The financial industry understands this too. It sells personas before it sells products.

One thing stayed with me. As a senior advertising executive at Ogilvy UK, Sutherland favoured hiring psychology students over business school types. The logic was clean: business students are trained to optimize what already exists, while psychology students are trained to understand why anyone wants it optimized in the first place.

Psychology graduates were also less demanding in terms of remuneration. The MBA would celebrate this as a cost-efficient operating model. Labour arbitrage opportunity. Optimized human capital allocation. The vocabulary of celebrating one’s own irrelevance.

Sutherland’s approach was not to start with the product. It was to understand the behaviours people recognize in themselves, then build the persona they would want to step into. That is not a master plan in the cinematic sense. Most marketing plans fail. The ones we remember are the ones that survived long enough to look inevitable. They tickled the right fancy, gathered the right crowd, and reached a tipping point.

The financial industry has its own frames: the prudent saver, the responsible diversifier, the believer in the future. These personas were fine-tuned for us to recognize, admire, then aspire to. The effect is to create a quiet psychological discomfort when we do not reach for one. This is marketing.

That matters because finance is rarely sold as greed. It is sold as responsibility. Development. Stability. Diversification. Access. A better future. The tragedy starts when those words become doors into arrangements that extract from citizens while telling them they are being taken care of.

For a real financial expert, diversifying a portfolio is serious work. There are research papers, mathematical models, considerations of sectors and businesses and the interconnections between them. The math is complicated and the interconnectivity is overwhelming. It feels like science, but diversification is often art justified with numbers. This is why it is often called investment philosophy, not investment science.

Warren Buffett is obviously a real person, but he is also a persona. The first is in Omaha studying companies. The second is in your inbox selling you a newsletter, a fund, a book, a course, a feeling about being smart with your money. Both exist. Only one is selling

Beware, personas are not descriptions: they are doors presented to you. They sell you an idea, a vision, a finished version of yourself. The pitch ends when you open the door. By the time you ask, you are already inside.

This essay uses personas to deconstruct the illusion. A teacher in Buenos Aires in 1985. A founder in Dakar in 2026. An engineer in the Bay Area in the near future.

Buenos Aires, 1985

In 1985, a teacher in Buenos Aires gets her paycheck. It is larger than the previous one, but barely enough to catch up to prices that seem to move faster than her salary. She experienced what is known as hyperinflation: the daily price increase of groceries, the unavailability of durable goods, the diminishing purchasing power of her pension, and the growing stack of bills in her hand. To illustrate, monthly inflation had just hit thirty percent, quietly eating her purchasing power before she could even deposit her paycheck.

Argentina had been prosperous in the early twentieth century, with per capita income among the highest in the world. Then the economy stalled in the 1930s, coinciding with the Great Depression.

The country needed to reinvest in its decaying infrastructure and modernize productive capacity. The capital required for this had long been expensive and tightly rationed.

The petrodollar windfall that changed everything

A defining moment came in 1973, when the oil-producing countries, under the OPEC cartel, quadrupled prices. Western banks filled with U.S. dollars: the petrodollars. The same price hike severely slowed developed economies and weakened domestic credit demand.

This will sound counterintuitive, but the petrodollars came in as a windfall for Argentina: the World Bank was now able to provide credit to finance the capital investment it needed to reinvigorate the economy. For context, the World Bank is governed by wealthy creditor nations and is instrumental in routing capital from surplus countries into debtor economies.

Banks holding petrodollars had a strong appetite to lend, or else the dollars would just be unproductive assets on their books. Argentina and other Latin American countries now qualified as borrowers at scale. The loans flowed because borrowing was relatively cheap and more specifically available.

Alas, the touted economic turnaround never materialized. In 1979, the U.S. Federal Reserve raised interest rates in its own fight against inflation. For a moment, rates rose well above 20 percent. The dollar rose with them. Argentina’s loans were payable in dollars, but more importantly, they were floating-rate loans tied to U.S. interest rates, meaning the payments were variable. The result: the cost of servicing them tripled. That was not part of Argentina’s budget, nor forecast. The peso had to be devalued to keep up. In 1982, Argentina could no longer pay back what it owed to foreign banks and defaulted. Mexico had paved the way months before.

As a side note, it is usually bad financial practice to service a debt in a currency other than the one you earn in. Argentina cleanly demonstrated why.

Then came the International Monetary Fund (IMF) to manage the default on behalf of the lending banks. The IMF had been created in 1944 as part of the Bretton Woods system, to help countries that ran out of foreign currency to pay their international obligations.

By the 1980s, its mandate had changed. The IMF was the institution that arrived after a default to convert a country’s unpaid debt into enforceable terms. The terms were consistent across cases: devalue the currency to make the country’s exports cheaper, cut public spending to free money for repayment, privatize state assets to generate dollars for reimbursement. Citizens then lost subsidies on fuel and food and electricity. Wages were compressed.

The logic is familiar from the game Monopoly: when the cash runs out, the assets start moving to whoever still has money.

Argentina was forced to resolve the situation because the alternative was being cut off from dollar liquidity entirely. Resolving the debt issue with the IMF was necessary before the World Bank could resume the lending program. Many infrastructure projects were significantly advanced but not yet completed. There was no alternate path other than economic isolation.

The teacher’s currency lost value because the country had borrowed dollars to build infrastructure that came in late, came in incomplete, or came in at many times the promised cost. The lending contracts required the involvement of specific international firms (consulting, engineering, construction, etc.). Unfortunately, the dollars did not stay in the local economy to create the growth that was supposed to repay the loan. The full debt stayed on the country’s books.

The real tragedy: the pension the teacher had paid into for thirty years melted away, along with her buying power. She had not borrowed money from a foreign bank. She had not seen a penny of it, while she would spend the rest of her life paying for it. She had invested her time in her country, and the country failed her.

Dakar, 2026

Senegal was supposed to be the counterexample: stable, democratic, investable, the kind of country where the diaspora could return home and build.

What follows is a live example of the same trap.

In 2024, a young entrepreneur returns to Dakar to launch a digital payments company. She had grown up in Senegal, studied abroad, and built her career in finance in Europe. She came home because the country seemed to offer a proper business environment for what she wanted to build. International firms wanting a presence in Africa had chosen Dakar as the stable ground to service the continent. The technology sector was booming. Talent was strong, both homegrown and returning. The economy was prospering. Public infrastructure was developing.

I was in Dakar in February 2026. I heard on the local news that Senegal was “in talks” with the IMF. I knew what that meant

The hidden number

Something happened in 2026 that drastically changed her business environment.

The newly elected President Faye’s promised public finance audit discovered seven billion U.S. dollars in hidden debt. The country’s true debt was now 132 percent of GDP, not the 74 percent that had been reported by his predecessor.

The hidden debt had not been invented. It had simply been spread across project loans, state enterprises, and off-budget accounts that the official figures never consolidated. It was deemed a conscious decision by the previous administration to underestimate the debt stock.

No good deed goes unpunished: the IMF froze its 1.8 billion dollar lending program. Senegal’s sovereign bonds dropped sharply. The president dissolved the government. That was the political consequence, which worsened the economic outlook.

For our young entrepreneur, the consequence immediately surfaced in the next funding conversation. Her investors were now applying a new risk discount. The country itself had not changed overnight. The audit had only made the hidden number visible. Her work, her time, her investment, her ambition, all now devalued by the market.

Same trap as in Buenos Aires, four decades later. The people inside the persona are doing the same work, providing the same value, but the market “corrected” their worth.

The Bay Area, in the near future

Sometime in the near future, an engineer in the Bay Area checks his retirement balance. The number is smaller than expected.

He works at Microsoft. He has been there ten years. His base salary is high by any standard outside the few zip codes around him. Every three months, part of his pay comes in company stock. His financial advisor told him to diversify, sell the stock as soon as possible, with the objective of not having all his eggs in the same basket. The advisor recommended some of the major index funds, including the S&P 500. He followed the advice.

The illusion of diversification

Let’s look at the S&P 500, the most popular index fund, which owns shares in roughly five hundred of the largest public corporations listed in the U.S. This investment vehicle simulates direct ownership in Microsoft (approximately 5 percent of the index). It also holds Apple, Alphabet, Amazon, Meta, Nvidia, and Tesla. When SpaceX goes public, it will not be in the S&P 500. The committee has refused to fast-track it... for now.

But Nasdaq did. The Nasdaq-100 will absorb it within weeks. Retail allocation absorbs it on day one. The system has multiple paths.

The S&P committee can still change its mind. The gatekeeper held the line, but the machinery routed around it anyway.

His employer is in the index. His employer’s customers are in the index. His employer’s suppliers are in the index. He diversified away from Microsoft by buying the universe Microsoft was inside. Expecting to diversify, his exposure simply changed shape. The marketing sells diversification with the boast of owning a share of corporate America.

From a strict financial perspective, this is not proper diversification by any account. Especially for our engineer who works for Microsoft. His mortgage compounds the problem. He bought at the height of the market at the highest rate in decades. The mortgage payment is fixed every month. The income that covers it is not. The base salary alone falls short. The stock covers the difference. If the stock falls, the mortgage does not.

A serious financial review would recognize the mismatch: variable income, fixed debt. Many of the people selling financial products to retail customers will not be bothered by that. They are not paid to do the analysis. They are paid to distribute the products that the financial analysis would have flagged for our engineer. This is exacerbated by self-directed investing apps that allow people to cut the intermediary. You spend weeks shopping for a new fridge, then move far more money in minutes because the persona inspired you.

Meanwhile, June 2026

This next part contains financial arithmetic, but it matters. Stay with me.

At the time of this writing, SpaceX is expected to go public in June 2026 at a targeted valuation of 1.75 trillion U.S. dollars. The company lost 4.28 billion U.S. dollars in the previous quarter. To justify the valuation at a 20x earnings multiple, SpaceX would need to generate roughly 87 billion dollars in annual profit. Comparable to Apple, one of the most profitable companies in human history.

No serious financial analysis would defend this valuation. A company losing billions is not casually assumed to out-earn Apple. The number does not have to be correct for the market to accept it. It only has to be large enough for the machinery to absorb it.

If you’re still with me but lost, what I’m trying to say is this: the planned SpaceX stock market entrance is just plain nuts on valuation.

The distribution tells the same story. Up to thirty percent of the offering is being directed toward retail investors. The historical norm is closer to five or ten. Fidelity lowered the account threshold to two thousand dollars. Robinhood, Schwab, SoFi, E*Trade, and other self-directed investing apps are part of the machinery. This is not incidental access. It is distribution design.

The marketing seduction also reaches across asset classes. Bitcoin dropped from 74,000 to a recent low of 60,000 U.S. dollars at the time of writing. I read this as exit liquidity rotation: retail selling one future-facing asset to buy the next one. This is my call, and I could be wrong. But the same persona that bought Bitcoin as the future of money is now buying space as the future of everything. The asset changed. The persona did not.

The institutional buyer sees a valuation. The household sees a door: space rockets, Mars, innovation, the future, all packaged in a company you were never destined to own. That is the frame. The arithmetic did not disappear. It was packaged.

Here is the mechanism: the index buys the stock once the stock qualifies for the index. Retirement contributions buy the index because the index is the prudent default. Retail buyers buy the story because the story was packaged for them. The inflow supports the price. The price supports the weight. The weight attracts more inflow. Closed loops do not correct themselves. They amplify until something outside the loop breaks the pattern. In plain terms, it is a microphone pointed at an amp at full volume.

The engineer thought he was prudent. He thought he was diversifying. He thought the house was an asset. The advisor told him each of these things and he followed the advice. The advice was what the financial industry tells every person in his position to do.

The financial industry has marketing departments too. Trillions in assets do not market themselves. The advice was correct inside the system that produced it. The advice did not show him that following it placed him at the centre of the system rather than at the edge of it.

They sold you 500 companies. They delivered concentration in seven, including the one that already pays your salary.

The engineer bought the persona. A financial professional should recognize the mismatch: fixed obligations, variable income, and a mispriced asset packaged as diversification.

No one reads the fine print. The people who build the persona know that. They make sure you recognize yourself before you understand the investment you have accepted.

The Collateral

The teacher in Buenos Aires, the founder in Dakar, and the engineer in the Bay Area are in the same position. Different decade. Different country. Different class. Same ownership. Argentina owed the bank, and the bank owned Argentina through the repayment terms. Senegal owed the fund, and the fund froze the programme through its compliance terms. The engineer gave his money to the S&P, and the magnificent seven own him through the allocation terms.

The system takes from one side and pays the other. The teacher's labour produced dollar surplus that serviced debt held by people whose accounts she would never see. The engineer's monthly buying gives the early investor billionaire an exit: market liquidity to sell his shares. The entrepreneur's value was repriced by the same market that benefited from her work. All three are inside the system. Only one side is paying for it to continue.

This is not a story about bad bankers in 1985 or bad index committees in 2026. The people inside the system ran the methods the system rewarded. No one planned the trap. It emerged from each actor responding to the incentives in front of them. It is not a market call. The unwind may arrive next year. It may arrive in ten years. The argument does not depend on timing. The argument is about who owns whom inside an arrangement that never tells anyone who owns whom.

The teacher, the entrepreneur, and the engineer thought they were doing what they were supposed to do, in good faith, inside the personas they had stepped into. Their value did not diminish, and their work did not become less real, but the terms changed around them. The system claimed the collateral, and their equity evaporated.

The system does not lend to the household so much as it borrows from it, calling caution participation until losses arrive and the collateral is taken away as if it were simply the normal settlement of accounts.

What stayed with me after listening to Rick Rubin and Rory Sutherland is that human decisions often begin before the numbers. The frame comes first. The numbers arrive later to explain why the choice felt reasonable. In finance, that gap becomes dangerous: the frame sells responsibility, while the contract organizes absorption.

That is why the persona matters. The system sells us the persona before it sells us the investment. We step into it because it feels safe and responsible: the prudent saver, the responsible diversifier, the believer in the future. The investment itself is not part of the pitch. By the time it becomes visible, the persona has already done its work.

We thought we were choosing a financial strategy. We were accepting a place in the marketing plan.

We are the collateral, and we are on auto-pay.

The opt-out is buried in the fine print.

It caught my attention because I do cold plunges too. It makes winter more tolerable.

The retreat’s founder has no medical credentials. He grew up on a farm and healed himself after cheating death in an accident the local doctor said was usually fatal. He offered the therapy to others, then proceeded to build a practice inspired by his protocol. The medical establishment dismissed him for years, but the clientele kept coming.

The business prospered. In a single year, the retreat hosted a reigning monarch, a duke and duchess, twenty-two princes, one hundred and forty-nine counts and countesses, and thousands of regular subscribers.

The retreat is called Gräfenberg. It opened in 1822. The founder was Vincenz Priessnitz. The mountains are in Austrian Silesia, in what is now the Czech Republic.

Sorry for the mind fuck. There is a point to all this.

When Priessnitz died in 1851, his protocol had spread across Europe. Twenty hydropathic establishments opened in Britain alone. American physicians made the journey to Gräfenberg and brought it home.

Priessnitz was the precursor of what became the wellness industry.

Some years later, a new wellness empire was born in North America, specifically in Battle Creek, Michigan. The Kellogg brothers started a sanitarium together, and one of them referred to himself as a health evangelist. The treatments were Priessnitz-inspired: hydrotherapy, vegetarian diet, fresh air, daily movement, rest. Their patients included Rockefellers, Fords, Edisons, and four US presidents.

You probably recognized the names Kellogg and Battle Creek. The Kellogg brothers developed the ready-to-eat breakfast cereal for their retreat, then commercialized it, launching another massive industry. Like a lot of kids my age, my breakfast reading was the cereal box. That is where I first learned about Battle Creek, Michigan.

Fast forward to today, the global wellness industry is worth nearly seven trillion dollars. Larger than the sports and pharmaceutical industries combined.

With social media, we are now exposed to more lifestyle and tech evangelists, selling their programs and philosophies, their services, their life-changing solutions. I call these figures Instrumental Evangelists. The word Instrumental matters: we keep building better tools, stronger systems, faster protocols. Even with what they promise, they do not give us better judgment. The Instrumental Evangelist is the person who preaches the instrument as if it could solve the deeper problem, maybe even transform you.

Big ideas tend to follow the same path. Someone notices something that works, and others begin to gather around the practice. At first, the exchange is alive. Literature circulates, recipes are shared, perspectives are written, and the thing itself remains the point.

Then the product is packaged, and a philosophy starts to form around it. The philosophy attracts followers; followers require doctrine; doctrine eventually needs a spokesperson. That is when the evangelist builds the church.

The business model wins, not necessarily the original intent. Apple is the clearest example.

Steve Wozniak started in the early 70s by drawing computers with a pencil and paper, even before he could afford to build them. In high school he had drawn hundreds of them, turning the work into a game. The rule he set for himself: can I design the same computer this month with fewer chips than last month? He did it over and over, redesigning machines he had no means to build, sharpening what he could see on the page. The redesigns were not meant for customers; they were not for a company he wanted to start. Leaving his dream job at Hewlett-Packard was never his plan. He designed them for the discipline of doing the work better than he had done it last month. Paper was cheap. Components were not.

He explained the reason later, in his own words: he wanted engineers to look at his designs and say, whoa, what a brilliant engineer. He wanted respect from his peers. He was not looking to build a following; he was building a piece of work where other engineers could recognize his talents.

When he walked into the Homebrew Computer Club in 1975, in the era of kit computers, he was doing the same thing he had done on paper in high school. He brought the Apple I schematics. He handed them out. He wanted to share with the members, show what he had accomplished.

Steve Jobs was at the meetings. He saw the same schematics, but he read them differently. He saw a product the world would buy. Where Wozniak saw a machine other engineers could understand, Jobs saw a philosophy that could be named around it. The two of them founded Apple Computer together in 1976. Wozniak designed and built the machines. Jobs built the image and the company.

They had a serious disagreement on what the next machine should be. Wozniak wanted the Apple II open, with expansion slots so other engineers could build on what he had made. Jobs wanted it closed. He wanted the experience controlled by the company. Wozniak won the first round. He did not win many after that.

Jobs saw the paradigm shift and built the church around it.

The name needed followers. The followers needed someone to convert them. But Apple needed developers to create software for its platform. In 1983, it hired Guy Kawasaki as a Software Evangelist. The title was lifted from the church without irony. Nobody at the company thought twice. The product was the gospel. The customer was the convert. The skeptic was the infidel.

Subscribe now

My father brought home an Apple IIe when I was eleven. I learned to work on a machine with expansion slots, a cover you could open to peek inside. The cool part was the computer games, as you only needed blank 5¼” diskettes and friends with the same machine. The games were cracked by someone who had figured out how, and we passed them around. The FBI never did find us, despite all their warnings. The culture trusted you to share what you had. Donne au suivant.

In hindsight, I was a user, not a convert. Later in the 1980s my parents bought a PC compatible. By then, the Mac was for designers and art students. PC was where the work was, and where university was, at least in my world. I spent the next decade and a half inside the Microsoft ecosystem, upgrading one version at a time, starting with Windows 3.11.

After years on PC, around 2008 I bought a Mac Pro tower. And when I bought my first smartphone I chose a Samsung Galaxy running Android over the iPhone. The iPhone was for consumers, fully packaged. I wanted a device for users, configurable from the inside. Samsung used to be that device. It is not anymore. Companies do not have principles. The business model wins.

What was once free became packaged, then sold back at a premium. Modern industry runs the same play. The cold plunge sells us back the swim, deconstructed, for a fee. Breathwork. Cold exposure. Sunlight exposure. Grounding. Vagus nerve stimulation. The slippers you forgot at home are the upsell.

Does exercise begin when we step on the treadmill? We park as close as possible to the gym entrance. The elders sound out of touch only because they remember the thing before the instrument replaced it. They used to walk to school by necessity. Now we drive to the gym to get some walking done, the instrumental substitute of a civilization that became privileged enough to remove walking from daily life.

Jesus of Nazareth preached a message of sharing, and Rome put him on a cross for it. Before we called them disruptors, empires called them threats. His followers kept the message alive. Eventually, a church was built around it, with Rome at its centre. Over time, the practice of the virtue became an instrument: a structure that could collect, interpret, administer, and preserve.

Will Durant wrote: “We repeatedly enlarge our instrumentalities without improving our purposes.” In other words, civilization accumulates comfort and capacity, while it does not accumulate wisdom.

That is the part worth remembering. Someone noticed something that worked. The instrument that came after was not useless. The trouble began when the instrument was allowed to stand in for the thing itself.

That is how the gospel of the instrument begins.

So we got talking. I told her I had an economics and finance concentration from McGill, class of 1994. I told her accounting and finance was a great combo, that whatever she ended up doing, not letting other people count your own money was an important life skill.

Moving on in the conversation, I said something a bit dramatic, not usually my style, but I was on a roll. I told her that most of the advanced finance theories I’d learned in my undergrad years had turned out to describe a world we don’t actually live in: Black-Scholes option pricing, CAPM, portfolio allocation, alpha and beta, and some other financial mumbo-jumbo. Still on the menu, even though the kitchen knows it’s gone bad.

Not wrong inside their assumptions. Just dangerous when you forget the assumptions are there. The Long-Term Capital Management implosion of 1998 was the first big crack, the 2008 crisis was the wider one. The main culprit in both cases was financial modelling built on the normal distribution, using the past as a predictor of the future.

Such severe events were never recorded before. Who could have known. The previous recorder probably said the same thing about the one before that.

Please allow me one more boring joke. I think it adds some context. A physicist, a chemist, and an economist are stranded on a desert island with a single can of food but no way to open it. The physicist suggests using force and gravity to crack it open. The chemist suggests heating it until the pressure bursts the lid. The economist simply says: “Assume we have a can opener.” (Gemini condensed the joke for me. The chain has been running for a long time.)

So where am I going exactly with my histoire de mon oncle? Nowhere fast, I bet you’re thinking. I currently run a brewery, work on a one-off project in Cameroon and Senegal, and write a Substack conveniently named I’m Not an Expert (psstt. this is essay nineteen), on topics including artificial intelligence, organisations, judgment, and so on. Not your usually interesting taproom conversation.

Subscribe now

Why I write

What was my trigger for starting writing in January? I turned off the TV. I had questions; nothing existential, not a midlife crisis, but about things I’d been carrying for years without spending adequate time to think them over: capital markets, Bitcoin as a refuge currency, quantum biology, institutional drift. Questions that had been sitting in a backlog while the TV ran, the idiot box they used to call it.

Writing turned out to be a great way to challenge my ideas, find out what I actually thought. Who would’ve thunk?

Last January, I was in Douala, Cameroon, working on an economic development mandate, spending the evenings in the apartment with some free time, and I just started writing what became my first published essay. As I write this, I’m back in Douala. They say fish is good for the brain. Good assumption.

Then came the harder question of whether to keep it private or publish. Taleb, one of my favourite modern thinkers, settled it for me, well, at least in my mind. Publishing is skin in the game, as he would say, and if you’re going to hold a position, put it where it can be challenged, improved, or discredited, because keeping it private is the same as not holding it. Yes, to answer the questions you didn’t think of asking: I sometimes feel bare in front of the room publishing all these essays about my personal experiences.

That’s it, I write for myself. The name of the Substack came naturally: I’m Not an Expert. Don’t expect answers. What you’ll get is someone trying to find out what he thinks, in public, where the thinking can be checked, discredited, fact-checked.

Why am I baring it all, spending my precious time writing this section? What comes next is going to sound like a critique of credentials. It is. And I have one.

Back to finance

As I was trying to say, my finance and economics degree taught me a lot of things that turned out not to hold. But not everything was useless. I did have some very useful courses, especially Real Estate Finance, even if I didn’t suspect at the time I’d end up as a real estate financial analyst and an investment banking associate. That course was basically maths, so no risk of challenging the equations, just make sure you pick the right one, and don’t mess up your assumptions.

Live and learn.

The late 90s and early 2000s was a good time to be in finance. The dot-com boom and bust, where the tech industry apparently had a meeting and decided that the laws of physics, P/E ratios, and NPVs no longer applied to them. The AI crowd is gospelling that their valuations should equal a meaningful share of the world economy, since AI is going to own it.

The plausibility machine

Speaking of parallel universes, my friend’s daughter is walking into a new one, and she doesn’t know it yet. University professors complaining that AI is destroying learning. Not all professors, but some “vocal” ones, the ones actually spending time writing (they probably have a book in the pipeline). Students completing essays in two hours, when it used to take twelve. TAs grading papers they know were written by ChatGPT, told by their supervisors not to fail anyone, ending up grading students on their ability to use the tool. As is my understanding, students are now clients, and the customer is always right. Right?

The educational apocalypse, as one observer called it.

I say the framing is wrong. The cheating panic and the hallucination panic, the worry that AI invents facts, are the same panic, and neither is really about AI.

Watch what the student is doing. You write an essay that sounds right inside the framework the professor is measuring. The framework rewards fluency in its own conventions. Your essay reads like an essay should read. The argument has the right shape. The citations are formatted correctly. The conclusion follows from the premises. Whether any of it is true is a question the framework wasn’t built to answer.

That move is older than ChatGPT. I was doing it in 1994. We all were. We sat in lecture halls learning to produce options pricing models that were elegant inside their assumptions. We didn’t ask whether the assumptions described the world. The professor wasn’t asking either. The exam measured fluency in the framework. The grade certified the fluency. Whether the model would survive contact with reality was a question that lived somewhere else, in a room I wouldn’t enter for another decade.

The student using AI is doing what I did, with better tools. Reciting plausibility inside a framework. The institution measuring the plausibility. The truth test, if it ever comes, happening somewhere else. AI didn’t introduce this move. The credential was already rewarding it. AI just industrialized it and made it cheap.

If the goal is the ol’ parchment paper, AI is the cheaper path. Cheap as in low value. If the goal is to prompt all the way, university was already the wrong path, and AI was never the question.

Quick note before going further: I’m writing here about North American universities. The universities I’ve worked with in Cameroon and Senegal operate in a different environment, where the feedback loop is tighter. I have not observed the same vulnerability there. Education is valued, real skills are recognized. Remember, I’m not an expert, and this is not a universal claim. I observe. I write.

It really comes down to this: if you go to university to learn to prompt, maybe you’re not that smart. You go to learn real stuff, or properly learn how to learn. Universities are not useless. Programs with external accreditation, like medicine, engineering, accounting, and nursing, have additional checkpoints. Same for college and trade programs that evaluate by triangulation: what you produce, what you can demonstrate, and what you can defend in conversation. The vulnerable programs are the ones where production is the only measure. That’s where AI is exposing what was already broken.

That said, even there, the evaluation still happens inside a framework. The difference is that reality gets more chances to interrupt. Galileo would have failed an astronomy class in his time.

A note on theft

I should mention that the title of a previous section, Why I Write, I stole outright from Orwell. Old school snatch. I read the essay a while ago, absorbed it, then got inspired to write my own self-serving section. He’s been dead long enough that it’s technically public domain in Canada, but I would have done it anyway. That’s how writers have always worked. Read what came before, absorb it, put it back out in your own voice for your own reasons. Orwell did the same with the people he read. The chain has been running for a long time.

Large language models like ChatGPT just industrialized the move and made it cheap. They train on stuff written by someone else, recombine it at scale, and put it back out with the attribution chain broken by design. The reader can’t tell what came from where. Neither can the writer using the tool. That’s not obfuscation exactly. Obfuscation implies someone is hiding something. The model isn’t trying to hide anything. Have you ever prompted one to summarize a book for you? The book lives rent-free on what I still imagine are hard disks somewhere.

The end is near

I was on X recently and saw a version of the post that keeps circulating: fed up university professors saying AI is destroying learning. An AI-powered, semi-illiterate workforce on the way. A glimpse into the educational apocalypse. The examples were the now-familiar ones. We already mentioned some.

The framing is wrong, and the wrongness is what this essay has been about. The apocalypse is not that students are cheating. The apocalypse, if there is one, is that the institution can no longer tell the difference between a student who knows something and a student who produces something that looks like knowing. Cheating is a symptom. An institution’s inability to measure is the disease.

I learned Black-Scholes models in 1994 that predicted outcomes in a world we don’t live in, and the institution gave me marks for it. The difference is I had ten years of work to find out which half of what I’d been taught actually held. The students writing essays in two hours instead of twelve won’t have that decade. They’ve outsourced the part where you find out you were wrong, before they ever learned what being wrong feels like. The customer is never wrong?

The cheating panic and the hallucination panic are the same panic. Both are about plausibility being mistaken for truth. Students do it on purpose. Models do it by design. Some institutions are stalled, because they no longer know what or how to measure.

The parchment paper is still printed if the tuition is paid up. But when everyone else is assuming a can opener, the useful person is the one who gets the damn thing open.

Share

The room mostly did not agree with him. I know, I asked around afterward. But you could not tell in the class. Nobody else stepped to the plate to correct him after I did.

Forty years later I called my high school buddy to make sure I had not made up a false memory. He remembered it too. We forgot the class. We remembered the teacher’s name.

We did not hallucinate.

The word hallucination was coined by Thomas Browne in 1646. An English physician. He needed a word for a vision that was off. Not a lie. Not a dream. Something the mind produces while convinced it is perceiving the world. From the Latin alucinari, to wander in the mind.

Browne watched humans hallucinate in the 17th century. Three centuries later, John Tait watched a machine do it.

Tait described a program called FRUMP, short for Fast Reading Understanding and Memory Program. The name already promised what the program could not do. FRUMP read news wires and summarized them. One day, it processed a sentence about San Francisco being shaken by the death of Mayor Moscone. FRUMP filed the story under earthquake.

Tait called it “the hallucination of matches.” The match was stronger than the meaning.

Subscribe now

That was 1982. Forty years before ChatGPT made the word a public concern. ChatGPT was released in November 2022. Within months, courts were sanctioning lawyers, professors were failing students, and editors were retracting articles. The same hallucinations entered medical records. ChatGPT had invented cases, quotes, and sources. Some humans had passed them on as facts. AI was giving us something we had not expected: smarter hallucinations. In 2023, Cambridge Dictionary named the verb, hallucinate, its word of the year. The new definition included AI hallucinations.

It sounds like a glitch. A misfire. Something the machine did while briefly disconnected from reality. The framing protects the system. The system is fine. ChatGPT just hallucinated. Sometimes humans confabulate too. The answer comes first. Then we make up the explanation.

What about the teacher in 1985? By Browne’s definition, he did not hallucinate. He believed it. He was not alone. The “fact” of fifty-two states was in circulation in the 1980s. Before the web, you went to the library for fact-checking or asked an authority figure, if you had access to one. The teacher was often the only authority in the room. The error survived. I had experienced a phenomenon that would not have a name for another twenty years.

Can you hallucinate if others experience the same thing?

The biological hallucination definition no longer holds. Hallucination now describes a system that is wrong, not corrected, and passed as knowledge. The burden of seeing clearly falls on the human, who read in good faith and assumed what they read was factual.

In 2009, a writer named Fiona Broome noticed that she had a clear memory of Nelson Mandela dying in prison in the 1980s. Other people online agreed. They remembered the funeral. They remembered the news coverage. They were certain.

Mandela was released from prison in 1990. He became president. He died in 2013, at 95, in his home.

Broome called it the Mandela Effect. The pattern kept showing up. Mr. Monopoly has never worn a monocle. Darth Vader never said “Luke, I am your father.” The line is “No, I am your father.” Some still think the United States has fifty-two states. People remember otherwise. Many of them with absolute confidence.

No machine was involved. No teacher pulled rank. Just a mistake that traveled. One person remembered wrong. Another agreed because they half-remembered the same thing. A third repeated it because the other two sounded sure. Somewhere along the way, the wrong version stopped being a mistake and started being memory.

The classroom has thirty kids. The social cost of disagreeing with the teacher is immediate. Most stay quiet. Some students could have walked out believing it.

The Mandela Effect operates on millions. The social cost is harder to see. You are not correcting a specific person. You are correcting a shared memory that has already hardened into fact. By the time the error is old enough to be named, too many people believe it for correction to travel cleanly.

We have more fact-checking tools now. Will it be easier?

We test Artificial General Intelligence (AGI) the way my teacher tested me: not to see if it can think, but to see if it provides the expected answer. Developers use benchmarks with technical names like MMLU, ARC-AGI, or HLE. These acronyms act as a shield. They discourage the question. The benchmarks do not track new ideas or complex problems. They check obedience.

Yann LeCun, Meta’s former AI chief, has admitted that his team used different models for different benchmarks. “Results were fudged a little bit,” he told the Financial Times. We are not testing cognition. We are checking compliance.

You rate an AGI based on the answer you expect to hear. Is that how you grade an essayist writing on AGI? A diplomat in a tense situation? A researcher with unexpected data?

AI nepotism: models feeding off models. When a hallucination enters the data, the next model learns it as fact. Error becomes norm. Some call it Habsburg AI, after the dynasty that went extinct through inbreeding. The model reproduces itself with itself until no outside signal can get in. Even before social media, rumours found a way to propagate. Urban legends too. Now, humans needn’t take part.

If ChatGPT hallucinates and nobody reads it, is it a hallucination? No. The hallucination needs a reader. And each ChatGPT session is its own environment. The same prompt can produce different answers between sessions, unlike traditional software.

Bugs used to have addresses. You found the cause. You wrote the fix. Large language models do not fail by bug. The error is not located. There is nothing to patch.

The industry has a response. Responsible AI. Human in the loop. Guardrails. Reinforcement learning from human feedback. A solution devised by the marketing department. The work looks like correction. It is closer to polish. A glass eye fills the socket. It does not see. You can keep it shining. It is still blind.

Meanwhile, the teacher still makes the questions and decides the answers. If he is wrong again, it is only a hallucination if he is caught... otherwise it becomes knowledge.

Subscribe now

Ça dépend.

He was not stalling. He was not flinching. I know that now. He is an electrical engineer who spent years planning telephone networks, then moved into venture capital investing. In both fields the honest answer is the same: it depends on conditions not yet evaluated.

Back then, it looked like paralysis. He couldn’t throw me an answer.

I pictured his boss in the early years, wanting to throw something across the room. Just give us the fucking answer. Tell us where to put that damn tower. Tell us which company gets the axe. Make up your mind. And my mom thought I had no imagination. I didn’t, but I watched a lot of TV.

But my brother was not equivocating. He was protecting the question. Why? He’s stubborn, I think.

There was a major power outage in northwestern New Brunswick. A man in my community posted on Facebook: Ok le manque d’électricité, y vont tu déduire ça de notre prochaine facture?

Was he asking for credit? Did he forget to use the sarcasm font? Let’s treat it as a joke. Or not... why do we feel the urge to resolve everything? I smiled and moved on.

Subscribe now

As a teenager, I was under the impression that I knew most of the answers, only to realize I was choosing the questions.

We ask a lot of questions. We rarely stop to ask if we are even asking the right thing. We assume we know enough to know what to ask. We do not.

Asking your doctor “will I die?” is not the same as asking about the probability of dying from a specific cause. Same person. Same room. Different conversation. The actual answer the doctor will think: yes, you will... eventually.

Reminds me of a meme. Everybody’s asking how’s the beer. Nobody’s asking how’s the brewer.

We do not fail to find answers. We often select the questions that have answers available and call that knowledge.

You ask a question and you want a clean answer. Think of a prism. Dark Side of the Moon style. You send the light. They hold the prism.

The prism does not explain what the light is. It reveals what it contains. What comes out is not an answer. It is the conditions the answer depends on.

But the light has to hit the right angle. Otherwise, nothing separates.

Silence, instead of an answer, is hiding the prism. The light is avoided.

Evasion, even langue de bois, is the prism with a filter. The conditions come out. The ones they chose for you, without your permission.

“I don’t know” is closer. At least the light is moving. But is it ignorance or apathy? One cannot find the proper angle. The other already left the room.

Ça dépend is not a bug. You asked for an answer. What came out is what the question actually contained. The conditions. The path to an answer, or another question.

Sometimes they hold it correctly and the light separates cleanly. You read, you move on.

My brother would not leave the question. That is what infuriated everyone. He held his ground until the conditions were legible.

He still built the networks. He still made the investments. Ça dépend did not stop him from pulling the trigger. It stopped him from pulling it before the path was legible.

The person who demands an answer before the situation is ready is not more efficient. They are rushing the process. Ça dépend does not delay the decision. It refuses the illusion that the decision is obvious. Meanwhile, the one who waited knows the ground.

The decisions are quick. The process behind them is diligent.

I recently took a week off and decided I would read three books. I started with the smallest one. Camus. Le Mythe de Sisyphe.

What a load to bear.

I carried that book everywhere I went that week. Read half a page. Put it down. Found something else to do. Picked it up again. It travelled with me like an unfinished obligation. I finally finished it on the flight home.

I was happy just looking at the finished book. Proud of the achievement. The boulder at the top. I closed the book and looked at the clouds through the oval window.

Then I remembered I had two more books to read. I smiled.

That is Camus. You do not arrive. You turn around. Sisyphus must be imagined happy. I understood that for the first time at thirty thousand feet, holding a finished book, already knowing the next one was waiting.

My brother figured that out at the dinner table. Camus needed about 150 pages to convince me.

I have been asking why since I was five years old. My cousins gave me a nickname that summer. Pourquoi. It was not a compliment.

Four decades later I am still asking. The asking did not produce answers. It produced better questions.

Don’t expect an answer. If you do, I’ll know you didn’t read Camus.

When we resolve x, and y, and z. We will reach supreme knowledge.

And then what?

One of the first stops was Picaroons in Fredericton. Sean Dunbar, the owner at the time, sat down and answered every question. Candid. To the point. No posturing. I was not the first person to show up unannounced and ask him how the business worked. I could tell.

At the end of the meeting, I asked if he was ok with me using any of what he had shared.

“Of course. This is all open source.”

It hit me in a way he probably did not intend. Some years back, I had spent a summer writing open source code for Google on the Open Syllabus project. I knew exactly what open source meant. The knowledge is available to anyone willing to show up. No gate. No licence. No proprietary filter. You take it, you build with it, you pass it on. For many, open source software meant something like free beer. For me, it meant something like free speech.

Linking open source software to craft beer made perfect sense to me. I have kept that approach with every brewer since.

A friend sent me a reel the other day. An English woman, probably in her seventies, was asked a simple question. If you were in your twenties again, what would you do?

Go to the pub.

Not travel. Not start a business. Not go back to school. The pub. She talked about her local spot in Durham. She went there many times a week. Parents, friends, strangers. Lawyers, doctors, teachers, lecturers. A total cross-section.

“We lived our lives in front of each other,” she said. “You went in, you were in a bad mood, or you had a row with your husband. You lived your life with that community.”

“Pubs are the last place where people can get together. Flaws and all.”

I received this reel at the exact moment I heard a bioethicist say on a podcast that the number one predictor of good health is socialization. Not exercise. Not diet. Not supplements. Being in a room with other people, talking.

The woman in Durham got it right without the study.

I have a strong suspicion that we will never have an app replicate what she described.

In 1992, a British anthropologist named Robin Dunbar proposed that the human brain has a limit on the number of stable relationships it can maintain. The number he suggested was roughly 150. The exact figure is debated. The constraint is not.

Below that ceiling, you know the people. You read the signals. Trust is built through repeated interaction. Social pressure is felt, not administered. You do not need a policy for cooperation. You need proximity.

The taproom replicates something older than any institution we have built. The regulars know each other. The staff knows the regulars. Reputation is tracked by showing up, not by profile. Nobody needs a rating system. The room reads you.

Sean Dunbar the brewer and Robin Dunbar the anthropologist share a name. One described the ceiling. The other built a business underneath it.

We used to have more of these rooms. Churches. Night clubs. Edmundston had plenty of both.

From my observation, the grocery store seems to be the new de facto meeting place. The aisle where you run into someone you have not seen in months. The checkout line where two neighbours catch up while the cashier waits. Nobody designed the grocery store as a gathering space. It became one because there is nowhere else left. Maybe there is a business plan behind this new social trend, but I digress.

Sean did not need the research. He opened a taproom in 2016. He said it plainly in an interview. “A quiet public place to talk about ideas is an amazing thing. Beer helps take you over that edge of, ‘Alright, this is going to sound crazy, and I normally would not tell you, but I have this idea...’ and if you talk to enough people and have that conversation, you run into enough people that go, ‘That’s not a crazy idea at all!’”

“That is how innovation happens, and community planning, and community thinking, and building businesses. That is why we do what we do. That is why we make beer.”

Not why we sell beer. Why we make it.

And yet, we tried to build this online. Social media was supposed to be the room at infinite scale. Everyone connected. No closing time. No geographical limit. No friction.

That was the problem. Social media removes the friction that makes the room work. I rarely see someone get infuriated in a taproom.

The awkward pause. The eye contact you cannot avoid. The fact that you cannot leave a conversation at a bar without the other person noticing. Online, you can exit without cost. You can perform without consequence. You can say things you would never say to someone sitting next to you.

The friction was not the obstacle. It was the mechanism.

A bioethicist will publish a finding. The finding will be reported. A programme will be designed. Funding will be allocated. The programme will measure engagement. A committee will evaluate outcomes. The measurement will become the objective. The objective will replace the thing.

And somewhere, a taproom will still be open. No programme. No mandate. No measurable outcome.

Just a seat, a pint, and the person next to you. Something that happened organically.

The headlines will tell you the pint is the problem.

I say, chat freely. Drink responsibly.

We have an ongoing discussion on supplements and growing older. He is a fiscalist. I run a brewery. Every spring, I distract him a little, out of empathy. I see tax season as the equivalent of hell incarnate for my accountant friends. They do not see it that way. Fair enough.

Whenever he thinks I am getting too confident on a subject, he has a line for me: “your CV reads finance and beer.” Recently he added another: “I stay in my lane.”

Obviously, I am more discerning. But he is not here to defend himself.

He is not wrong that expertise matters. It does. But what is expertise, exactly? Is it the credential, the department, the title, the stamp? Or is it the refusal to stop looking once the room has moved on? Every field has a gate. The people inside it, the gatekeepers, decide what counts as signal and what counts as noise.

That is why John Ott caught my attention.

Ott was a banker in Chicago who spent his free time doing time-lapse photography. He made his first film in 1927, still in high school. His hobby became a career after twenty years in banking.

His subjects were flowers. To film them properly, he built elaborate studios in the form of greenhouses, because a plant is not a prop. It has its own timing, its own rhythms, its own environmental demands. Keeping the lighting consistent from frame to frame while keeping the plant alive was an engineering problem in its own right.

Disney noticed. They hired him to shoot footage for their nature documentary Secrets of Life. They wanted a stem to pumpkin sequence.

In the studio, the vine would grow. It would flower. But the female flowers turned brown and dropped off. The male flowers developed vigorously.

The next year, the original fluorescent tubes had burned out. The hardware store was out of the same ones, so he grabbed daylight-white fluorescents instead. Then the pattern flipped. The male flowers dropped off and the female flowers developed vigorously. A pumpkin needs both to pollinate. No pollination, no pumpkin. He found pollen from another source. He finally grew his pumpkin.

But he kept wondering what had happened. He went back to the greenhouse and tested. He confirmed his thesis: switch the tubes, switch the outcome. Every time. The plant needed what artificial light could not give it: the full spectrum it would get from the sun.

The scientists who reviewed his work moved on. Most people who run experiments outside their field produce noise, and the gate cannot tell the difference in advance. It filters correctly most of the time, I assume. The cost is that it filters at the same rate regardless of what is coming through.

Ott kept going. He did bona fide botanical and biological experiments. He coined a term for what he thought was happening to people living under artificial light: mal-illumination. The light equivalent of malnutrition. Loyola University gave him an honorary doctorate in 1958. The credential arrived after the achievement, not before or during it.

What Ott observed was never controversial when it came to plants. A serious grower knows that a tropical plant moved north does not just require water and warmth. It needs the right spectrum, the right length of exposure. Ask your florist.

No one finds that mystical. No one calls it fringe.

People will worry about their orchid’s light and never once think about their own lighting environment. Plants have mitochondria. So do you. Photosynthesis runs on visible light. Mitochondria contain enzymes that absorb red and infrared light. Different parts of the spectrum. The unfiltered sun provides all of it.

The ancients did not need a microscope to notice something was there. Three thousand years before anyone had a word for spectrum, the Greeks were filtering sunlight through coloured cloth in healing temples. Ailing people were brought there. They called the city Heliopolis. City of the sun. They did not know why people recovered from their ailment. They knew that they did.

Nobody had to refute sun therapy. Something more convenient arrived. The question was not asked.

But not everyone forgot.

In 1903, Niels Finsen won the Nobel Prize in Medicine for treating disease with light. He was sick himself. Weak, anaemic, living in a north-facing house in Copenhagen. He noticed that time in the sun made him feel stronger. That was the observation. Everything else followed from it.

That same year, Auguste Rollier opened a heliotherapy clinic in Leysin, in the Swiss Alps. He was treating tuberculosis, primarily of the bone. Eventually he ran thirty-six clinics. Over a thousand beds. His recovery rates were remarkable enough that the clinics kept expanding.

Then antibiotics arrived and cured tuberculosis, the same disease Rollier had been treating with mountain light for twenty years. The sanitariums closed. Nobody asked whether the light was doing something the drug was not. I guess the pill was easier to swallow.

Allow me to go out of my lane for a moment. There is an enzyme in your mitochondria called cytochrome c oxidase. It sits at the end of the energy chain that keeps your cells running. Scientists first noticed it in 1884 because they could see it absorbing light at specific wavelengths. They named it in 1925. The observation went into the literature and stayed there. What it might mean for cells living under artificial light was not a question anyone thought to ask. Nature does not build light-absorbing molecules for nothing, I assume. Nature is thrifty. A molecule that absorbs light at specific wavelengths is doing something with it.

In March 2026, Nature published an article on red-light therapy. The article found that cells exposed to red and near-infrared produce energy more efficiently. When cells are healthy, the difference is small. When cells are stressed, it matters.

The article does not mention it, but cytochrome c oxidase is known to absorb light in that range. What it does with it is still debated. The absorption is not. The study is precise about what it measured. One variable, in isolation. The body does not receive light that way in nature.

Take away something nobody was measuring, and you will not notice until you need it.

The sun provides these wavelengths for free. But we always substitute for efficiency purposes. Do we really?

“LEDs are efficient.” “They cut consumption.” “They last longer.” We are all familiar with the marketing pitch. Meanwhile, nobody counted the components, the circuits, the rare earth materials, the cost. But then again... efficient at what?

The old incandescent bulb produced a broad range of light, including red and infrared wavelengths. LEDs replaced them with a narrower spectrum, optimized for one thing: lumens per watt. Visible brightness per unit of energy. That is what efficient means. The wavelengths your cells respond to were not part of the equation.

Where I live, in northern New Brunswick, the incandescent bulb’s so-called waste heat was heating the living room I was sitting in. The bulb was doing several jobs. The energy savings were real. The other metrics were never in the room.

My cousin would say I have too much time on my hands.

He is not wrong about the gate. The world is full of people with theories and no signal. The scientists who ignored Ott were doing exactly what the gate is built to do.

Finsen still won the Nobel.

The gate was right to be skeptical. The gate was wrong about Ott. The gate has no way of knowing which situation it is in.

Neither do I.

Sometimes the banker notices one thing. Sometimes the brewery guy reads outside his lane. Sometimes the knowledge was sitting there in plain light.

The hype is real. The progress is real. But most of what is being sold is still a sales pitch. Especially the ‘G’ in AGI.

Artificial General Intelligence, the version of AI nobody has built yet. A machine that reasons, learns, and solves problems across any domain. On its own. Not just chess, not just code. A brewery. A hospital. Everything.

It is not an ambition. It is a tell.

The definition keeps moving. Every time the machine falls short of General, the people selling it move the goalposts.

Jensen Huang runs Nvidia. His chips power most of the world’s AI. He declared AGI achieved, then admitted in the same conversation that the odds of AI building something like Nvidia are zero. That is not a milestone. That is a marketing claim eating its own tail.

What we call reason can be automated. The machine that beats every grandmaster is not a trick. It is reason at scale. It does not read the game the way a grandmaster does. Millions of positions evaluated per second. In a closed system with fixed rules, computation is sufficient. Judgment is not needed. The machine only needs what is on the board.

Chess, code, medical scans. Within those boundaries, it is often superhuman.

The promise of the ‘G’ is that you open the boxes. The machine reasons across any domain, on its own. Generally.

But reason is still reason. Scale it as far as you like, a machine cannot walk into a room mid-conversation and know they were talking about “it”. For humans, just a faint blush says everything. The machine was not in the room.

Friday evenings, the Petit-Sault taproom gets festive. Conversations overlap. Glasses clink. Music runs underneath. Your brain handles this without effort. You hear the person across the table. Not because the other sounds disappear. Because your brain, in real time, reads what matters and drops the rest. It filters. Not everything gets through.

Last summer my parents drove in from out of town. We sat together in the taproom on a busy Friday.

My father wears a hearing aid. That night, he struggled to follow the conversation. I watched him smile and nod at things he couldn’t quite hear. He had probably turned it down.

Everything came at him at the same volume, like a phone call from someone standing in a crowd. The group at the far end of the bar. The clink of glass behind them. The music. My mother beside him. All of it arriving with the same priority. The brain filters. The hearing aid does not. That was the problem.

Not all misreadings cost the same.

In 2018 and 2019, two Boeing 737 MAX aircraft crashed within five months. 346 people dead. The pilots never had a chance.

Boeing had redesigned the 737 with larger engines set further forward on the wing. It changed how the plane handled. Rather than redesign the airframe, Boeing added a system called MCAS. Boeing gave MCAS one sensor input per flight. The aircraft had a second. Engineers wrote the system to read only one. When that sensor misread the data, MCAS pushed the nose down repeatedly. The pilots fought it. The software’s instructions overrode their judgment.

Observe a first-time flyer on a commercial flight. Every sound registers. Every shudder. Every turn, every drop in altitude, every change in engine pitch. They grip the armrest. They look at the flight attendant’s face for reassurance. Not because the plane is in trouble. Because nothing in memory tells them what is normal. Everything arrives as potential signal because nothing has been filed in memory yet.

The pilots in both crashes were reading the aircraft. Their body, their instruments, their experience. One sensor contradicted all of it. The software was not built to question it.

The first-time flyer has no instrument yet.

I have been tasting our beer for thirteen years. I know when something is off. Not always immediately. Not always in words. But I know.

The midwife who reads a labour room in seconds, the mother’s breathing, posture, skin colour, the rhythm of the contractions, built that reading across hundreds of deliveries. Each one different. Each one adding something that belongs to her and no one else. The emergency room nurse reads triage in a glance. The pilot knows in their seat what the sensor cannot read or misread.

This is judgment. Specific, not general. Accumulated over a lifetime. You paid the cost when wrong. You got the reward when right.

It fails sometimes. The difference is that the person who exercised it faces the consequences, however minimal. No training data does that.

No two people read and sort the same way. That is why you bring different people into the room for each important decision.

The assumption is that the problem is sensors. Not enough data. Not enough inputs. Not enough compute. Scale the machine and you close the gap.

But our advantage was never in receiving more. It was in knowing what to ignore.

Reason computes. Judgment reads and sorts. Insight arrives where experience ends.

Pattern recognition finds what has appeared before. Artificial intelligence does this remarkably well, at superhuman speed, across scales no human can match. Insight is the moment something arrives that experience could not have produced.

Sometimes I taste something I cannot name. Not a deviation. Something new. The palate registers it before any previous experience covers it. It arrives from the body before the mind has caught up.

The midwife who has attended a thousand births still encounters the delivery that feels different before she can say why. The moment of knowing precedes the knowing. The pilot felt it. The instruments confirmed it later.

Can AGI ever have insight? Nobody has shown how. You do not get there by scaling reason. Improvement from data is not the same as improvement from consequence. The machine does not pay when it is wrong. Someone else does.

As I write this, my dog is monitoring the front of the house from the window. Intensely. My wife has been away almost a week. She’s driving back from the airport, about 100km away. The dog is waiting.

Now tell me exactly where that sensor is.

In 2001 or 2002, a few co-workers and I went for drinks after work at the 737, which happened many times in the summer. Altitude 737 sat in the penthouse of Place Ville Marie, forty-seven storeys above downtown Montreal, named for its elevation in feet above sea level. We worked on the fourth floor of the same building, at RBC Capital Markets. Through a friend’s connections we had direct elevator access to the top. No lineup.

“5 à 7” was what we called it, though we rarely left the office before eight.

One of those evenings, someone mentioned that the boss had Cuban cigars in his office. Good ones. From a contact. Word was he did not mind people helping themselves. Good enough.

The boss held the top investment banking title in the Montreal office, though he was not really our boss. He did not manage people. He managed relationships. His value to the institution was his network, and everything about him oozed that. A lawyer by training, a Rhodes Scholar, he had run the corporate-commercial practice at one of the most prominent law firms in the country before moving to a chief legal officer role at a national railway. Extremely well connected. Thousand-dollar suits. Brand-name suspenders. The kind of person who never had to explain where things came from. And in that kind of office, at that level, you did not correct the man. The reflex was to humour him, although interaction was minimal. That was understood by everyone, even people who had never been told.

So we headed to his office to check them out. A few drinks will give you that kind of courage. I saw the cigars from afar. Right away I knew they were fake.

Not because I was smarter or particularly gifted. Because I had spent a lot of time at Blatter & Blatter in the 1990s, say after lunch on Friday with colleagues. The funny thing is that I had my own version of the boss’s connections. I had a tobacco shop where I was a regular. Blatter & Blatter were cigar and pipe royalty in Montreal, recognized over larger territory. He had contacts who got him Cuban cigars. I had a shop that taught me what authentic Cuban cigars actually looked like.

Blatter & Blatter has been in Montreal since 1907, when Ernest Blatter and his family arrived from South Africa and opened a pipe factory and a retail shop. Five generations of Blatters. Pipes made by hand on the premises. The family had been offered opportunities to expand over the years and turned them all down. They preferred to keep the operation small enough that they could control the quality of every piece.

The shop sold serious Cuban cigars. Mainly not Cohibas and Monte Cristos. Those are what tourists buy and counterfeiters fake. Cohiba was Fidel Castro’s trademark, originally made for him and his communist comrades in 1966, first offered to the public in 1982, and counterfeited almost immediately.

The real inventory at Blatter’s was Bolívar Belicosos Finos, Hoyo de Monterrey Epicure No. 2, Rafael González, Ramón Allones. Brands nobody bothers counterfeiting because nobody buying them is trying to impress anyone, except maybe other cigar aficionados. You smoke a Hoyo Epicure No. 2 because you like how it smokes.

Pierre Blatter got asked the same question several times a week. Someone would walk in with a box: “I went to Cuba,” or “A friend brought me Cohibas.” He could tell from across the room. So could I, eventually. A real box of Cohibas is not cheap, so the manufacturer puts the effort in. Every cigar the same colour. Same length. The band placed precisely, printed cleanly. Fakes get the broad strokes right but not the details. The uniformity is off. The printing on the band is usually wrong. Once you have handled enough real product, the fakes announce themselves.

The boss’s cigars were fake. Classic fake Cohibas, probably bought in good faith from someone who bought them in good faith from someone on a beach in Varadero. That is how it works. The supply chain is not dishonest at every link. It only needs to be dishonest once, early enough that everyone downstream believes the story.

I told the guys at work. They did not believe me. So the next day, when he was out, I took the box to Blatter & Blatter. Pierre made the call from maybe ten feet away. Fake.

As a reference, a box of twenty-five Cohiba Esplendidos now sells in the thousands. A good Cuban cigar is amazing. Those were definitely out of my budget. I did see some legit contrabands though.

Now, here is where it gets interesting. I was told the boss was planning to give a box of those cigars to a client. A Quebec businessman mostly known by his initials, who had inherited one of the largest media empires in the province after his father’s sudden death a few years earlier. Anyone from Quebec would know the name.

The career move was silence. Let the cigars go to the client. If someone in the client’s circle spotted the fakes, that was the boss’s problem, not mine. I would have been invisible. No risk, no cost, no exposure.

I could not do it. I made a fuss about it. Told my co-workers he needed to know, that he could not give those cigars to that client. But I did not want to tell him directly. I probably said something like: maybe we should let him know somehow. But there was no somehow. There was no memo template for reporting fake cigars. No way for “your cigars are counterfeit” to move from a junior associate to the vice-chairman through normal operation.

By making a fuss, I blew the whistle without meaning to, which is not a bad joke given that the boss’s previous employer was a railway. I do not remember exactly how it reached him that an associate had identified his prized Cubans as counterfeit. But it reached him.

I wanted it to reach him. Maybe not with my name along for the ride. I had a delusion of being Deep Throat over a box of Cohibas. I am quite sure my name travelled with the gossip. My cover was blown.

Did I wonder whether it might cost me a promotion, a bonus, my job? It crossed my mind. But I rationalized it quickly enough: I had prevented a man from looking like an idiot in front of a major client. Who would be against that? Maybe I should have dropped my coffee on the cigars. Being a useful idiot would have been safer. I am not a great actor.

What only occurred to me later is that a man in his position could not afford to ever look like an idiot. Not to a client. Not to a colleague. Not to a junior associate. The image was the job. If something messes up his image, you fix it quietly or you do not fix it at all. The person who reveals the flaw is not doing a favour. He is creating a problem. Catch-22.

I never found out what happened. I never saw him with cigars again. RBC never gave me a heroic medal. And the fake cigars were now my fault.

In May 2024, our payment terminals started refusing transactions at the brewery.

I called Square. The person on the phone was polite. They walked me through the steps. The people who have been there know the drill: reset the terminal, check the connection, try again.

I tried again, and again. Same problem.

Word on the street, or in the Taproom, was that other businesses in Edmundston were having the same problem. The restaurant down the street. The coffee shop around the corner. A system failure, I assumed. I checked the Square status page. Dashboard showed all green. I guess Edmundston was just a margin of error.

We were all calling Square. We were all getting the same script. Reset. Reconnect. Try again. Nobody could tell us why. My day had just switched to debug mode.

I was helpless. Cash was king at one point. Then plastic made every business dependent on a third party to complete the most basic act of commerce. Most customers were not going to the ATM. I could not properly serve customers until this was fixed. Meanwhile, every Square user in town was calling customer service, and they could not fix the problem either. Felt like a quagmire was developing.

Then a pattern emerged, I figured. The businesses having the problem were all along the Canada–US border. Edmundston sits right on it. That is the kind of pattern you would want to know if you were tasked with troubleshooting at Square.

I called Square again, armed with what I had found. The person on the phone had no procedure for receiving a diagnosis from a customer. They had a checklist. Screenshot of the error. Run a test payment. Try another network. The checklist had a terminal flaw: no slot for my problem. It was designed for a broken terminal, not a geographic mislocation affecting many businesses along an international border. The call ended the way the others had. The problem was my fault. So they thought.

After a couple of weeks I went to Twitter. I described the problem publicly. That changed the dynamic completely. Within minutes, a direct message from Square. The escalation was triggered. Suddenly I was in contact with an engineer named Jack who had access to the system and the authority to look at what was actually happening.

He was not following a script. He could hear.

I told Jack my theory. He confirmed it. The terminals along the border were reading their location as the United States. They would not process Canadian transactions because, as far as the software was concerned, they were not in Canada.

Jack overrode my settings manually. Bienvenue au Canada. The problem recurred a week later while I was in Europe. A system update reverted the manual bypass. Jack fixed it again within minutes, then a permanent software update rolled on the next Wednesday. Done. The whole thing took about three weeks from the first escalation to the final fix.

With the help of hindsight, I later came up with a theory about what could have triggered the location issue. My hamster brain tends to work overtime with stuff like that.

The new Madawaska–Edmundston International Bridge opened to traffic on June 6, 2024. The new bridge replaced a crossing that had stood since 1921 on a new axis, just meters from the original.

I do not know how Square locates its terminals. But my problem happened just the month before, in May. This one time, for me, correlation was causation. I was satisfied. I could be wrong. Most customers with theories are, I assume.

Multiple businesses were calling in through the proper channel. The channel could process a complaint. It could not carry an insight. But Jack could.

There is always a Jack inside those large organizations. The path to him exists. It is not available from the outside. Most systems are designed to make sure you never reach him. At one point, I was a Jack when my parents had printer issues, and they had my phone number. I quickly figured out I was not customer service material. I digress.

Twitter forced Square’s hand. It did not create the path to Jack. It made the failure visible enough that Square activated it.

And Square was a company with a phone number.

There was a time when having a problem with something you bought meant going back to where you bought it. Someone behind the counter heard you and acted, or told you plainly they could not. The distance between you and the person who could help was a few feet.

You want to tell our brewer personally that you enjoy our pilsner? Come to our taproom. We have a counter for that. Not a difficult task.

For large companies, the hotline replaced the counter. The phone tree replaced the hotline. The chatbot replaced the phone tree. Each step cost less to run. Each step removed a person who could read what you were actually saying.

The endpoint is not worse service. It is the absence of it, all designed to look like its presence.

Every interaction the institution eliminated saved money. A dollar removed from expense is a dollar moved to profit. What was lost is now invisible.

The person with the problem can no longer reach the person who can fix it. The institution did not stop listening on purpose. It stopped because listening gets harder, fast, when a business grows. No big time executive gets rewarded or promoted for arguing that customer service should grow at the same scale as the business.

The product stops improving because the feedback that would improve it no longer arrives, unless it pops up in the numbers (you know, big data). The customer stops trusting because trust is built on being heard. The cost compounds quietly. Not in any single interaction. In the aggregate. In the product that drifts further from the people using it while the dashboard says everything is fine.

Choosing the dead ear is not free. It is just not a cost anyone is tracking.

But that was a payment terminal problem. Imagine when you don’t know Jack.

Now consider the business that runs on a platform it does not own. Your customers find you on it. Your revenue depends on it. The way you reach your customers runs through someone else’s system, on their terms. Not negotiable.

The platform suspends your account. You look for someone to call. There is no one to call.

Billions of users. No phone number. Try finding one for Facebook.

Unless you are visible enough to force one.

Even the companies trying to get it right hit the same wall.

You pay for a tool. You use it daily. Something goes wrong. You write a detailed report through the usual channel. Days pass. Sometimes more. A reply arrives. It is templated. It addresses a version of your problem that is not quite your problem. You reply with the specifics again. Silence. I know. It was my recent experience with Anthropic (Claude.ai).

The institution did not ignore you. It processed you. Processing and hearing are not the same thing. Unless you are part of a trend on a dashboard, you are a rounding error.

The bigger the company, the larger the market capitalization, the more users it serves, the less likely it is that you can reach a person who can act on what you are telling them.

Market capitalization and reachability run in opposite directions.

The store counter had a feedback loop. The person behind it could hear you, read what you meant, and act on it or walk it to someone who could. Every step that replaced the counter was designed to handle more interactions. Volume and quality are not the same thing. A system optimized for volume processes ten thousand complaints identically. It hears none of them.

The company that made itself unreachable did not just save money. It lost the ability to hear the thing that would have told it what to fix next.

That is not the efficiency that technology promised. The consumer lost on that equation.

Jack could hear. The system made it difficult for me to reach him. The real cost is not what I went through to get there. It is that they will never have enough of him. When you scale by removing the person who listens, you lose the conditions that produce people who can.

The Dashboard Is Not the Body

Bryan Johnson Optimised the Map and Forgot the Territory

My son had terrible eczema as a baby. We did what concerned parents do. Asked around. Researched. Heard one brand of skin cream was the best. We finally brought it up with the local pediatrician.

He was old school. Un vieux routier. We mentioned the cream and he dismissed it before we finished the sentence. Use the pharmacy base, he said. Same active ingredient as every other one on the shelf. Without the scent, without the additives, without the price tag. I appreciated his destabilizing frankness. No half measures.

Thirty years as a no-bullshit clinician. He did not need the packaging to know what was inside it. He never said it would cure the eczema. He said it would provide relief. That distinction mattered.

I recognized what the doctor expressed with his action and his sharp words. Years in finance and business had already trained me to identify a pitch before I heard what was for sale.

That seed stayed with me. Over the years I grew skeptical of the supplement industry, the wellness industry, the longevity industry. I could smell the fee structure underneath the clinical language. What I keep seeing is a business model dressed as a solution.

But skepticism is not an argument. It tells you something is off. It does not tell you why. You may have the instinct, the experience. If you cannot put the reason into words, you cannot communicate the point. What you cannot formalize into words, you cannot defend.

Years later, I could finally put words to what I had always felt. Years of reading biology, systems, philosophy. How organisms regulate themselves. What happens when they stop. Not a curriculum. A question I kept following because the answers kept connecting to things I had already seen in institutions, in businesses, in markets. My analytical mind tends to work overtime. Strange subject for someone who works in a brewery, I know.

A brewery operator in New Brunswick reading quantum biology. A tech bro biohacker in Los Angeles optimizing his blood. The algorithm that put us in the same feed may be the only thing working as designed.

Bryan Johnson documents everything. Every biomarker, every protocol, every measurable detail of his attempt to reverse aging, all of it on X. What began as a personal experiment has now been packaged into a protocol called Blueprint and offered to anyone willing and able to pay for it. He outsourced sleep, appetite, and exercise to an algorithm. Forget about what your body tells you.

Whether he outsourced fun is not explicitly mentioned.

The problem is not the ambition, I admire the dedication. It is what the approach replaces.

Most animals act directly on signal. Humans receive the same signal and think about it first. Johnson has replaced that thinking with an algorithm. The signal still arrives. The algorithm answers first. Every time the algorithm answers instead of the person, the capacity to read those signals weakens a little further. Not because the capacity was removed. Because it was no longer consulted, or plainly ignored.

The protocol does not simply guide judgment. It assumes judgment was never necessary.

The old pediatrician read the patient. The algorithm reads the data. One had decades of feedback between what he prescribed and what happened next. The other has metrics. The difference is not technology. It is whether the feedback loop is intact.

The organism is a sensing system before it is a set of measurements. It reads the environment. It integrates. It corrects. When a cell is no longer serving the organism, the signal arrives and the cell dies. Not because something attacked it. Because the system is still listening and the appropriate instructions are executed. That capacity requires the cell to be functioning as a reader, not by using a third party dashboard.

I do not interpret Blueprint as a wellness protocol. It optimizes what it can measure and calls that health. Are you healthy if you need to be convinced you are?

A man who has driven his resting heart rate to 37 bpm and his body fat to 5 percent has achieved a goal. Whether he is in good health is a different question.

Blood work is a picture in time, measured with certain tools. Optimising toward specific biomarkers is not the same as understanding how the organism is doing.

Supplements follow the same logic. Take a molecule out of the system that gave it meaning and deliver it in a pill, a liquid, a powder. The body evolved to extract nutrients embedded in fibre, water, timing, and the physical act of eating. More than bioavailability. Most omega-3 extracts are oxidized before they reach you. An optimal way to get your DHA is to eat a fresh oyster or a fatty fish like mackerel or wild salmon.

Strip that context and you have not delivered the nutrient more efficiently. You have delivered it differently. What the body does with it is no longer the same. The body did not evolve to process bulk nutrients all at once.

Imagine a lab that reproduces the full spectrum of sunlight in a laser. Every measurable property matches. Wavelength, intensity, frequency. They aim it at a plant. The plant dies. Not because the measurements were wrong. Because sunlight is not its measurements. It is the relationship between the light, the atmosphere, the timing, the angle, the season. Match the measurements and you still miss what made it work.

The supplement is the laser. The food it came from is the sun.

Concentration is not simply purification. Whatever was deemed beneficial gets concentrated. So does whatever was harmful. The protocol cannot tell the difference. Neither can the dashboard.

That problem goes beyond the pill bottle. Johnson is not just isolating molecules. He is isolating himself. We evolved within environmental signals. Light cycles, temperature variation, movement, food structure, microbial exposure. Strip that context and the measurements remain. But their meaning becomes less certain. The dashboard may show green. It may also be reading the wrong environment.

When the magnesium supplement improves sleep, the real question is not whether magnesium helps sleep. The question is whether low magnesium was the signal or a symptom of something else the supplement is now masking.

The dashboard shows green. The system may still be misaligned.

This is the iatrogenic risk Blueprint cannot see. The protocol improves the marker. The dashboard shows green. The protocol takes credit. The customer who paid for the protocol sees green and feels the confirmation. The dashboard is the gold star on the homework.

A statin lowers the cholesterol marker. The physician continues the prescription. Whether the patient is better is a question nobody is asking.

This logic is not unique to Blueprint. Modern medicine has been making the same trade for a generation: clinical judgment out, protocol in.

Meanwhile the capacity to read your own body gets a little quieter every day. Nobody would ever care more about your health than you would. Who is really listening?

This is the same problem I keep finding in every system I look at. The institution that replaces judgment with procedure. The business that replaces signal reading with metrics. The organism handed a protocol where self-correction used to be. Same mechanism. Different scale.

The project promises optimisation of the organism. What it actually produces is optimisation of the dashboard that describes it.

We started the brewery in 2014 with beers named after local historical figures.

Tante Blanche, who ordered the rich to share with the poor during a famine. Bob LeBoeuf, the legendary prospector. Louis XVII. Soeur Catherine, a hard-nosed IPA named after a teacher who took no shortcuts.

The stories were real. The connection to place was genuine. In the taproom, the whole system worked. People loved it.

Then the beer left the taproom.

Shipped outside the francophone Northwest of New Brunswick, a stubby bottle sits on a shelf next to hundreds other craft beers, most of them local, all of them telling a story. There is no story in the air, no local pride, no one to explain who Tante Blanche was or why it matters. A customer with no connection to the place picks it up, reads the label, and either buys it or puts it back.

The beer is completely alone.

Then the homophone arrived.

Some anglophone drinkers were reading soeur as sour. That’s fair. Damn.

In 2015, sour beers were rare in New Brunswick. The misread did not just confuse people. It sent them toward a style expectation the beer could not meet.

We took nearly ten years to act on that signal. Commitment to the original vision, mostly. Which is its own lesson.

We eventually dropped the ‘Soeur’ part. The beer stayed. It remains one of our flagships today, doing what it was always supposed to do. It is now Catherine Old School IPA.

A message arrived about another beer name. A story about the person it honoured. Unconfirmed, but credible.

The name changed the same day.

No spreadsheet justified it. No positioning exercise optimized it. There was no meeting about brand equity or sunk costs.

There was just a clear answer, arrived at in about thirty seconds.

Three different tests. Three different kinds of problems. No framework told us what to do in any of them.

That is where I want to start.

Simon Sinek is right.

Knowing your why matters.

Organizations that understand why they exist tend to make clearer decisions. They attract people who resonate with them. They maintain coherence longer than organizations that operate purely on incentives or momentum.

There is nothing controversial about that.

But his framework stops at the inspirational moment. It explains why the question matters. It does not explain what comes next.

And what comes next is the hard part.

A why is not a destination.

It is a constraint.

More precisely, it is an energy-saving device.

Without it, every decision forces you to relitigate first principles. Every disagreement becomes philosophical. Every operational problem risks turning into an existential debate (and some did!).

A real why narrows the decision space.

You do not start from zero each time something breaks. The why compresses the range of acceptable options before the discussion even begins.

That is the real value of the idea, and it is rarely explained this way.

At the beginning, a why is almost always vague.

It has to be.

Most founding statements sound like this: community, quality, connection, authenticity. Naive. Sincere. Broad.

That is not a flaw. It is the raw material.

The real meaning of the why is not decided when you write it down. It is decided later. Quietly. Through stress tests the business did not plan for.

No framework can tell you what to do in those moments.

That is where judgment enters.

The local heroine, the homophone, and the message were not anomalies.

Is this a distribution problem? Or a values problem? The homophone took a decade to resolve, not because the signal was unclear, but because the why was doing its job. It protected what mattered while the noise sorted itself out.

Eventually the expression had to change. The foundation did not.

The message was the opposite. No ambiguity, no delay. A why that lives in you makes certain decisions fast. A why that lives on a wall or in a deck creates friction at exactly the moment when speed and clarity matter most.

That is the gap I kept running into in Sinek’s framework.

He is right that you should know why you exist. He is right that it shapes everything downstream.

What he does not tell you is that the why is a hypothesis, not a conclusion. The business is the stress test.

And the stress test requires something no framework can supply: the judgment to read ambiguous signals accurately, to distinguish a naming problem from a values problem.

The why does not make decisions.

It makes certain decisions fast.

And over time, if you pay attention, it stops being the vague founding cliché you started with. It becomes something earned through the decisions you actually made.

That is not inspiration.

That is craft.

The ultimate taproom is the one where people don’t feel the need to look at screens for distraction.

That may sound like a small thing. But it’s surprisingly rare.

Since the beginning, I’ve resisted adding televisions in the Petit-Sault taproom. A couple of times, I was pressured into putting on the game. Apparently, who won had direct implications for the future of geopolitics.

So no TVs. Not out of some aversion to television. As an 80s kid, I was raised and lectured and entertained by TV. The reason is simpler. My brain is wired to notice screens. Put one in a room and attention drifts toward it automatically, whether the content matters or not. I’m bored, I reach for the phone. Not intention. Reflex.

I realized this years ago in a restaurant with my family.

A golf tournament, or maybe cricket, was playing on the television across the room. I don’t follow golf. I don’t understand cricket. I couldn’t have cared less who was playing. And yet I caught myself watching it. Not out of interest. Out of reflex.

The screen was there. My brain responded. Like a cat following a laser dot.

Once you notice this, it’s hard to ignore. Screens don’t just display information. They pull attention. Even when the content means nothing to you.

That’s why I never installed televisions in the taproom.

The goal wasn’t to create a rule. It was simply to create a room that didn’t compete with itself for attention.

Phones are different. Someone pulls one out to text a friend to join them. Someone else snaps a photo of a beer they liked.

But the best taprooms share a particular quality. Once people settle in, the room becomes more interesting than the device.

Conversation takes over.

Someone asks what you’re drinking. Someone else recounts the idiotic thing their co-worker did, again. A person recommends a playlist to the bartender. A group at the next table starts debating which game to play, Trivial Pursuit or UNO. Two people who arrived separately end up sitting together.

None of this is dramatic. That’s the point.

Beer is a slow technology.

It asks people to sit down, stay a while, and share the same space. It doesn’t speed anything up.

And when that works, something small but noticeable occurs.

The phone stays in the pocket.

For a while, the room is enough.

The ultimate taproom is not the one with the most taps or the newest releases. It’s the one where the room itself holds your attention.

It’s the one where people don’t feel the need to look at screens for distraction.

That’s the room we tried to build at Petit-Sault.

Around ten years ago, a group of craft alcohol producers in New Brunswick got together to find a common voice. We were trying to define what craft meant, mainly to have something to say to the regulator and the provincial liquor monopoly. Big beer had the shelf, the taps, their ears. We were fighting for inches. The people in that room had simply taken a turn making something they loved, at considerable personal cost, bound by a simple thing: making good beer and good alcohol in their province. Nobody was getting rich fast.

In the middle of one meeting, someone asked a simple question: if a brewer used malt concentrate instead of mashing grain, would their beer still be considered craft? It was exactly the kind of question the group had assembled to answer. The room reacted as if a bad word had been spoken. Not an argument. A reflex. The question had landed before anyone had evaluated it.

The reaction made some sense. Elsewhere in the province, some distiller licence holders were buying industrially produced neutral grain spirit, adding botanicals, and calling the result craft alcohol products. That fight was real. The line was genuinely under pressure. But in that atmosphere, the room could no longer tell the difference between a producer gaming a definition and a brewer asking an honest question.

What nobody mentioned: some extract beers have won competitions. Blind taste tests have fooled experienced judges. The quality gap that once justified the stigma had closed. The room was not arguing about quality. It was arguing about identity. About what the group was trying to be, and who got to belong to it. The room never found out, because the room never asked.

I saw the same mechanism more recently, in a university in Cameroon, when I mentioned ChatGPT to a room of trainers. Immediate. I could read it in their faces. Not curiosity. The word had landed before the question had, another bad word in another room. This could have been any classroom, anywhere.

This alarm also made sense. A student has to write a book report. The student hands it to an AI tool, runs it through an obfuscation tool, submits it. The assignment was designed to build something. That process has been bypassed entirely.

But the programme we were developing was for women entrepreneurs. Say one of them needs to understand a concept she cannot grasp. She finds a book, reads an article. She uses AI to explain the passages that stop her. She is not skipping the reading. She is doing more of it. The tool is extending her reach, not replacing her effort.

Same tool. Opposite relationship to the work.

The institution could not tell the difference. Its response was the same either way.

In the eighties, a teacher could usually tell when a student had copied from an encyclopedia. A ten-year-old does not write like that. I know because I tried. The signal was readable. The receptor still worked. Today the output looks the same whether the work was done or bypassed. The tool closed the gap the teacher used to read through.

What has changed is not the temptation. What has changed is the capacity to read what is actually in front of you.

The craft alcohol story ended quietly. A question was never asked and nobody noticed. The Cameroon story has not ended. Those trainers went back to their classrooms. I may have shifted something for some of them. For the others, I cannot say.

Every institution eventually does the same thing. It takes the judgment of its best people, writes it down, and runs the rule forward. The rule is cheaper than the original. It is also always a little wrong.

What AI replaces is not jobs. What AI replaces is the rule that stood in for judgment. The procedure that was already standing in for something else. The judgment was already gone. AI just made the procedure cheaper to run.

Education systems penalize students for using AI because they cannot tell the difference between use that bypasses learning and use that deepens it. Corporations mandate AI adoption without distinguishing procedural tasks from those that require judgment. Regulators write policy that addresses what they imagine AI does, not what it actually does. They call it ‘responsible AI.’ Human in the loop is the reassuring phrase. The judgment has left the room. The language remains.

This is not a virus model. There is no external pathogen. The institution’s own immune system has turned against itself. It cannot distinguish between a threat and native capacity. The receptors still fire. They fire at everything.

Institutions do not lose judgment because they misunderstand the situation. They lose it because the cost of exercising it becomes higher than the cost of running the rule.

A rule does not have a shelf life stamped on it. But ignore it long enough and it stops protecting what it was written to protect.

The problem is not AI. The problem is the kind of work that institutions have already reduced to procedure.

The argument that AI will make most human work obsolete follows the same misreading. It confuses the rule that stood in for judgment with the judgment itself.

AI is extraordinarily good at generating and running rules. If your work was executing it, the ground has shifted. But the rule never created the value. The value came from whoever noticed what the situation actually required, knew when the rule was wrong for this case, and made a call that cost them something if they were wrong. That capacity does not get cheaper when AI arrives. It gets rarer.

I use AI in my own work. It does not tell me what to think. It gives me more surfaces to think against. The meaning comes from my side. The tool amplifies what is already there. If nothing is there, there is nothing to amplify.

If you think writing better prompts will save your job, it might. It will not save your career.

The beer at my brewery is mashed by Michaël. The team around him is part of what you taste. If the person serving you in the taproom makes you feel genuinely happy to be there, the beer tastes better. Not many people account for that. You can analyse every molecule and reproduce the recipe exactly. You will not reproduce any of it.

AI can match the thing. It cannot match the person behind it. For certain things, the person is the value.

The same is true of your grandmother’s cooking.

AI runs on electricity and silicon. It does not metabolize. It does not pay a cost to exist. It does not die.

A system that pays no cost for being wrong has no reason to develop judgment. Judgment emerges from exposure to consequence. The feedback loop requires that something be at stake.

AI has no stake. It cannot have an autoimmune response. It cannot mistake native capacity for a threat because it has no native capacity to protect. The failure mode playing out right now in education, in corporations, in regulatory bodies, belongs entirely to us. AI is the stress that made the condition visible.

The processing gets faster. The judgment gap gets wider. Biology and physics do not negotiate with institutions.

Whether anyone inside those institutions is still reading clearly enough to manage it is a genuine open question.

The trainers in Cameroon are part of that answer.

So is every room that has already decided it knows what the word means.

On Friday afternoon, the first of Ramadan, my colleague suggested we visit the African Renaissance Monument before the weekend. A must when you’re in town, she told me.

So we went.

I climbed the stairs. There were very few people. The sky was blue. The harmattan dust had followed us all week. At the top, fifty-two meters of bronze: a man, a woman, a child lifted toward the horizon. The man points northwest. Toward the Atlantic.

I was impressed. I took a photo and sent it to a friend.

His reply came back quickly. I’m paraphrasing: a monument to Africa’s emergence from ‘obscurantism’, built by North Korea.

I stopped walking.

That single sentence didn’t diminish what I was looking at. If anything, it made the monument more compelling. In that contradiction, something became clear. Liberation, funded by one of the most closed regimes on earth.

A monument is not a policy. It is not a memo, or a committee recommendation, or a five-year plan. It is judgment made permanent. Stone, bronze, concrete on a hill. You cannot walk it back. You cannot add a footnote. You cannot call a meeting to revisit the decision.

A head of state wanted to mark something. He had a vision: an Africa emerging from darkness, reaching toward its future. He also had a problem: the vision required a builder, and the builder required payment.

He hired Mansudae Overseas Projects, North Korea’s state-owned sculpture company. The same hands that build the statues of Kim Il-sung. He did not pay in cash. Public land was transferred. Later, he claimed a significant share of tourism revenues.

Those contradictions are now fifty-two meters tall. The decision will survive the man who made it real.

None of that was written on the plaque.

He paid a price. The protests were real. The criticism was sustained. The friction existed.

It did not disappear in a week. It did not dissolve in a news cycle.

A monument must survive the reaction to it. A social post does not have to. A monument is built to outlast its author. A social post is not meant to.

The impulse is not new. The arena is.

What has changed is not human behaviour. People have always sought advantage. What has changed is what friction costs the person who receives it.

A scandal once required sustained institutional pressure to become consequential. A protest had to persist. An opposition had to organize. Time was required.

Visibility now moves faster than consequence. What once took months to absorb takes hours to scroll past. The influencer monetizes the audience without apology. The executive claims credit for what the institution built. The friction arrives, generates engagement, and disappears.

The behaviour is not new. The cost of receiving friction is.

Senegal is predominantly Muslim. The imams objected to the figures. Thousands protested in the streets on the day of the unveiling. Their own judgment, paid in public. The opposition argued that a head of state cannot claim intellectual property over an idea conceived in public office. The monument had been paid for with public land, between thirty and forty hectares of it.

Most leaders, facing that bill, choose a committee instead. They commission a study. They wait for consensus. They build nothing.

I never asked anyone there what the monument means to them. An oversight.

I came down the stairs the same way I went up. Almost alone. The harmattan dust was still following us.

The harmattan dust can blur the horizon. It can sand the bronze over time. It cannot undo that a decision was made and paid for in public.

In an entrepreneurial class in Dakar, I heard a case that stayed with me.

One of the participants runs a small handmade goods business. She was describing a recurring problem with her clients.

A customer would see pictures of handmade handbags in what she called her catalog. They would then place an order expecting the exact same bag. Same tone. Same grain. Same stitching. And when their bag arrived, they were disappointed.

The leather was slightly different. The shade warmer or cooler. The texture not identical. The model was the same. The dimensions were respected. But it was not a replica.

She explained this defensively, as if she was letting the customer down, as if she had somehow failed to control a variable that industrial producers manage through grading, blending, and standardization.

What struck me first was simple. There had never been a promise of industrial replication. These were handmade goods. Artisans were taking responsibility for the whole process, from sourcing materials to assembling the final piece. Even the leather itself had been prepared by another artisan upstream. Each bag began with a different hide, cut at a different moment, stitched under slightly different tensions.

But what struck me second was this. She called it a catalog. It was not. It was a portfolio. A catalog promises replication. A portfolio shows what the process produces. The bags in those photos were already sold. They had existed once, in the hands of someone else. There was no original to reproduce. Each piece begins again.

The variation was not a defect.

It was the process.

The real problem was not quality. It was the standard being applied. An industrial language of replication had been imported into a craft context. Identical had quietly become synonymous with excellent.

The shift does not start with the customer. It starts when the producer recognizes that variation is part of the making.

Craft beer went through this shift years ago.

Early craft brewers were judged against industrial lagers. Consistency meant sameness. Batch variation was interpreted as instability. But over time, the category reframed the conversation. Brewers stopped apologizing for seasonal shifts, for hop harvest differences, for the living nature of fermentation. They explained it. Customers followed. Wine had shown the way.

No one expects two hop harvests to produce identical aroma, or fermentation to behave the same across seasons. Industrial brewers work against this reality. Specifications exist to flatten whatever the season produced into a predetermined target. Same beer, regardless of where or when it was made.

Craft brewers work with it instead.

That was a turning point. When aggressive filtration was no longer assumed, haze stopped being a flaw. It became visible evidence of process. Beer was no longer presented as a sterile beverage engineered for indefinite uniformity. It was allowed to show signs of life.

That is how beer was made before industrial production reframed uniformity as the standard.

What changed was the expectation.

The contrast becomes clearer at the edges of industry.

Consider Hermès. Each bag is made by a single artisan who leaves a discreet stamp inside, pressed into the leather where only a repair technician or a curious owner would find it. A coded mark identifying who carried the piece from beginning to end. Not a brand. A signature. The system is disciplined, but the discipline protects continuity, not replication alone. Within that continuity, subtle variation appears. Differences in tension. Differences in pressure. Differences in rhythm. Not flaws. Signatures.

Now consider pre-distressed denim. The factory reproduces the appearance of wear, abrasion, fade, tear, but compresses what would normally unfold over years into a controlled treatment applied in hours. Something does happen. Sanding. Washing. Chemical fading. But the garment skips the lived sequence of friction, movement, and time. The experience is simulated rather than accumulated.

There is a reason we can usually tell. The fade on processed denim does not follow the logic of a body moving through the world. The wear lands in the wrong places, distributed too evenly, like an AI-generated image that resolves light too perfectly. Close enough to pass a glance. Wrong enough to feel off on contact. Authenticity leaves traces that are difficult to fake precisely because they were never designed.

Only one contains duration.

Only one records the hand that made it.

Industry achieves continuity by narrowing variation within tolerances through grading, blending, surface treatment, and statistical control. Craft achieves continuity differently. It preserves responsibility across the entire piece.

The difference is not between imperfect and perfect.

It is between fragmented and continuous.

The factory can standardize output.

It cannot standardize attention.

And over time, that difference shows.

The entrepreneur in Dakar was not describing a quality problem. She was describing a framing problem that began the moment she called her portfolio a catalog.

The bags in those photos were already gone. Made for someone else.

The next bag would be made with the same hands, the same attention, the same process.

It would not be the same bag.

It would be the next one.

While working in Douala, Cameroon, I was part of a program with local trainers focused on entrepreneurship and digital adoption. Our conversations often turned to pricing, positioning, and customer hesitation. We were trying to understand how businesses signal who they are and what they stand for.

It was my first time there. After long days in the classroom immersed in discussions from the business point of view, I found myself in a different role.

Now I was a customer.

So we asked a simple question:

Where should we go for dinner?

We didn’t say we wanted something local and informal. We didn’t say we wanted to avoid places designed for business travellers. We didn’t describe the experience we were looking for.

We just asked.

The answer came back quickly: You should go to the Vault.

Have you been there?

No.

Multiple people. Same recommendation. Same admission.

Recommending something “safe” also protects the person making the recommendation. If it goes wrong, the risk is lower.

When we arrived, the positioning was obvious. Top floor. Great view. Attentive service. Polished experience. It was clearly built for business travellers. The kind of place where you take an out-of-town guest for a reliable, impressive dinner.

It just wasn’t what we were looking for.

Nothing deceptive happened. We left a gap in our question, and people filled it.

In the absence of clarity, we became a persona: foreigner, on a business trip, looking for something reliable. Safe choice.

We were not the intended market in terms of what we wanted. But we fit the intended persona. The restaurant had positioned itself clearly enough that even people who had never been there understood what it represented.

Before you ever see an advertisement for a restaurant, a product, or a service, you already have an idea in your head about what it will be like. We rarely process everything. We look for coherence. When information is incomplete, we complete it.

It only takes one signal: a confident recommendation, a familiar name, a story that fits.

Marketing isn’t what a business says about itself. It’s the set of expectations that form around it.

Markets don’t wait for you to define yourself. They infer.

The Tension

I see the same pattern constantly with fellow entrepreneurs.

They say they need marketing. What I observe instead is customers hesitating. Asking the same questions. Seeking reassurance before moving forward.

Often, the instinct is to improve the surface: better visuals, a cleaner logo, a stronger online presence.

But hesitation rarely comes from weak graphics. It comes from uncertainty. A better logo doesn’t resolve a flawed offer. Logos and signage don’t matter much when everyone tells you the guy on that street corner has the best fresh braised fish. The signal that spreads wins.

Customers hesitate when they don’t know the price until the last minute, when they don’t understand how the process works, when they don’t know what happens if something goes wrong, or when they’re unsure what choosing you says about them.

Sometimes the risk is functional: will this work? Sometimes it’s social: what does this choice say about me? Most decisions contain both.

The Vault wasn’t just “safe.” It signaled modern, reliable, appropriate. That signal travelled among people who had never been there. They weren’t recommending a meal. They were recommending a position.

If you’re not clear about the tension you resolve, the market will decide for you.

The Cost

In resource-constrained environments, this becomes visible fast. Feedback loops are tighter. Confusion costs more. When a customer hesitates and walks away, there’s no ad budget to bring them back. What isn’t coherent collapses quickly.

In more comfortable markets, confusion can survive longer. Budgets absorb friction. Advertising can mask weak positioning for a while. But the underlying dynamic is the same. Tension that isn’t resolved doesn’t disappear. It just gets more expensive to work around.

The Work

Marketing reduces tension. It makes the outcome feel predictable.

It does this through clarity, consistency, and repetition. Clarity reduces friction. Consistency builds trust. Repetition turns perception into memory. And memory becomes expectation.

When this works, customers move toward you before you say a word. When it doesn’t, you re-explain yourself constantly.

You’ve succeeded at marketing when you’ve filled a void for your customers, and your business will succeed when the expectations you’ve built correspond to a market large enough to sustain you.

Clarity doesn’t mean narrowing yourself to nothing. It means choosing which ambiguity you keep.

When you leave your position undefined, someone else defines it for you.

The market is already forming expectations. Deliberate or not, they become real.

Their portfolios are diversified.

They believe their investment fees are low.

They stick to glancing at a performance chart, just to see how much they earned.

Everything important seems to be handled.

That is usually when substitution has already taken place.

This is the fourth essay in a series about what we stop seeing when we choose rules over judgment. The earlier essays traced the same pattern across different domains. They showed how resilience weakens when real signals are replaced by mediated ones, how judgment fails to develop when tools remove us from the feedback loops that build it, and how craft erodes when attention is replaced by systems that cannot notice what falls outside their design.

Taken together, they show that some kinds of judgment disappear once you turn them into rules.

The Pattern

As systems scale, judgment becomes expensive. It is slow, contextual, and difficult to standardize. Rules, procedures, and abstractions are cheaper. They travel well. They promise consistency. They feel neutral.

So judgment is not eliminated. It is substituted.

The substitution works best when conditions are stable. It produces efficiency, repeatability, and speed. It also thins sensitivity. Over time, the system stops reading signals and starts reinforcing its own structure. What looks like control is often a loss of feedback.

The danger is not failure at first. It is success without understanding.

In the years before 2008, financial engineers built instruments of remarkable sophistication. Securitization. Collateralized debt obligations. Credit default swaps layered on top of other credit default swaps. The people building them understood the mechanics. Few understood what they were building inside the larger system. The models said safe. The pressure accumulated where no model was looking. They were playing catch with a grenade because they liked the way it felt when they threw it.

In practice, this also looks like competence that cannot respond when something drifts. A brewer who relies on a supervision dashboard designed to catch overheating misses a stuck valve that is overcooling the tank. The temperature is displayed. The alarm stays silent. The system is doing exactly what it was built to do. The judgment that notices the mismatch has been delegated away.

Nothing breaks immediately. What fails first is the ability to notice.

The crash appears sudden.

The erosion was not.

Substitution at Scale

Finance offers the cleanest example.

Exchange-traded funds, or ETFs, are now the default way most people participate in markets. They are praised for simplicity, low fees, and broad exposure.

Price discovery predates modern finance, but for most of market history it depended on dispersed human judgment at the margin. Buyers and sellers disagreed about value, risk, and uncertainty. That disagreement was not noise. It was the signal.

Judgment scales poorly. It requires attention.

So it was substituted with a rule.

When you add an S&P 500 ETF to your portfolio, your money does not enter the market evenly. It follows a rule. Every new dollar is allocated in proportion to market capitalization. The largest companies receive the largest share of the flow, not because anyone reassessed their fundamentals, but because they are already the largest.

In practice, that means a significant portion of each new investment is directed into the same handful of companies that dominate the index. The so-called Magnificent Seven absorb a disproportionate share of inflows simply because the structure requires it.

You think the Magnificent Seven are overpriced. You avoid them individually. But you buy an S&P 500 ETF, and roughly a third of your money flows there anyway. You are not choosing them. You are not rejecting them either. The structure decides.

As more money follows the same rule, prices rise where flows concentrate. Rising prices increase market capitalization. Increased market capitalization increases index weight. The next dollar follows the same path. The feedback loop tightens.

Diversification was meant to reduce overall risk by combining exposures that behave independently. Many modern tools preserve the appearance of diversification while quietly increasing correlation. Judgment about how risks relate is embedded in the structure itself, expressed through the rules that allocate capital. What looks spread out on paper can still fail together.

The abstraction works by removing the need to look.

When conditions change, the system does not adapt through discretion or thought. It reacts mechanically.

Financializing Coherence

Bitcoin presents a different expression of the same mechanism.

Whatever one thinks of it, Bitcoin does not pretend to be something it is not. It has no cash flows, no management team, no roadmap. Its appeal rests on explicit properties: fixed supply, censorship resistance, self-custody, and independence from institutional intermediaries.

Those properties are not incidental. They are the point.

The Bitcoin ETF does not alter the protocol. It alters what ownership means. Self-custody becomes custodianship. Sovereignty becomes convenience. Ownership becomes exposure. Counterparty risk and institutional dependence are reintroduced by design.

The properties that gave Bitcoin coherence are abstracted away. What remains is price behavior. Volatility. A ticker symbol.

The question shifts from what this is for to how it trades.

This matters because people do not only remember losses. They remember embarrassment. When an asset becomes part of identity, a statement about values or being right, its collapse wounds more than a portfolio. It wounds self-conception.

Abstraction dulls that signal too. Exposure feels safer than ownership until accountability suddenly returns.

This is not a moral claim. It is a structural one. A system designed around direct alignment is re-expressed through an abstraction that scales better but adapts worse.

The Same Drift

The same substitution appears wherever tools promise to collapse judgment into procedure.

In Douala, I watched academics reject ChatGPT reflexively. Predictably, some students would use it to avoid reading. The tool encourages shortcuts.

But the entrepreneurs in the room saw it differently. If a book mattered to their business, they would read it. They would use ChatGPT to clarify sections they struggled with, not to replace engagement. The tool amplified judgment they were already exercising.

The difference was not the tool. It was where responsibility sat.

Refusal can become its own kind of abdication. Using a tool as authority outsources judgment upward. Rejecting a tool as a badge outsources it sideways. Both avoid responsibility.

Automation improves precision while thinning sensitivity. AI accelerates access to information while tempting us to mistake output for understanding.

A user who trusts the output no longer notices when the question itself was wrong.

Nothing breaks immediately. What fails first is the ability to notice.

The question is not whether tools are used. It is where judgment sits.

Tools can amplify judgment. They cannot replace it.

A Closing Posture

You can use the tools and still think.

You can automate and still pay attention.

You can delegate execution and still remain accountable.

What matters is whether responsibility has been handed off quietly.

The first essay in this series was about refuge, the abstractions we retreat to when signals become hard to read. What I can say now, having followed this pattern from Douala to the trading floor, is that every substitution in this series is a refuge. The feedback loop we skip. The dashboard we trust instead of the tank. The rule we follow instead of forming a view.

Refuges feel like answers. They are often the place where reading stops.

A refuge you cannot leave is not a refuge. It is just a different kind of exposure.

Using the tools well means knowing what they cannot do.

The feedback loop is not inefficiency. It is how judgment stays alive. You notice. You act. You learn. You do it again.

I’ve run a craft brewery since 2014. Long enough to know that tradition is mostly a story we tell after the fact.

This is the third essay in a series about judgment and substitution. It uses brewing to look at a wider question: what happens when we stop paying attention and hand off responsibility.

What Are We Protecting?

Is that cheating? Are you really a craft brewer? Will the craft police raid your brewery to strip you of your mash paddle and revoke your beard privileges?

A brewer asks if you use ChatGPT for recipes.

The question is not really about software. It is about trust.

The implication is familiar. If a machine helped, something human was replaced. Something sacred was violated. The beer is less honest. The brewer is less authentic.

But before answering that, we should ask a better question.

What exactly are we trying to protect?

Purity

Before hops, brewers used gruit. Blends of herbs, roots, and whatever else seemed to work. The mix varied by region and season. Some of it was harmless. Some of it was not.

One documented ingredient in some regions was black henbane. A nightshade. Hallucinogenic in small doses. Toxic in larger ones. It caused dry mouth, which made drinkers thirstier, which made them drink more, which pushed some of them into the danger zone.

That was not tradition. That was chemistry without understanding.

Fast forward to Québec in the 1960s. Dow Brewery added cobalt sulfate to improve foam stability. It worked. It also killed around fifty people.

At some point, someone should have paid attention to what we now call the Reinheitsgebot, often translated as the German purity law.

But even that story is not what we pretend it is.

It was not about romance. It was not about soul. It was an economic and safety regulation, established in 1516, designed to control inputs, protect grain supplies and bread prices, and avoid poisoning people. It restricted beer to barley, water, and hops. Yeast was not excluded. It simply had not been identified yet.

For centuries, it was not even described as a “purity law.” That framing only emerged in the early 20th century, when the regulation was retroactively branded as a symbol of quality and tradition.

We retrofitted the romance later.

This obsession with purity was distinctly German. Belgian brewers took a different path. They worked with what they had: leftover grains at the end of winter, local fruits, spices, and later sugar. Sugar was often added deliberately to raise alcohol without overloading the beer and, in some periods, to work around malt-based taxation, which was tied to grain use rather than final strength.

Not rebellion. Adaptation.

Brewers love a good story. Especially the one you tell between sips in the taproom. But tradition did not descend from the heavens intact. It was edited, enforced, revised, and eventually marketed.

This matters, because when people invoke purity, they usually are not talking about history. They are talking about the moment they personally stopped feeling comfortable with change.

The Line You Drew

If tradition is constructed, what are we really protecting when we reject new tools?

Gas-fired vapour-barrier kettles. Glycol chillers. Laboratory yeast. Water chemistry adjustments. Cryo hops. Centrifuges. Automated carbonation systems.

None of these existed in 1516.

Is it authentic to brew a stout in a place where the water has none of the mineral profile of Dublin? Of course not.

But it can be deeply authentic to the style of a dry Irish stout.

We manipulate the water precisely so the beer behaves the way the style demands.

The line between acceptable and unacceptable technology is rarely principled. It usually sits exactly where the speaker got comfortable.

Everything before that point is tradition. Everything after it is cheating.

The Wedding

On a recent trip to Cameroon, a local collaborator asked me a question.

“So your recipes, you must protect them so no one can replicate them?”

Not really.

We focus on clean tanks, the best ingredients we can get, and a motivated brewer who knows their system. After that, replication is not the risk people think it is.

Craft is not defined by secrecy. It never has been at its best.

It is defined by three things:

Real ingredients. Engaged technique. Human supervision.

Craft allows borrowing. It allows iteration. It allows improvement.

Tradition works a lot like a wedding. The vows matter. The commitment matters. But no one pretends the relationship should freeze at the ceremony. Craft is not about preserving the moment you said “I do.” It is about what you build after.

Something new is great. Something borrowed from another brewery is inevitable. Something blue, to turn the beer blue, probably not.

What craft does not allow is abdication.

Four Camps

When new tools appear, brewers tend to fall into one of four camps.

Purists Clean tanks. Clean beer. Everything else is noise. We ferment sugar. We do not serve it. If it needs glitter, the beer failed.

Experimenters Curious. Iterative. Willing to try. Prepared, when necessary, to dump a batch.

Undercover users Quietly using new tools and techniques while publicly criticizing them.

Unaware brewers Do not know what they do not know. Focused on output, with only a partial understanding of the inputs. Rare, but they surface when the system behaves outside its normal assumptions.

By default, in a living craft, we should all be experimenters.

Imagine if beer had stopped at its first iteration. Most early beers were not the result of choice, but of environment. Local water. Available grains. Wild yeast they did not even know existed.

Yeast was not identified as the driver of fermentation until the mid-19th century. Before that, brewers worked by observation, not explanation.

The purists are not wrong to care. They are wrong to freeze there.

The craft evolved because curiosity outpaced comfort.

The Solenoid

ChatGPT can generate a recipe for a Belgian Tripel in seconds.

It cannot tell me if my yeast is stressed because I underpitched. It cannot tell me if my thermometer or flow meter is drifting. It cannot tell me if fermentation is throwing heat instead of character. It cannot tell me if the right move today is patience, intervention, or dumping the tank.

The craft is not following the recipe. The craft is knowing when to ignore it.

We have a PLC that controls fermentation temperature. It opens a solenoid to circulate cooled glycol whenever the beer drifts out of its target range. It is precise. There is an alarm if the temperature stays out of range for more than a few minutes. But the system is designed to cool, not warm. There is no reason for it to do the opposite in normal operation.

One Saturday, I stopped by the brewery and, out of habit, checked the fermentation temperatures. Not because an alarm went off. Just a reflex. A solenoid had stuck open. Glycol was still flowing. The temperature was dropping. Fermentation was slowing fast. The system thought everything was fine. The recipe was correct. The dashboard was calm.

If no one had physically checked the tank, that batch would have stalled and died quietly.

We caught it in time. Not because the system knew. Because someone did.

AI can surface patterns. It can summarize known techniques. It can remind you of things you already forgot.

It cannot taste. It cannot smell. It cannot notice that something feels off.

And it certainly cannot take responsibility for the outcome.

The Gap

This is not AI versus no AI.

It is expertise versus unexamined execution.

Give the same recipe to two brewers in the same place. One may see a starting point. The other may see an answer.

Information scales instantly. Experience compounds slowly.

The deeper problem is not bad answers. It is unasked questions.

You do not Google a problem you do not know exists. You do not prompt ChatGPT about underpitching if you have always used the same amount of yeast and never realized the cell count had changed.

Experience does not just answer questions. It reveals which questions matter.

That gap is where craft lives.

A recipe can tell you how to brew a beer.

Only time in the brewery tells you whether you have learned anything.

If a brewer outsources judgment, they stop practicing the craft.

But if your brewer is trying to make better beer, willing to do the research, to test different approaches, and to understand the parameters they are working within, then you have a great brewer.

Tools do not erase craft. Abdication does.

It looks at how real know-how is built through experience, and what gets lost when we lean too heavily on instructions or tools. The later essays return to this idea in other contexts.

I recently went to Douala, Cameroon, for the first time, for a business contract involving a digital literacy program. Traveling with someone who had been there before, everything went smoothly when we arrived at the airport. There were more checkpoints than usual, but she knew exactly where to go, which steps mattered, and which did not. What looked complicated from the outset was, in practice, straightforward if you knew the system.

That pattern, complexity that dissolves with expertise, repeated itself the moment we left the airport.

We were heading to the apartment. It was Saturday night, around 8 p.m. The roadsides were crowded. Motorcycles were everywhere, often overloaded, moving quickly through tight gaps. People filled the streets. It felt less like disorder and more like a city fully awake, people out enjoying a hot, humid evening.

I have driven in Europe and South America, in conditions that demand attention. Still, I do not think I would drive in Douala. What I saw looked like anarchy. I would have decision paralysis. Cars, motorcycles, and pedestrians shared the same space, with little visible regard for lanes, formal priority, or even direction of travel.

And the honking. Constant.

Not like Manhattan, where honking often sounds like frustration spilling outward, an extension of what is already happening inside the driver.

The next day, well rested and sitting outside, I heard it again. Unrelenting honking. Noise. Chaos, or so it seemed. But it did not take long that first week to realize it was not noise at all. It was language.

In Douala, honking communicates:

“I’m beside you,” to a motorcyclist. “Passing left,” to the car ahead. “Watch the gap,” to the truck.

A communication protocol learned through immersion, not instruction.

From outside the system, it sounds like meaningless noise. From inside, it is precise, functional, and understood.

I would not drive there. I do not know the code. You cannot learn this from instructions. No manual. No tutorial. No prompt will teach you. You have to be in it. You have to put skin in the game and learn.

With a feeling of inferiority, I would tell my driver that I had great skills driving on ice in Canada. Plenty of practice.

This is what we lose track of when we talk about AI and expertise.

The difference between know-how and how-to.

We need know-how. But we default to how-to. It is the path of least resistance to ask for instructions. The harder question is whether those instructions are actually the right path.

The False Promise

In the literacy program I am working on, before anything else, I have to put AI in context for small business owners. What can it actually do for them?

We keep hearing that ChatGPT promises to collapse the gap between novice and expert. Some people are starting to believe it.

Information scales instantly. Experience compounds slowly.

I recently heard Anthony Scilipoti on a podcast discussing how young accountants acquire expertise. For an ambitious CPA, this traditionally means years of grunt work: audits, endless questions, and long days dealing with difficult, impatient clients.

Now many assume AI will simply remove that grunt work altogether. I can almost hear the collective sigh of relief from accounting students around the world.

Scilipoti offered a warning. Those years of grunt work are not wasted time. They build pattern recognition. You develop a sense that something is off. A useful bullshit detector.

Those years are a feedback loop: mistakes, corrections, and thousands of iterations. The real value of a CPA is not rule application or tax code recall. It includes the ability to read financial statements and sense what feels right versus what feels wrong.

That judgment comes from exposure, not information.

AI can process data, perhaps faster, perhaps more thoroughly. But can it replace the judgment that forms through lived exposure?

The answer matters. Because if we are wrong, we are not just making work easier. We are quietly eroding the very thing that makes expertise valuable.

When know-how defaults to how-to, we mistake having instructions for understanding the system.

I learned this gap firsthand while working in finance in the 1990s.

My Boss and the Spreadsheet

Working in investment banking in the 1990s, I became proficient with Excel. It was a tool I relied on constantly: financial projections, discounted cash flow models, sensitivity analysis.

Need to change an inflation assumption? One entry, and the entire model updates. Sensitivity analysis? Fine. How many scenarios?

It was fast. Accurate. Powerful. In negotiations, we could test outcomes on the fly. As an associate, I mastered the tool. I ran thousands of iterations.

My boss was exceptional. Sharp, quick, deeply knowledgeable. He once told me he had learned to build discounted cash flow models using a pen, paper, and a calculator.

That sounded barbaric to me.

Change an assumption? Then you had better tell me everything up front, because I am about to sit down and redo the entire calculation by hand. No undo button. No instant recalculation.

The cost of a mistake was high. You had to think through the relationships between variables before committing anything to paper.

That friction had an unexpected benefit. It made him never take the structure for granted.

Whenever he asked me for scenarios, he would glance at my spreadsheet for maybe thirty seconds.

“Something’s off,” he would say.

My reaction: Oh shit.

He was always right.

I had not made a calculation error. The formulas were correct. The numbers reconciled. But the structure underneath did not hold. The logic was wrong.

He did not just see the numbers. He felt the structure beneath them.

Not because his Excel skills were better. They were not. He had internalized what coherent models feel like and could sense when one assumption pulled everything out of shape.

When you build projections by hand, you cannot hide from the logic. Every assumption ripples through the entire model, visibly. Over time, you develop an intuition for what holds together and what does not.

Excel amplified my analytical capacity. No question. It made me faster, more efficient, able to test scenarios that would have taken him hours.

But it could not manufacture the judgment that comes from high-cost iteration.

The tool gave me speed. Experience gave him signal detection.

This is not unique to finance or Excel. The pattern repeats whenever we confuse tool proficiency with expertise.

The Double Trap

Two things happen when we default to how-to. Both feel like efficiency. Both quietly hollow out judgment.

The first trap is the illusion of access. We think having the output means we have the expertise.

You want to start a business. You prompt AI for a business strategy. It gives you a framework that sounds sophisticated. But you have not developed the judgment to know whether it fits your specific context, your constraints, your customers.

The output looks like expertise. It is not.

The second trap is distraction from the core task. This one is more insidious.

Tools do not just offer shortcuts. They redirect attention away from the activities that build expertise.

The business owner who used to write her own email campaigns learned which subject lines got opens, which offers fell flat, which tone made people reply. Now she prompts AI for “a promotional email for my spring sale.” The output is fine. She sends it. But she is no longer in the feedback loop. The intuition that would have compounded over dozens of campaigns never forms.

This happens gradually. The tool keeps producing results. The outputs look fine. You feel productive.

But you are outsourcing the very activities that would have compounded into expertise.

You do not realize what you have lost until you need to make a call without the tool. Or the environment shifts. And you discover you can no longer read the signals yourself.

The Highway and the City

Not all tasks require lived experience. Some environments are stable enough that how-to actually works.

Changing the shock absorbers on my golf cart? I watched a YouTube tutorial, extrapolated to my model, and got it done. The task was bounded. The variables were few. The system did not push back in unpredictable ways. Instructions were enough.

Does this mean I can now call myself a “leisure vehicle ride dynamics specialist”?

Think about driver assist. On a highway, it works. Predictable environment. Clear rules.

In Douala, it would be worse than useless. The car is programmed to follow traffic rules. Douala runs on a communication protocol no AI has learned.

The difference is not complexity in the abstract. It is whether the environment is stable or shifting, whether the rules are explicit or emergent, whether mistakes are reversible or consequential.

Most valuable work does not happen on highways. It happens in complex, shifting environments where pattern recognition matters and judgment cannot be automated.

That is where defaulting to how-to leaves you exposed.

What Gets Amplified

This is not an anti-technology argument. Using Excel in the 1990s let me work on far more deals than I could have on paper. Technology can be a powerful amplifier.

But amplification cuts both ways.

A microphone amplifies your voice. It is still your voice, just louder. If you have nothing to say, it is just noise.

The question is not whether to use the tool. It is when. Tool before judgment produces dependency. Judgment before tool produces leverage.

Information scales instantly. Experience compounds slowly.

We are tempted to collapse that gap with tools. But some things cannot be shortcut.

Pattern recognition develops through exposure. Judgment forms through high-cost iteration. Signal detection requires time in the system.

The honking in Douala makes sense when you are inside the system. You cannot learn it from a manual. You cannot prompt your way into understanding it. You have to be in the traffic. You have to pay attention. You have to survive it long enough to read the signals.

That is not romanticizing difficulty. It is recognizing how complex systems actually work.

Expertise does not come from instructions. It comes from surviving reality long enough to learn from it.

The path of least resistance is to ask for instructions. In Douala, instructions will tell you how to honk. Only in the traffic will you know if you’ve learned anything.