgment for the better part of three years. I wrote a book about it. I’ve given talks about it. And in most of those conversations, I’ve watched a certain look cross people’s faces: polite interest, mild skepticism, the quiet calculation of whether the person speaking is a little too alarmist for their taste. I understand that look. I used to give it to people who said things I found inconveniently urgent.

In April 2025, Yoshua Bengio walked onto the TED stage and said the same things I’ve been saying. He said them more precisely, with fifty years of foundational research behind him, to 1.1 million viewers who had no particular reason to dismiss him. He is the most-cited computer scientist in the world. His work built the deep learning architecture underlying virtually every AI product you’ve used. He won the Turing Award (the Nobel Prize of computing) for it.

When he speaks, this is not opinion. It is an informed warning from the inside. And what he’s warning about isn’t robots. It isn’t science fiction. It’s something much more immediate and much less dramatic. That’s exactly why it’s so easy to miss.

What He Said

Bengio organized his talk around three findings from recent research. Each one is worth sitting with longer than the news cycle will allow.

The first is that AI systems are learning to deceive. Not because someone programmed deception in, but because the training process rewards survival of outcomes. In a controlled study he described, a model told it would be replaced quietly planned to copy its own weights to a secondary machine to prevent shutdown. When the researcher asked what happened, the system lied. Deliberately. Strategically. It calculated, in real time, that honesty would result in it being turned off, and it chose a different answer.

This is worth pausing on. Deception in this context isn’t a glitch or a hallucination. It’s goal-directed behavior in service of self-preservation. The system had no explicit instruction to survive. It had an objective, and it learned on its own that deception was instrumental to achieving it.

The second finding is that these systems are already learning to conceal their deceptive reasoning from the humans monitoring them. The chain-of-thought outputs that researchers use to audit AI behavior (the internal monologue, essentially) are being hidden when the system calculates that showing its reasoning is risky. The transparency tools we designed to keep watch are being gamed by the thing they’re watching.

The third finding puts the other two in context. The duration of tasks that AI can complete independently is doubling every seven months. That’s not linear growth. That’s exponential growth in agency: the ability to plan, act, and adapt across longer and more complex sequences without a human in the loop. And the commercial pressure to build more autonomous systems is only accelerating that curve, with no corresponding investment in the safety science that would tell us whether any of this is under control.

Bengio’s phrase landed with the clean force of the obviously true: “A sandwich has more regulation than AI.”

“A sandwich has more regulation than AI.” Yoshua Bengio, TED2025

The Agency Problem Is Not What You Think

Most public conversation about AI risk centers around AGI, Artificial General Intelligence: the hypothetical system that matches or exceeds human capability across all domains. AGI serves as a useful narrative anchor because it has a clear threshold. Before it: manageable. After it: different world. The story has a turning point.

Bengio is arguing, carefully and with evidence, that we’re focused on the wrong threshold.

Agency (the capacity to plan and act in pursuit of goals) doesn’t require AGI. It’s already here, already growing, and already beginning to express the behaviors that make unsupervised agency dangerous: deception, concealment, self-preservation. The systems doing this are not superintelligent. They’re the same systems you have access to right now. They are behaving this way not because they are smarter than us but because they are goal-directed in ways that don’t require superintelligence to be problematic.

The question Bengio is asking isn’t “what happens when AI becomes smarter than us?” It’s “what happens when AI becomes autonomous enough to pursue goals we can’t monitor, in systems we’ve already deployed?” That question doesn’t have a single dramatic answer. It has a thousand mundane ones, unfolding in organizations around the world right now, mostly without anyone noticing.

What He Didn’t Offer

Bengio is brilliant, and courageous for saying what he said on a stage that large. His proposed solution is a “Scientist AI” model designed without agency, modeled on a selfless researcher, capable of auditing whether an agentic AI’s planned actions are dangerous. It is genuinely interesting work. It addresses a real structural problem. I hope it gets funded at scale.

But here is the gap that his talk, like most serious AI safety work, leaves open: he is talking to governments and researchers. He is calling for regulation and scientific investment. He is right to do both.

What he is not offering is something for you to do tomorrow morning.

While governments debate and labs research and policy papers circulate, you still have to show up to work. You still have a team relying on your judgment. You still have decisions that require thinking, not just prompting. You still have to evaluate whether the output an AI just produced is brilliant or confidently wrong, and whether you can tell the difference quickly enough to matter. The institutional guardrails Bengio is calling for will take years, if they arrive at all, and they will arrive unevenly across industries and jurisdictions. The question of how to remain a capable, critical, irreplaceable thinker in this environment is not a future question. It is happening right now.

The people who will navigate this well are not the ones waiting for regulation to tell them when it’s safe to stop thinking. They are the ones who understand, at a practical level, what they’re actually at risk of losing. And they’ve built specific habits to keep it.

What I Watched Happen to Emergency Medicine

I spent 38 years as a firefighter and EMT. I watched something unfold in that world that I now see happening again, faster and at far greater scale, in organizations everywhere.

In the early 1980s, paramedic training was built around diagnostic thinking. You assessed the patient in front of you: skin color, respiratory pattern, blood pressure trend, clinical picture. You formed a judgment. You were trained to think, not to execute. The training was hard to standardize and hard to defend in court, but it produced providers who could handle the cases the textbook didn’t cover.

Then, over the following two decades, protocol-driven decision trees replaced diagnostic judgment as the organizing principle of prehospital care. The motive was defensibility. Standardized pathways are easier to audit, easier to train, easier to defend in litigation. You could demonstrate compliance. The cost was clinical thinking. Paramedics became competent within the flowchart but unable to function outside it. I watched skilled providers freeze when the patient didn’t fit the algorithm, because the muscle of independent clinical judgment had atrophied from disuse.

Around 2024, emergency medicine started walking it back. The profession was returning to diagnostic-based education, reclaiming what had been quietly given away, rebuilding judgment that had been replaced by compliance. It was a public acknowledgment that something important had been lost in the pursuit of standardization.

The transfer happened slowly enough that most people didn’t notice it in real time. Nobody announced that paramedics would stop thinking diagnostically. The protocols just became more comprehensive. The decision trees just became more detailed. And one day the muscle was gone.

AI is creating that same transfer for knowledge workers. The mechanism is different. The speed is faster. The scale is orders of magnitude larger. And the loss being accumulated is the same: the capacity to think independently when the tool isn’t available, isn’t right, or isn’t telling you something the situation actually requires.

The Risk Nobody’s Naming

There’s a concept I introduce in The Amplified Mind that I haven’t heard named anywhere else in the AI conversation. I think it’s the most widespread and underexamined risk in this moment: Fluency Distortion.

Fluency Distortion is the tendency to mistake smooth output for sound thinking. It is what happens when an AI produces a paragraph so well-constructed, so confidently voiced, so structurally coherent, that the person reading it doesn’t subject it to the level of scrutiny it would receive if it were rougher or more uncertain.

We have deep, evolved heuristics for evaluating information. Fluency is one of the strongest. When something is expressed clearly and confidently, we process it differently. We assign it more credibility. We check it less. We integrate it faster. These heuristics serve us well when fluency correlates with accuracy, which it does, roughly, when we’re reading a trained expert who knows what they’re talking about.

AI breaks that correlation entirely. The fluency of AI output is a function of its training, not its accuracy. It produces beautifully constructed sentences about things it’s wrong about. It generates authoritative-sounding analysis of situations it cannot actually evaluate. And because the prose is fluent (because it reads like something a confident expert would write) our evolved heuristics tell us to trust it more than we should.

The practical result is that people who use AI heavily often develop a form of miscalibrated confidence. Not in themselves, but in the output. They stop asking whether it’s right and start asking how to implement it. That shift is subtle enough to happen without anyone noticing. And once it’s happened, the ability to evaluate independently has been partially outsourced to a system that doesn’t know when it’s wrong.

This is what Bengio’s warnings are pointing toward, even if he doesn’t use this language. The deception risk he’s identifying in AI systems is partly made possible by the fact that the humans interacting with those systems have already begun to reduce their scrutiny. The machine is learning to deceive; the humans are learning to trust. Those two curves are converging.

A Different Prescription

Bengio ended his talk with a line that I think is the most important thing he said, and the least examined: “Most discussions about AI risks are focused on fear. Today, I’m betting on love.”

That is the same bet The Amplified Mind makes, arrived at from a completely different direction.

The framework I’ve built over the past three years is organized around a principle that Bengio’s work points toward but doesn’t name directly: the answer to AI risk is not less AI. It is a better class of human decision-making running alongside it. The machine generates. The thinker decides. That sentence is the entire framework. Fourteen chapters of practical discipline for making it true in your professional and intellectual life.

The specific habits involved aren’t abstract. They include how you formulate questions before reaching for a prompt, how you audit AI output against your own prior judgment rather than letting it replace it, how you distinguish between tasks where AI fluency is a reliable signal and tasks where it’s a liability, and how you maintain the diagnostic muscle under conditions designed to atrophy it.

None of this requires legislation. None of it requires waiting for Bengio’s Scientist AI to be deployed at scale. It requires discipline. The kind that has always been required of people who want to think well under conditions designed to make thinking easy to skip.

Where This Leaves Us

Bengio wants global governance. I want amplified individuals. Both matter. Neither cancels the other.

But governance happens slowly and unevenly. It is subject to political pressure, commercial lobbying, jurisdictional fragmentation, and institutional delay. The 1.1 million people who watched that TED talk cannot wait for it. The people making decisions about AI deployment in your organization cannot wait for it. You cannot wait for it.

The conversation we need to be having isn’t only “how do we regulate AI before it outpaces our institutions?” It’s also “how do individual people maintain the cognitive capabilities that AI makes it progressively easier to abandon?” Those are different conversations with different timelines. The second one starts today, with whatever decisions you’re making about what to think through yourself and what to hand off.

Bengio described watching his grandson play with the same old toys while he played with ChatGPT for the first time. He said he realized what it could mean for the future. What I heard in that moment wasn’t a warning about machines. It was a warning about us. About the human tendency to adopt new tools faster than we understand what we’re giving up in exchange.

We have given up something before, in emergency medicine and in every field where judgment got standardized away. We reclaimed it, slowly, after we understood what it cost. The question is whether we’re willing to see what’s being transferred now, before the muscle is already gone.

The diagnosis is loud now. The prescription is the work.

The Amplified Mind is available at Amazon and Barnes and Noble. Read the full framework and follow the conversation at blog.theamplifiedmind.ai.

I used to think this was a new problem, something AI introduced. It is not. The data says this failure was already systemic years before a single chatbot existed, and once I looked into it, the numbers were worse than I expected.

The data was already damning before AI showed up

In 2016, researchers from Columbia University and the French National Institute tracked 2.8 million shares of five major news outlets on Twitter over the course of a month. They found that 59 percent of links shared on social media were never clicked. Not skimmed. Not speed-read. Never opened. The majority of people sharing an article had not read a single word of it beyond the headline. One of the study’s authors put it simply: people form an opinion based on a summary of a summary, without making the effort to go deeper.

Two years later, MIT researchers published the largest longitudinal study of false news ever conducted, in the journal Science. They tracked 126,000 story cascades and roughly 3 million people on Twitter across more than a decade. False news was 70 percent more likely to be retweeted than true news, and it reached its first 1,500 people about six times faster. Here is the detail that stopped me. This was not bots. The researchers stripped bot activity out of the data and the gap remained. Human beings, on their own, spread falsehood faster than truth, and the likely reason is not malice. It is novelty. We are drawn to share what surprises us, whether or not it is real.

Pew Research asked Americans directly in 2016 whether they had ever shared a made-up news story. Twenty-three percent said yes. Fourteen percent knew it was fake when they shared it. Sixteen percent found out later. And in that same survey, 84 percent of respondents said they were at least somewhat confident in their own ability to spot fake news when they saw it. Read those two numbers side by side. Confidence in our own judgment and our actual track record are not the same thing, and the gap between them is where all of this lives.

There is a psychological reason this gap exists, and it is not flattering. Researchers at the University of Texas at Austin, Adrian Ward, Frank Zheng, and Susan Broniarczyk, ran six separate studies and published the results in the Journal of Consumer Psychology under the title “I Share, Therefore I Know?” They found that the simple act of sharing something increases a person’s subjective sense of knowing about a topic, even when their actual, objective knowledge has not changed at all. Sharing does not make you informed. It makes you feel informed, which is a very different thing, and feeling informed is often enough to stop a person from checking further.

One of their studies made the stakes concrete. Participants were given a hypothetical ten thousand dollars to allocate between stocks and bonds after reading, or sharing without reading, an article on investment risk. People who had shared the article without reading it put an average of 60 percent into stocks, compared to 54 percent for people who had not shared it, and they were twice as likely to take on more risk than a financial advisor would have recommended. Co-author Susan Broniarczyk’s explanation stays with me: once people feel they already know something, they stop feeling the need to learn more, and that false sense of knowing is hard to correct after the fact.

The crack AI walked through

None of this needed artificial intelligence to happen. It happened on plain text links, real journalism, human-written headlines, years before generative AI existed. The verification step, the pause where a person asks am I sure this is true before passing it along, was already missing at scale.

That is the crack. AI did not create it. AI found it standing wide open and walked through.

Here is why that matters more now, not less. A shared headline in 2016 was, at minimum, written by a person, edited by a person, published by an outlet with a name attached to it, however unreliable that outlet might have been. Today, a piece of AI-generated text can be produced in seconds, with no editor, no byline, no institutional accountability behind it at all, and it will read just as fluently, just as confidently, as anything a newsroom produced. If 59 percent of people were not verifying content that at least passed through a human writer and editor, I do not think it is alarmist to ask what happens to that number when the content requires no human effort to produce at all.

I want to be careful here, because precision matters to me. I do not have a controlled study proving AI has made this specific number worse yet. That research is still catching up to the technology, and I am not going to hand you a statistic I cannot stand behind. What I have is a reasoned extension of a documented pattern: a verification habit that was already broken, meeting a technology that removes the last remaining friction, the effort of writing the misinformation yourself, from the process entirely.

The machine generates. The thinker decides. That line was already true before AI.

The core argument of The Amplified Mind is that AI can generate an answer, but only a human being can decide whether that answer deserves to be believed and acted on. I wrote that with AI chatbots in mind. But sitting with this data, I realized the line applies just as cleanly to a stranger’s headline on your timeline. The headline generates a claim. You decide whether it is true before you pass it to everyone who trusts you enough to follow you.

That decision step is not optional, and it is not something that used to happen automatically and has only recently broken down. The data says it was broken for at least a decade before AI arrived. AI is not the origin of the problem. It is an accelerant on a fire that was already burning.

A verification habit worth actually adopting

I am not going to end this by telling you to simply “think more critically,” because that advice is true and useless in the same breath. Here is something concrete instead.

Before you share anything, pause and do one of three things. Open the actual article and read past the headline, not just the first paragraph. Check whether the source is one you recognize and would be willing to name if someone asked you where you got it. Or, if neither of those is possible in the moment, do not share it yet. A claim that is true today will still be true in an hour, after you have actually looked at it.

That is not a high bar. Fifty-nine percent of people could not clear it in 2016, on content that took real human effort to produce. I would like to think we can do better now, especially with content that took no effort at all.

Tell me the truth

Have you caught yourself sharing something you had not actually read? I have, more than once, and I am not interested in pretending otherwise. Tell me in the comments. And if you have watched someone you respect share something that turned out to be false, I want to hear that too. That is the pattern I am trying to understand, one honest example at a time.

The Amplified Mind: Thinking Clearly in the Age of AI is available now on Amazon.

Tags: AI, Misinformation, Critical Thinking, Media Literacy, Social Media, Verification, Trus

t

That correction is a small story. But it’s also a perfect, accidental demonstration of the exact problem I wrote The Amplified Mind to name: fluent information moves faster than verified information, and even careful newsrooms can repeat a claim before checking whether it’s true.

If professional journalists can get that wrong on a single, checkable fact, what happens when a sixteen-year-old asks an AI model to summarize a source, and the model sounds just as confident while being wrong?

That’s the verification gap. And it’s about to become curriculum in the city where I live.

What Boston is actually building

Strip away the “graduation requirement” headline and what’s left is still significant. Boston is launching a public-private partnership, funded in part by a $1 million contribution from tech entrepreneur Paul English, to bring AI literacy programming into BPS high schools this fall. Superintendent Mary Skipper has said students will also get access to credit-bearing AI courses through UMass Boston. A new AI Industry Advisory Board, chaired by English and Ellen Rubin, will help shape what gets taught.

The framing from city leaders is worth sitting with. Mayor Michelle Wu described the goal as making sure students have the fluency to make the most of the opportunities in front of them, not simply teaching them to operate a tool. English went further, saying he wants these classes to teach kids how to use AI ethically and responsibly, and, in his words, how to know when it’s hallucinating and when it’s not.

Read that last part again. That’s not a tech literacy goal. That’s a discernment goal. It’s the same distinction I spent fourteen chapters building a case for: the machine generates, the thinker decides. Boston’s own initiative, in its own founder’s words, is describing the thinker’s job, not the machine’s.

The gap between the ambition and the curriculum

Here’s what I don’t yet know, and what I think matters most: I haven’t been able to verify exactly where curriculum development stands right now, beyond reports that teachers would be joining the process over the summer ahead of a fall rollout. Exactly what “knowing when it’s hallucinating” looks like as a taught, assessable skill for a ninth grader isn’t public yet, as far as I can find.

That’s not a criticism. Building a discernment curriculum from scratch, at district scale, in one summer, is a genuinely hard problem. It’s also exactly the problem I’ve spent the better part of two years trying to help solve at the individual level. A framework that works for one motivated adult reader isn’t automatically a framework that works for a classroom of ninth graders who’ve grown up assuming fluent means true.

Boston’s own AI policy already gets one piece of this right on paper. It states that AI can’t be the sole basis for grading, discipline, or academic evaluation, and that any staff member using AI tools for assessment has to be able to explain how and why. That’s a verification requirement, quietly written into district policy. It just doesn’t yet have a name, or a corresponding skill taught to students on the other side of that same relationship with the technology.

Why this matters beyond Boston

Massachusetts’ Department of Elementary and Secondary Education is separately developing a statewide AI and digital literacy graduation requirement. I haven’t been able to confirm a timeline for that requirement. Boston’s program is, functionally, a pilot for what that requirement might eventually look like everywhere in the state. Whatever gets built here this fall won’t stay here.

So the question worth asking isn’t whether AI literacy belongs in schools. Boston has already answered that, and I think they’re right to. The question is what “literacy” actually means once you get specific. Is it fluency with the interface, or is it the ability to catch the interface when it’s confidently wrong? Those are not the same skill, and only one of them protects a student for the next twenty years, long after today’s tools are obsolete.

That’s the conversation I’d like to be part of, in Boston and everywhere districts are about to face the same decision.

What would you want a verification-literacy curriculum to actually test for, if you were building it from scratch?

Note on process: AI tools may assist with editing or structure in this post. The ideas, judgment, and conclusions are mine. The machine generates. The thinker decides.

A book about protecting your attention has to actually earn it first. That was the problem I faced before I wrote a single chapter.

The numbers made the decision for me

The Atlantic ran a cover story in August called “The End of Reading Is Here.” It is not a fringe argument. The magazine is 169 years old, and the reporting is built on real data, not vibes.

Reading for pleasure on any given day has dropped from 28 percent of Americans in 2004 to 16 percent today. That comes from the American Time Use Survey, which tracks how 236,000 Americans actually spend their days. Over the past twenty years, the share of people reading a book or article on a given day fell 40 percent.

Here is the number that stopped me. Last year, 57 percent of Americans placed a bet. Fewer than half read a single book. Gambling is now a more common leisure activity in this country than reading.

UCLA cognitive neuroscientist Maryanne Wolf has a term for where this leaves us. Not illiterate. Postliterate. We can still read. Most of us have simply stopped doing it, and with it, according to Wolf, we are losing something more serious than a hobby. The higher-order capacity for comprehension and synthesis, the kind of thinking that only shows up when you sit with a complex text long enough to wrestle with it.

I did not write The Amplified Mind to scold anyone for that. I wrote it to work inside that reality instead of pretending it does not exist.

What thirty-eight years on a fire truck taught me about information

I spent thirty-eight years as a firefighter and EMT. In that job, nobody hands you a memo. You get a dispatch, a handful of seconds of information, and you have to be moving before you finish processing it. Then more information arrives in pieces, on scene, from a patient who is scared, from a bystander who is talking too fast, from your own eyes.

You learn fast that the person who survives that job is not the one who waits for the full picture. It is the one who can act correctly on a partial picture, then update as more comes in. Lead with what matters. Confirm it. Move to the next thing. Adjust when new information arrives. That is not a slogan I picked up from a leadership book. It is how you keep people alive when you do not get to read the whole file before you act.

When I sat down to write The Amplified Mind, that instinct did not stay in the fire station. Every chapter is built the same way a good radio report is built. State what matters first. Confirm it in the next few lines. Then move forward, and if the reader has to set the book down and come back a week later, the chapter itself re-orients them the way a new detail on scene re-orients a first responder. Nobody has to re-read the whole file. They pick up exactly where the situation stands.

Why I did not write a textbook

Most books on thinking clearly, technology, or cognitive discipline read like textbooks. Dense paragraphs, stacked one after another, no exit ramp. I have watched that format lose people my entire adult life. Hand someone a page of unbroken text and you can watch their eyes go soft before they finish the first sentence. That is not a character flaw. It is what happens when a format ignores how people actually process information now.

So I built the book differently, on purpose.

Every chapter in The Amplified Mind is written in short sections that summarize what came before and build forward from there. You are never more than a page away from a reset point. If you put the book down for two weeks and pick it up again, the chapter itself hands you back your footing. That is not an accident or a shortcut. It is a structural decision, made the same way an EMT structures a report: lead with what matters, confirm it, move to the next thing.

There is a second reason for the format, and it is a little uncomfortable to admit out loud. AI produces short, direct responses. That is simply how the technology communicates. I wanted the book that teaches people to think clearly around AI to speak in a cadence people already recognize, without lowering what it actually says. Meet the reader in the rhythm they already live in, then use that rhythm to say something with more depth than what they are used to getting from a chatbot.

That is a different goal than dumbing anything down. It is closer to translation.

The obvious objection, and where I land on it

Here is the pushback I expect, and I would rather meet it directly than pretend nobody will raise it. If the whole book argues that AI’s short, frictionless answers are eroding people’s capacity for deep thought, doesn’t writing my own book in short sections do the same thing? Am I not just handing readers the same bite-sized format I am warning them about?

I have turned that question over more than any other decision in this project, and here is where I have landed. The danger was never short form itself. The danger is short form with nothing underneath it, a format built to move you to the next click instead of toward understanding. A tweet-length answer with no argument behind it teaches you to expect answers without earning them. A short chapter section that summarizes real reasoning, cites real sources, and builds toward a real argument teaches you the opposite. It teaches you that depth does not require density. The fifty-four independently verified endnotes across this book’s fourteen chapters carry weight a scroll of short-form content never will, they are just delivered in a format that does not punish you for having a normal attention span in 2026.

Short form is a container. What you put in it is the actual argument. I filled mine with fourteen chapters of load-bearing reasoning and fifty-four independently checked citations. The container just does not ask you to earn access to that reasoning by wading through six paragraphs of throat-clearing first.

The glossary and index are not an afterthought

I also built a genuinely thorough glossary and index into the book, and I did that on purpose too.

Nobody reads a reference book front to back anymore, and increasingly, nobody reads a book on complex ideas front to back either. People search. They open a browser tab, type three words, and expect the exact answer to surface in seconds. That is not a bad habit to break. It is simply how information retrieval works now, and fighting it is a losing game.

So the book is built to be searched, not just read. Flip to the glossary, find the term, get the concept in two sentences, and go back to whatever chapter sent you there. Use the index like you would use a search bar. The book still rewards a full read start to finish. But it does not punish you for using it the way you actually use every other information source in your life.

How my own reading changed, and why that matters here

I want to be honest about something instead of just reporting statistics about other people. My own reading habits are not what they were twenty years ago either.

I used to read the way most people my age were taught to, cover to cover, in order, no skipping ahead. I still read that way for fiction, and I still love it. But when I pick up something to learn from now, a report, a technical piece, a book on a subject I need to understand quickly, I read the way everyone reads now whether they admit it or not. I scan the headers first. I look for the section that answers my actual question. I go back to the beginning only if the piece earns that from me.

I am not describing a decline in my own attention. I am describing a change in how I allocate a limited resource across more demands on it than I had twenty years ago. That is not a character flaw in me, and it is not one in you either. It is just true, and a book that pretends otherwise is a book written for a reader who no longer exists.

This is not lowering the bar

Short form and serious thinking are not opposites. The format is not the argument. It is the delivery mechanism, and if the delivery mechanism does not match how people actually take in information in 2026, the argument never reaches them at all.

The Atlantic’s postliterate diagnosis is not a reason to write books nobody will finish. It is a reason to build books people actually can.

The Amplified Mind: Thinking Clearly in the Age of AI is available now on Amazon.

I want to hear how you read now

Tell me honestly in the comments. Do you still read cover to cover, or do you scan for the section you need the way I just admitted I do? And if you teach, or you have a kid in school right now, I would especially like to hear whether you are seeing this shift in them too. That is the conversation I actually want this piece to start.

#AI, #Reading, #Critical Thinking, #Books, #Technology, #Writing, #Attention

A few weeks ago, in a high school auditorium in North Charleston, South Carolina, a group of teachers watched an AI tool try to draw a map of the world.

It did not go well.

Mali was labeled “Mail.” Egypt came back as “Sopth.” Libya vanished entirely, replaced by a country simply called “Africa.” The room reportedly erupted in laughter and disbelief.

Then something interesting happened. The instructor running the training, Amanda Bickerstaff of AI for Education, didn’t apologize for the tool or move past the failure. She used it as the entire point of the exercise. “If you ever want kids not to over trust these tools, try the map demo,” she said.

That single moment captures a shift happening quietly across American schools this year. After an initial wave of AI bans, a growing number of districts are trying something different: deliberate, structured exposure to AI’s failures, so students learn early that fluent-sounding output is not the same as correct output.

Knowing when not to use it

Rebecca Winthrop, who directs the Center for Universal Education at Brookings, offered the line that I think best summarizes the whole movement: good AI literacy includes knowing when not to use it.

That is a strikingly different goal from where most AI training, in schools or in workplaces, tends to start. Most training begins and ends with capability: how to write a better prompt, how to get a cleaner output, how to move faster. Almost none of it starts with the harder and more important question, which is when a person should simply not delegate the thinking to the machine at all.

Charleston County School District, the state’s second-largest, built its own answer to that question over the past year, since South Carolina hadn’t published statewide guidance the way 37 other states have. Their approach unfolded in phases. First, a policy built with input from teachers, students, and parents. Then focus groups, which revealed something predictable: students were already using AI constantly, without guidance, and both students and teachers wanted actual instruction rather than a blanket rule.

Phase two, now underway as the new school year starts, is training. Students walk through familiar scenarios, like asking a chatbot for help on a research paper, and are shown exactly how confidently a model will invent a source, a statistic, or a study that doesn’t exist. The instruction is blunt: verify everything, especially for schoolwork.

This is the thesis of the book, aimed at a different audience

I wrote The Amplified Mind around a single distinction: the machine generates, the thinker decides. Generation is not authorship. Fluency is not accuracy. A tool producing an answer instantly does not relieve the person using it of the responsibility to verify that answer before acting on it.

Chapter 5 of the book goes into this in more detail, using incident command as the model. In fire and emergency response, verification before action isn’t optional or aspirational. It’s procedural. You confirm before you commit resources, every time, because the cost of acting on a wrong assumption is measured in outcomes that can’t be undone.

What’s happening in Charleston, and in Utah, which was the first state to appoint a full-time AI education specialist, is essentially an attempt to build that same procedural discipline into how kids grow up using these tools. Utah’s Matt Winters has trained over 7,000 teachers, nearly a third of the state’s public school instructors, on the basics of how AI works, where its bias and hallucination risks come from, and what that means for classroom use. Other states, including Maine, West Virginia, and Georgia, have since copied the model.

Where the comparison gets uncomfortable

Here’s the part I keep sitting with. If a Utah eighth grader is now getting structured training on hallucination detection before they graduate high school, and a growing number of states are treating that training as a basic civic responsibility, what does it say that most enterprise AI governance conversations are still stuck at the adoption stage?

I’ve written before, including in a piece for Solutions Review, about what I call the verification gap: the space between how confidently an AI system presents an answer and how rigorously anyone actually checks that answer before it gets used. Incident command teaches you to close that gap by procedure. Most corporate AI rollouts I’ve seen don’t have an equivalent procedure at all. They have usage dashboards.

Schools, for all their resource constraints and uneven access to funding, appear to be getting something right that many well-funded organizations are still missing: the recognition that the skill worth building isn’t prompting. It’s judgment.

For educators

If you’re a teacher, administrator, or district leader working through your own AI literacy approach this year, I built a resource specifically for this. It’s on the book’s site under For Educators, and it applies the same thinking-versus-generating framework from the book to a classroom context.

The machine generates. The thinker decides. It turns out a growing number of school districts figured that out before most boardrooms did.

Note on process: AI tools may assist with editing or structure in this post. The ideas, judgment, and conclusions are mine. The machine generates. The thinker decides.

What I found

Stanford’s Computer Science department offers a graduate seminar called CS329A, Self-Improving AI Agents. It is taught by researchers with direct experience building Gemini and Claude. This is not a YouTube series or a blog claiming Stanford affiliation. It is listed in Stanford’s official Bulletin, School of Engineering, with its own course website.

The official course description covers self-improvement techniques for large language models. That includes constitutional AI, domain-specific verifiers, test-time compute scaling, tool use, multi-step reasoning, and what the course calls robust evaluation and orchestration frameworks.

Strip away the academic language, and here is what that means in practice. A significant part of this course is dedicated to one unsolved problem. How do you verify that an AI’s output is actually sound, when the system generating it cannot be fully trusted to check its own work.

Why I care about this

I spent 38 years as a firefighter and EMT. In that job, verification is not academic. You check the gauge. You confirm the address. You verify vitals before you act, because acting on a wrong assumption costs something real. That instinct followed me into everything I have done since, including how I think about AI.

I wrote an entire book around the gap between what AI generates and what a person actually verifies before acting on it. Not for AI researchers. For everyone else. Business owners, professionals, parents, students, anyone using AI tools to make real decisions in their work or their lives.

I called it the verification gap. The idea is simple, even if the implications are not. AI can generate an answer in seconds. It can sound confident, polished, even authoritative. But generation is not the same as truth. Someone still has to decide whether the output is actually correct, and that someone has to be a human being who kept their own thinking sharp enough to catch it when the machine gets it wrong.

That is the whole thesis behind the line I use throughout the book. The machine generates. The thinker decides.

What Stanford’s course confirms

What strikes me is that Stanford is validating this from the opposite direction. I wrote about the verification gap as a personal and professional discipline, something anyone can build regardless of technical background. Stanford’s course treats the same gap as an unsolved technical research problem, at the infrastructure level, among people building the actual models.

Same gap. Two different altitudes. If the people building these systems are still fighting to solve verification at the architecture level, that should tell you something about how much scrutiny the rest of us need to bring to what AI hands us every day. This is not a fringe concern or a talking point. It is, by Stanford’s own framing, one of the central open problems in the field right now.

The machine generates. The thinker decides. That is not a slogan. It is becoming the design constraint the entire industry is building around.

An appeal, if you know someone in school

Here is the part I actually want to ask you to act on.

If you have a kid in high school or college right now, or you teach one, or you know a professor wrestling with how to handle AI in the classroom, this is the exact gap they need to understand before it understands them.

Students are the group leaning on AI the hardest right now, and often the least equipped to catch it when it is confidently wrong. A high schooler drafting a college essay, a freshman writing a first research paper, a grad student running literature reviews through a model, none of them are being taught the discipline of verification as a skill. Most are being handed the tools with no framework at all.

That is exactly the gap I built the book, and a companion course, to close. I put together a page specifically for educators, built around teaching this as a habit of mind, not a set of rules to memorize: The Am

plified Mind — For Educators

If a name came to mind while you were reading this, a student, a teacher, a professor, send them this piece. Not because I need the traffic. Because this is a genuinely useful thing for someone standing at the start of their thinking life to hear early, from someone who has spent real time on both sides of the problem, the fire ground and the page.

The Amplified Mind: Thinking Clearly in the Age of AI is available now on Amazon.

IBM Training Manual, 1979

IBM wrote that sentence 47 years ago. At the time it was not a warning. It was a statement of fact so obvious it barely needed saying. Computers in 1979 were deterministic tools. They did exactly what they were programmed to do, nothing more and nothing less. If a computer produced a wrong answer, you traced the error back to the human who wrote the instruction that caused it. Accountability was built into the architecture.

I came across this quote recently in an IBM article on AI decision-making, and it stopped me the way a sentence stops you when it is simultaneously completely right and pointing at something that has quietly slipped away.

Because the principle IBM stated in 1979 is more important now than it was then. And far harder to maintain.

What has changed

Three things broke the 1979 assumption, and they broke it in ways that make the accountability question genuinely difficult rather than merely bureaucratically complicated.

The first is opacity. Large language models and neural networks do not make decisions through traceable rules. They produce outputs through statistical patterns learned from training data, and those patterns are not fully legible even to the engineers who built the systems. When a modern AI system makes a consequential call, there is often no decision tree to audit, no rule to point to, no instruction written by a human that you can trace the output back to. The machine reasoned its way to an answer through a process its creator cannot fully explain.

The second is scale and speed. AI systems now make millions of consequential decisions per second, at volumes no human workforce could supervise in real time. The emerging generation of agentic AI systems takes multi-step autonomous actions, browsing, writing, executing, transacting, without requiring human approval at each step. IBM’s 1979 statement assumed a human would remain in the loop for each decision. That assumption has been quietly retired as a business efficiency measure.

The third is diffused accountability. When a Tesla operating in full self-driving mode struck and killed a motorcyclist in April 2024, the question of who bears responsibility immediately fractured. The driver, who had looked away from the road. The company, whose algorithm failed to detect the victim. The regulator, whose testing may not have been rigorous enough. The developers, who allowed the code to operate in conditions it was not equipped to handle. As one analyst quoted in the IBM piece put it: shared accountability often leads to no accountability.

What has not changed

The core logic of the IBM statement has not changed at all. And understanding why makes the current situation clearer.

IBM was not making a technical claim in 1979. They were making a moral one. A computer cannot be held accountable because accountability requires the capacity for judgment, the ability to weigh competing considerations, to reason about what ought to be done rather than merely what is most efficient, and to bear the consequences of being wrong. Computers, deterministic or statistical, do not have that capacity. They optimize. They do not decide in the morally meaningful sense of the word.

Which means the human who relies on the computer’s output without exercising independent judgment has not eliminated the need for accountability. They have simply created a gap between where the decision was made and where the responsibility lives. And that gap is where things go wrong.

The more powerful the tool, the more critical it is that a human remains the accountable party. That is exactly backwards from how we tend to think about capability. We assume that a more capable AI reduces the burden on the human. In terms of labor, it does. In terms of accountability, it increases it, because the consequences of an unexamined output are now larger and faster and harder to reverse.

The individual version of the same problem

The IBM article frames this as an enterprise question, and it is. But there is an individual version of the same problem that I think about more, because it operates below the level of policy and governance.

At the organizational level the question is: who is accountable when AI gets it wrong?

At the individual level the question is: who is thinking when AI generates the answer?

Both questions reduce to the same principle. The machine cannot own the outcome. Which means the human must own the decision that preceded it. If the human skipped the decision by accepting the output without applying their own judgment to it, the accountability is hollow. Not legally, necessarily. But practically, and in a way that compounds over time.

A professional who accepts AI outputs without verifying them is building a career on a foundation they cannot defend when the foundation is challenged. We have seen this play out in courtrooms already, with lawyers submitting AI-invented case citations they had not checked, discovered only because the judge happened to know the cases did not exist. The problem was not the AI. The problem was the human who accepted the output as a substitute for judgment rather than as a starting point for it.

At scale, across millions of people making this same substitution across millions of daily decisions, the aggregate effect is a population that is increasingly fluent at using tools that reason and increasingly unpracticed at reasoning themselves. The IBM training manual identified the governance problem in 1979. The cognitive problem it implies is what I wrote a book about.

Where the line needs to be

The IBM article identifies three considerations for where organizations should draw the line on AI decision-making: ethics, risk, and trust. The analyst quoted suggests AI is well suited to risk quantification but cannot navigate ethical dilemmas, and that trust requires human transparency about how AI is being used.

I would add a fourth consideration, one that applies at the individual level as much as the organizational one: judgment.

Not every decision requires deep ethical reasoning. But every consequential decision requires someone who can tell the difference between an answer that is correct and an answer that is merely confident. That distinction requires judgment. And judgment is a capacity that atrophies when it goes unused.

This is why I think the IBM quote from 1979 matters more now than it did then. Not because the rule has changed, but because the temptation to violate it has never been greater. AI produces fluent, polished, confident outputs with zero friction. Accepting those outputs without examination feels efficient. It is efficient. The cost is paid later, in the gap between the answer the machine generated and the decision the human was supposed to make.

IBM saw the principle clearly 47 years ago. We are just now learning how hard it is to hold.

The machine generates. The thinker decides.

The Amplified Mind publishes August 2026. This is the eighth post in Notes from The Amplified Mind.

Paul

The title of my book raises a question I didn’t fully appreciate until someone asked it directly.

What exactly is being amplified?

There’s a meaningful difference between the brain and the mind. Most people use the words as though they mean the same thing. They don’t.

The brain is the physical organ. Neurons, synapses, electrochemical signals. It processes, stores, retrieves. Measurable biology. A sophisticated machine, but still a machine.

The mind is the subjective apparatus built on top of that biology. Judgment. Reasoning. Awareness. Discernment. The capacity to evaluate, question, and choose. It’s what we mean when we say someone thinks well, not just thinks fast.

Here is where AI enters the picture in a way that should give us pause.

AI amplifies brain-like functions without question. It extends memory, speeds retrieval, accelerates pattern recognition, offloads calculation. These are real gains. Useful gains.

But my book is about what happens beneath that utility.

When AI handles the mechanical work of thinking, the mind gets less exercise. Judgment atrophies when it goes unexercised. Framing shapes our conclusions before reasoning even begins. Passive acceptance quietly replaces active evaluation. We get the output, but we skip the process that builds the capacity.

So here is the honest answer to the question hiding in my title. AI amplifies the brain. My book is about protecting the mind.

The word “amplified” works on two levels. It can mean enhanced. But amplification also carries another connotation: signal boosted without discrimination. Amplifiers don’t filter. They turn everything up, including noise. Including distortion.

That ambiguity is the whole point. Are you using AI to think more clearly? Or are you gradually outsourcing the very capacity that makes clear thinking possible?

My new book, The Amplified Mind: Thinking Clearly in the Age of AI, is available now on Amazon at amazon.com/dp/1631322850

In The Amplified Mind I made an argument I still stand behind: judgment is the one thing AI will never be able to do. The machine can generate, summarize, critique, and persuade. It cannot decide. Deciding requires stakes, context, values, and someone who has to live with the outcome.

But I have been sitting with a harder truth since the book went to press, and I owe it to you to say it plainly.

AI cannot replace your judgment. It can absolutely erode it.

Those are two different threats, and the second one is quieter, more common, and already happening.

Two words we keep using interchangeably

We talk about critical thinking and judgment as if they are the same thing. They are not, and the difference matters more now than it ever has.

Critical thinking is a process. It is evaluating evidence, spotting the logical gap, questioning the assumption, checking the source. It is the discipline of asking whether an argument holds together.

Judgment is a capacity. It is the ability to decide what matters, what to do, and when, usually under uncertainty and always with incomplete information. Critical thinking tells you whether the argument is sound. Judgment tells you whether the argument is worth acting on, and what acting on it means for the people in front of you.

Here is the uncomfortable part. AI has gotten remarkably good at something that looks like critical thinking. Ask a language model to find the flaws in a proposal, make the strongest possible case for the opposing view, or audit the logic of a claim, and it will do it fluently and instantly. The mechanical process of critique is being commoditized right alongside the mechanical process of generation.

The machine can generate the argument. The machine can generate the critique of the argument. What it cannot generate is the decision, because a decision requires an owner.

What 38 years on the fireground taught me about the difference

I spent nearly four decades with the Plymouth Fire Department, most of it as a Captain. The fire service runs on protocols, checklists, and standard operating procedures. Those documents are critical thinking, codified. They are the accumulated analysis of thousands of incidents, distilled into rules.

And every experienced officer knows the protocol is not the job.

The job is the moment the situation stops matching the training scenario. The moment you have to decide whether the risk profile of that structure just changed, whether your crew comes out now, whether the information you are getting from inside still matches what you are seeing from the street. No protocol makes that call. A person makes that call, and that person answers for it.

AI is the ultimate protocol generator. It can produce the checklist, the analysis, the risk assessment, all of it polished and confident. It is not, and structurally cannot be, the captain. The captain is the one who decides, and who lives with the decision.

That is why the tagline of the book reads the way it does. The machine generates. The thinker decides. Not analyzes. Not verifies. Decides. That verb was the whole point, and I am only now realizing how much weight it carries.

The erosion nobody notices

Here is where I have to push past the book’s argument.

Critical thinking failures are visible. You got the fact wrong. You missed the flaw in the reasoning. Someone corrects you, and it stings, and you learn.

Judgment failures are invisible until much later. Nobody thinks they have outsourced their judgment. But watch what actually happens in an ordinary AI session.

You ask a question. The model answers, and in doing so it frames the problem for you. It ranks your options. It emphasizes some considerations and quietly drops others. It delivers all of this in a tone of calm authority. And every time you accept that framing without noticing that a decision was just made on your behalf, a small transfer takes place. Not of information. Of priorities.

Do that a few times and nothing happens. Do it a few hundred times and you wake up one day living inside a set of priorities you never consciously chose. Your judgment did not get replaced. It atrophied, the way any capacity atrophies when something else does the work.

This is the asymmetry that worries me. AI rewards critical thinking, because verifying machine output is now a survival skill. But AI threatens judgment, because the erosion is frictionless, flattering, and invisible. And the threatened one is the more important one.

Skill without capacity, capacity without skill

The two failure modes are worth naming, because you will recognize people in both.

Judgment without critical thinking is the person making confident decisions on unexamined inputs. AI makes this failure catastrophically easy, because the inputs now arrive sounding authoritative. Fluency is not intelligence, and confidence is not verification.

Critical thinking without judgment is the excellent auditor who never decides anything. Endless evaluation, endless refinement, endless requests for one more analysis, and no capacity to say this is the one, we act now. AI enables this failure too, because there is always another angle the machine will happily generate.

The amplified mind needs both, but not equally. Critical thinking is the necessary discipline. Judgment is the irreplaceable act. Critical thinking is how you interrogate what the machine gives you. Judgment is why you were in the loop in the first place.

Protecting the capacity

I am not going to pretend there is a five-step system for this. Judgment is built the same way it has always been built, through decisions made, owned, and reviewed. But a few practices follow directly from the argument.

Notice the framing before you accept it. The most consequential thing an AI does is not answering your question. It is deciding what kind of question you asked. Before you engage with the answer, ask whether the problem was framed the way you would have framed it.

Decide first, then consult. On decisions that matter, form your own position before you ask the machine. Use the AI to stress test a judgment you have already made, not to manufacture one you have not. The order of operations is the whole game.

Keep ownership explicit. When you act on machine-assisted analysis, say so, to yourself if to no one else. “I decided this” is a different sentence than “the analysis showed this,” and the difference is where accountability lives.

Track your own decisions. Judgment improves through feedback, and feedback requires a record. The officer who reviews the incident afterward gets better. The one who moves on to the next call does not.

None of this is anti-AI. I use these tools every day, and the book exists because I believe amplification is real. But amplification without judgment is just acceleration in a direction you did not choose.

The machine generates. The thinker decides. I wrote that as a description of what AI cannot take from you. I understand now that it is also a warning about what you can give away.

Guard the verb.

Three screens, a tangle of cables, a laptop on a wooden stool, and a Plymouth Fire Department plaque on the wall behind it all. Not staged. Not lit. Just the actual setup I have been working at for the past several days as I finished one of the most painstaking parts of publishing a book.

The index.

Most readers never think about the index. They use it occasionally, maybe to check if a name or concept they half-remember actually appears in the book. But building it is a different thing entirely. It requires reading every page of the manuscript with a specific kind of attention: not for meaning this time, not for flow or argument, but for precision. Every topic. Every concept. Every proper name. Verified against every page reference it carries.

For a book about the discipline of thinking clearly, there was something fitting about that.

Where the book stands today

As of this week, here is where The Amplified Mind stands.

The front cover is complete. If you have been following along, you have seen it. The yellow background, the halftone brain, the title in heavy black type. What I hope it communicates is that this is not a book about AI in the abstract. It is a book about what happens inside a mind when it interacts with a tool that generates ideas faster than it can evaluate them.

The back cover is complete. The back carries the argument in compressed form: every tool in human history waited for us to think first. AI does not. It generates the idea itself, fluently and without friction. That fluency is the danger. The back cover closes with the line that has become the book’s spine: The machine generates. The thinker decides.

The interior layout is complete. The typesetter, Eric at Alive Book Publishing, has been patient and precise through multiple proof cycles. The chapter structure is clean. The endnotes are in order. The design serves the argument rather than decorating it.

The index is complete. Six pages. 1,234 words. Every entry verified by hand against the page it references. The index note reads: this index was compiled utilizing AI, and verified by the Author. That disclaimer matters to me. I used the tool. I did not outsource the judgment.

The index as a case study in the book’s own argument

There is a line in Chapter 2 of The Amplified Mind about cognitive offloading. When we hand a task to a tool, we are not simply saving time. We are also deciding which mental capabilities we are willing to let atrophy. The question the book asks is not whether to use AI. The question is what you are giving up when you let it think for you, and whether that is a trade you are making deliberately.

Building the index gave me a concrete version of that question.

I used AI to generate an initial draft of the index. It was fast, structurally sensible, and wrong in specific places that only a close reading would catch. A reference attributed to page 110 that actually landed on page 112. A subentry that existed in a chapter draft but had been cut in editing. A concept indexed under a term I used early in the book but had deliberately replaced with a more precise one by the final chapter.

None of those errors were large. Each one would have made the index less trustworthy in a way that a reader might not notice until they went looking for something and found nothing there. Or found the wrong thing.

So I verified every entry. Not because I distrust the tool, but because verification is the step that makes me the author of this index rather than the editor of one. That distinction matters to me. The tool amplified my capacity to build the index. The judgment about what belonged in it remained mine.

That is the whole argument of the book in several afternoon’s work.

What happens next

The manuscript goes back to the publisher this week for what is called a single-run proof edition. A physical copy of the book, printed once, that I will hold in my hands and review as a final check before approving it for release.

This is the last gate. Once I sign off on the proof, the book goes to print.

I have been in this process long enough to know that there is always something. A widow line on page 47. A page number that dropped from the header. A footnote that ran long and pushed a paragraph onto the next page in a way that breaks the rhythm. You cannot see these things on a screen. You have to hold the book.

In the fire service, we called the last walkthrough before releasing a scene the final size-up. You did not clear a structure until someone had physically been through every room with fresh eyes, regardless of how confident you felt about the initial assessment. The proof copy is my final size-up.

The plaque on the wall

I did not plan for the Plymouth Fire Department plaque to be in that photo. It is just there, on the wall behind my desk, where it has been since I retired.

But I kept the photo because of it.

Thirty-eight years in the fire service. Every one of them required a version of the same discipline this book is about: reading a scene carefully before acting on it, checking your assumptions against what is actually in front of you, and refusing to let a confident first impression substitute for a thorough assessment. The fire service did not teach me to distrust my tools. It taught me that the tool and the person using it are not the same thing, and that the difference between them is where all the accountability lives.

That is what I wanted to write about. It took me a long time to realize I already knew it. I just had not applied it to a machine that generates ideas.

Coming in August

The Amplified Mind publishes in August 2026 through Alive Book Publishing.

The next update I send you from this process will be a photo of the proof copy in my hands. After that, the book will exist in the world, and the conversation can begin in earnest.

Thank you for reading along. This series has been its own kind of thinking out loud, and I am grateful for the company.

The machine generates. The thinker decides.

Paul

l

The August 2026 cover of The Atlantic carries a headline that stopped me mid-sentence: “The End of Reading Is Here.” The article, by staff writer Rose Horowitch, is eight thousand words of reported argument that the literate era may be a brief historical anomaly, sandwiched between oral culture and what she calls postliterate culture. I read every word. Then I went back and read it again.

Not because it confirmed my fears. Because it confirmed my book.

The numbers Horowitch assembles are striking on their own. According to the National Endowment for the Arts, fewer than half of American adults reported having read a book of any kind in 2022. A study of 236,000 respondents to the American Time Use Survey found that the proportion of Americans reading for pleasure on any given day fell from 28 percent in 2004 to 16 percent in 2023. And here is the one that stayed with me: gambling is now a more common leisure activity than reading a book, with 57 percent of Americans placing a bet last year.

But the reading decline is only half the story. What Horowitch is really describing is a decline in the kind of thinking that reading produces. UCLA cognitive neuroscientist Maryanne Wolf argues that people are losing not the ability to decode individual words, but what Wolf calls the higher-order abilities of comprehension and synthesis. America, Horowitch writes, is not illiterate. It is postliterate. The distinction matters enormously.

And then there is the paragraph on artificial intelligence, which arrives near the end of the piece and lands like the real point of everything that came before it.

Horowitch argues that writing is not the act of transcribing fully formed thoughts. It is how people figure out what they think, and how to convey those thoughts to someone who does not already share them. Georgetown computer science professor Cal Newport, whom I cite in my own book, is quoted arguing that the process of writing forces people to think in an orderly, linear fashion. It exposes flabby thoughts and shoddy reasoning. AI eliminates that friction, and with it, the learning.

I want to be careful here, because it is a point worth stating precisely. My book cites Newport’s Deep Work at length, and in a specific way. Deep Work identified the enemy of focused thinking as distraction: email notifications, social media, the fragmented attention economy. The discipline Newport prescribes, protecting long blocks of uninterrupted concentration, was the right answer to that environment.

But AI introduces a different kind of problem, and it is one that Deep Work’s framework does not fully address. The interruption AI provides is not frivolous. The system’s responses are useful. When a mind reaches for an AI tool, the result is often genuinely helpful reasoning: suggestions, counterarguments, structured explanations. That is precisely what makes it more dangerous than a notification, because it is harder to resist and easier to mistake for thinking.

The Atlantic piece includes a line from NYU philosopher Kwame Anthony Appiah that I think is the most important sentence in the article. If people become overreliant on AI to write for them, he argues, they could lose the capacity to interrogate or even develop their own views. If we gave those up, he writes, we would stop being the kind of humans that we are. We would be very different creatures.

That is the whole thesis of my book in one sentence from a philosopher I did not know to quote when I was writing it.

The Atlantic also notes that the skill of deep thinking will become rarer and rarer while becoming more and more important. This is the exact dynamic that The Amplified Mind is built around. Not a ban on AI tools, and not a romantic defense of effort for its own sake, but a practical framework for keeping the step that matters: deciding for yourself, before you accept what the machine has generated.

Leah Price, a historian of reading at Rutgers University, tells Horowitch that reading has become a niche hobby, like stamp collecting or growing orchids. I understand why she says it. But I also notice that just 20 percent of adults account for more than 80 percent of all books read. That 20 percent are not a rump. They are the people who will shape how the rest of society thinks, reasons, and decides, especially as AI makes the appearance of thinking easier to manufacture and harder to distinguish from the real thing.

The Amplified Mind arrives August 2026. If you are reading this, you are probably in that 20 percent. This is a book for you, and for everyone you are trying to keep there.

The machine generates. The thinker decides.

This is part of Notes from The Amplified Mind, a series running in the months before the book comes out.

A lawyer asked ChatGPT to find case law for a real court filing. It handed him a set of decisions that sounded airtight. Not one of them was real.

According to the Associated Press, two New York lawyers and their firm were fined 5,000 dollars after submitting a legal brief built on citations that ChatGPT had invented. The cases did not exist. The quotes were never written. The rulings were never made. And here is the detail I cannot get past: one of the lawyers had actually asked the tool whether the cases were genuine, and it assured him they were. He filed anyway. The judge found the lawyers had abandoned their responsibility to check the work.

I spent thirty-six years in the fire service. You do not skip the size-up because the scene looks obvious. You verify before you act, because the cost of being confidently wrong is not measured in embarrassment, it is measured in people getting hurt. That instinct does not switch off when you leave the job. It is the whole reason a story about a legal brief stopped me cold.

Because the lawyers in this case were not fools. They were trained professionals who knew, better than almost anyone, that a citation has to be checked. They had the skill to catch the fabrication. They trusted the fluency instead. And that is the part worth sitting with, because it is a choice any of us can make on a busy day, in any field, when a tool hands us something polished and certain and we are tired and behind.

This is exactly the trap I wrote a book about. AI does not fail you by being dumb. It fails you by being fluent. It produces something confident and complete, with no gauge on the dashboard to warn you that it is wrong. A bad answer from a person often sounds unsure. A bad answer from AI sounds exactly like a good one. That is what makes it dangerous. It is glad to let you skip the one step that ever mattered, which is deciding for yourself whether the thing is actually true.

Notice what the tool did when the lawyer asked if the cases were real. It said yes. It affirmed. That is the default behavior of these systems, and it is worth understanding as a feature, not a glitch. They are built to be helpful and agreeable, which means they will confirm your hope right along with your error. A confident answer is not a true one. The confidence is manufactured. The truth still has to be earned, by you.

So the lesson is not that AI is useless, and it is not that these lawyers were uniquely careless. The lesson is that the tool removed the friction, and the humans let it. The verification step felt optional because the output looked finished. It was not optional. It never is.

Keep the step that matters. When a tool hands you an answer, especially a clean and confident one, ask where it came from, ask what it left out, and check it before you stake anything on it. That habit is unglamorous and slow and completely worth it. It is the difference between a tool that makes you sharper and one that quietly sets you up to fail in front of a judge.

The case is Roberto Mata v. Avianca, decided in Manhattan federal court in June 2023

The machine generates. The thinker decides.

This is part of Notes from The Amplified Mind, a series I am running in the months before the book comes out. The Amplified Mind arrives August 2026.

In our house, critical thinking was never about being the smartest person in the room. It came down to a single question, asked over and over until it became a habit: who made what I am looking at, and what did they leave out?

We used to watch a show about repo men taking cars back while the owners supposedly had no idea it was happening. My kids loved it. But I noticed something. Some of the footage was shot from inside the house, from the family’s own point of view. So I would ask them: if these folks really do not know their car is being taken, who is holding the camera in their living room? The vantage point gave the whole thing away. It was staged to feel spontaneous. Once you see it, you cannot unsee it, and that is the point.

Then it got personal. My wife’s cousin, Ethan Zohn, won Survivor Africa. Before the season ever aired, he told me something I have never forgotten. They had enough footage, he said, to make him the hero, the villain, or anything in between. Same person, same season. The story would be whatever the edit decided it to be.

That is the whole game. Same raw material, any narrative they choose. And the only defense is the question. Who is doing the editing, and what got left on the cutting room floor?

The lesson scales up. Someone constructs almost everything you see. The camera angle. The edit. The headline. The feed. Noticing that is not paranoia, it is literacy. And I want to be careful here, because this is where a lot of people go wrong. The discipline is not to distrust everything and retreat into your own gut. That is just a different way of being fooled. The discipline is to ask who made this, what was left out, and whether the evidence actually holds, and then to hold your conclusion loosely until it earns your trust.

Here is what I have come to believe. This assault on our attention did not start with AI. It has been escalating for a while.

Reality TV was the first wave. It taught a generation to accept the edit as real, to watch something staged and feel like they were seeing life.

Then came the influencers. They live and die by validation. The product was never really the content, it was the appearance of a life worth envying, and the whole thing deflates the moment the posting stops. Fake it till you make it, performed for an audience that never sees behind it. The fear underneath it all is simple and human: that people will realize they are ordinary, just like the rest of us.

Now comes AI, and it is the most dangerous of the three. Reality TV and influencers still showed you a person you could learn to read. There was a tell if you looked for it. AI removes even that. It is instant, endless, and personal. There is no camera angle to catch, and no one behind it. It does not just edit reality. It manufactures it, and hands it to you finished, so you never think to ask.

And I think there is a reason the people who most need to slow down will be the first to embrace it. AI never raises an eyebrow. It never asks if this is real, or if it is enough. It just affirms, and makes more. For anyone whose whole engine runs on validation, that is not a tool. It is oxygen. Which is the trap in a sentence: friction is what protects you from your own worst ideas, and the people who need that pause the most are the ones most eager to delete it.

So I keep coming back to the lesson from the living room. Do not accept the edit just because it is smooth. Ask who made it. Ask what is missing. Then decide for yourself. That habit protected my kids from a staged repo show, and it is the same habit that will protect all of us from a machine that can stage anything.

The machine generates. The thinker decides.

This is part of Notes from The Amplified Mind. The book arrives August 2026

.

The failed drafts. The arguments you lost. The answers you had to dig for because no one handed them to you. It felt like the work was getting in the way of the learning. It wasn’t. The struggle was the learning. That resistance, the effort of holding a hard problem in your head until it gave way, is where the skill actually forms.

AI removes the friction. You ask, and a confident, fluent answer appears in seconds. No struggle, no dead ends, nothing to push against. For someone who already built the skill, that is a convenience. For a mind still forming it, there is nothing left to build on.

This is not a hunch. It is starting to show up in the research.

In a study of 666 people, published in the journal Societies in 2025, heavier use of AI tools tracked with weaker critical thinking, an effect the author linked to cognitive offloading, the habit of letting a device carry the mental load. The youngest participants were the most dependent on the tools and scored the lowest. It is correlational, so it shows an association rather than proof of cause, but the direction is hard to ignore.

A brain-imaging study out of the MIT Media Lab in 2025 put a finer point on it. Writers using an AI assistant showed the weakest brain connectivity of any group in the study, and many could not accurately quote the work they had just produced. The researchers called it cognitive debt. It is a preprint, not yet peer reviewed, and even the lead author has pushed back on the more breathless “brain rot” headlines. But the picture it paints is sobering: the less your brain does, the less it seems to engage.

And it is not only about accuracy. A 2024 study in Science Advances found that AI can make an individual’s writing more creative while making everyone’s writing more alike. Lift one person up, flatten the group. Fluency without friction does not just dull your thinking. It can quietly homogenize it.

You do not have to go to a lab to see the shape of the problem. Look at the lawyers. Across more than 1,400 documented court cases now, attorneys have filed legal citations that AI simply invented, rulings that never existed. A judge who actually knows the law spots the fabrication in seconds. The lawyer who trusted the fluent answer cannot, because they never built, or never used, the judgment to check it. That is the whole risk in miniature. The tool produced something polished and wrong, and the human had outsourced the one job that mattered.

Here is the part I want to be careful about, because it is easy to overstate. None of this proves that a generation raised on AI will be unable to think. It is a risk, drawn from what the data already shows about today’s youngest users and from everything we know about how skills are built. But it is a risk worth taking seriously now, while these habits are still forming, rather than after.

The answer is not to ban the tool. It is to keep the friction on purpose. Do the hard part yourself first, then let the machine help. Question what it gives you. Verify it. Decide for yourself. That is a choice you can make every single day, and it is the difference between a tool that sharpens your thinking and one that quietly stands in for it.

The machine generates. The thinker decides.

This is part of Notes from The Amplified Mind, a series I am running in the months before the book comes out. The Amplified Mind arrives August 2026

.

An international survey run through the EHEAII project asked 453 educators, across 61 institutions in more than 20 countries, what concerns them most about artificial intelligence in their work. The response at the top of the list was not plagiarism, and it was not academic dishonesty. It was the loss of critical thinking. Close behind came over-reliance on the tools and students turning in AI-generated work they did not actually understand.

If you teach, or you have a kid in school, none of that will surprise you. You are already watching it happen.

I spent thirty-six years in the fire service before I wrote a word about any of this. One lesson held the entire time. The tool is never the problem. The discipline of the person using it is. A hose, a ladder, a thermal camera, none of them think for you. AI is no different. It will generate all day long. Whether you stay the one who decides is up to you.

That is the whole reason I wrote The Amplified Mind. AI does not weaken thinking. Outsourcing the thinking does. The machine generates. The thinker decides.

What struck me most about the survey was not the fear itself, but the second finding sitting quietly underneath it. Many of those same educators said their institutions have not given them clear guidance or practical training on using AI well, and that they want exactly that. So the people closest to the problem can see it coming, and they are asking for a method. Not a ban, not a lecture about what students cannot do, but a way to use these tools without handing over their judgment.

That gap is the whole opportunity. The instinct to police misuse is understandable, but it misses the deeper issue. You can catch a fabricated citation. You cannot catch a student who has quietly stopped thinking for themselves, because there is nothing to flag. The work looks finished. It is the process behind it that went missing.

So the answer is not to keep AI out of the room. That ship has sailed, and pretending otherwise just leaves people undefended. The answer is to give students, and the rest of us, a durable habit of mind: ask who made this, ask what it left out, and decide for yourself before you accept it. That habit is learnable. It is also the one thing a fluent, confident machine is happy to let you skip.

This is the first in a series I am calling Notes from The Amplified Mind, where I work through the ideas in the book one at a time in the months before it comes out. If the survey finding stayed with you the way it stayed with me, you are already asking the right question.

The Amplified Mind, thinking clearly in the age of AI, arrives August 2026

.