I was in a McDonald’s near a train station outside Lisbon, by a highway overpass and a bus depot.

The kiosk had a picture of coffee. A photograph of a coffee. But none of the buttons said “black coffee.” There was a button for a latte. A button for a cappuccino. A button for something called a “McFlurry Coffee” that I did not want to understand. I pressed latte, then tried to customize it to just coffee and nothing else. The screen asked me if I wanted whole milk, skim milk, or oat milk. It asked me if I wanted a flavor shot. It asked me if I wanted whipped cream. I declined all of it, and when the order appeared, it was still a latte. I abandoned the transaction.

The man next to me was maybe fifty, wearing a polo shirt and the expression of someone who has been asked to solve a problem he did not create. He had the same mission: a black coffee.

“I just want coffee,” he said to no one. The kiosk buzzed cheerfully.

A worker came around from behind the counter with a headset still hanging around her neck. She looked at the kiosk, looked at him, and did not ask what he wanted. She walked to a register that was dark (the kind that sits unused because the kiosks are supposed to handle the front lobby). The drive-thru was where she was needed. Cars cannot operate touchscreens, so the drive-thru still requires humans, and she had been up to her elbows in those orders before she saw him standing there. She tapped the screen a few times. A cup came down from the stack.

“Just coffee,” she said. Not a question.

“Just coffee,” he said.

Subscribe now

She rang it up on a register she was not supposed to be using. Took his cash. Handed him the cup. A short transaction that bypassed the entire system that was supposed to handle it. The whole thing took thirty seconds. The kiosk had been trying for five minutes and had not managed it.

That interaction is a parable about technology in 2026. We built a system that works great for the complicated order and fails at the simplest one. If you want four custom burgers with no pickles, extra cheese, and a specific ratio of ketchup to mayonnaise, the kiosk is faster than any human could be. If you want a black coffee, the machine becomes a puzzle.

This pattern shows up everywhere. The self-checkout at a grocery store demands that you “place item in bagging area” and then freezes when it cannot weigh an avocado. The automated phone system asks you to “say or press 1” and then cannot understand the thing you said. The airline check-in kiosk offers fourteen options before it lets you print a boarding pass, and by that time the human at the counter has already checked you in, tagged your bag, and smiled. In each case, the machine handles the complex use case better than it handles the simple one. The straightforward request becomes a bug.

The kiosk at that McDonald’s was not designed for the man who wants a black coffee. It was designed for the median order. The 80th percentile. The customer who wants a customized latte with oat milk and a flavor shot. The Gen X guy at the front of the line is the edge case that got optimized out of existence. The person who wants one thing, plain, without options, does not fit the interface. There is no button for no.

I thought about this again when McDonald’s had its global system outage in March 2024. Stores across Australia, Japan, and the UK shut down or accepted cash only because the network went dark. The kiosks froze mid-transaction. The app stopped working. Suddenly, the only thing that mattered was the person behind the counter with a register and a pen. The machines were bricks. The humans were the only option.

The worker who walked to that dark register and rang up the coffee by hand, she is the safety valve. The human override. She exists to catch the edge cases the machine cannot handle. That is fine until companies decide the override is too expensive and remove it. They will frame it as efficiency. Fewer people on the floor, more automation. The kiosk handles 80 percent of orders. Why keep paying someone to stand at a register all shift? But the 20 percent the kiosk cannot handle includes the simplest request in the restaurant. A black coffee. No customization. No upselling. Just coffee.

Take away the override and what happens to the man with the polo shirt? He either learns the machine or goes somewhere else. If enough people like him go somewhere else, you have a business model problem dressed up as a technology success.

I ordered my coffee from the same woman. She remembered me from watching the man before. Saw me hang back. Saw me waiting. She poured it black, handed it over, and said something in Portuguese that I took as friendly. I sat by the window and watched the buses arrive and depart. The kiosk blinked its menu at an empty lobby.

The air in late spring in Split is this perfect temperature where you don’t notice it. Sun high. Breeze off the Adriatic cutting the heat. Boats drifting in the harbour through the coworking window.

And I’m sitting here thinking about how the entire AI industry is about to have the rug pulled out from under it.

It won’t be regulation. It won’t be a breakthrough. The thing coming for AI is much more boring and much more inevitable. Economics.

Here is what I mean by “AI’s Spotify Moment.”

When Spotify launched, it didn’t invent streaming. It stepped into a market where the underlying product had been stripped of its value. A CD cost fifteen dollars. A stream cost fractions of a cent. The music itself became a commodity. So Spotify had to win on something else. Playlists. Discovery. Network effects. The wrapper became the product.

The AI industry is hitting that same curve. Faster than most people realise.

Subscribe now

Every week there is a new model release. Llama 4. Mistral Large. Qwen 3. DeepSeek V4. Gemini 2.5 Pro. Claude Opus 4. They all claim to be the best at something. They are all within spitting distance of each other. And they are all getting cheaper at a rate consumer hardware has never seen.

Run the numbers. In 2023, getting GPT-4 class performance cost around $30 per million output tokens. As of mid-2026, that number has fallen below $2 for several providers. Claude Opus 4.7 charges $5 per million input tokens. GPT-5 Nano runs at $0.10. Some open models run on consumer GPUs at fractions of that. The cost curve is not linear. It is falling off a cliff.

When inference becomes that cheap, the barrier to entry becomes zero. Everyone has access to near-frontier intelligence. The differentiation vanishes.

This is where the grey market becomes instructive.

There is a thriving economy of API proxy services, aggregators, and arbitrageurs who sell AI access at a fraction of the official price. In a piece on the hidden cost of cheap AI tokens, Souk documented how these “transfer stations” route traffic through unofficial accounts, swap models without telling you, and harvest your data as the real profit centre. The price difference looks like a hack. The reality is your data being monetised in ways you cannot see. (That Cheap Claude API You Bought)

But here is the part that matters for this argument.

That grey market exists at all. It exists because AI inference is already being treated as a raw commodity. The same way grey market GPU dealers treat compute as fungible units. The proxy economy proves that the market has already decided AI output is interchangeable. It is a widget. You buy it by the token, by the prompt, by the request.

And when the market treats something as a widget, the only question is who can sell it cheapest.

That is the Spotify Moment arriving early. The official market hasn’t caught up yet. The shadow market already lives there.

The implications are uncomfortable for anyone building in this space.

The model companies themselves are in the toughest spot. They are burning enormous capital on training runs, infrastructure, and talent. They are differentiating on benchmarks the market is increasingly ignoring. Every quarter, their product becomes cheaper to replicate. Open models are closing the gap. Hosted providers are racing to the bottom on price. The investor narrative is shifting from “who has the best model” to “who has the best distribution.” That is a brutal place to be if your entire company is built on the first thing.

The winners in this environment are the bundlers. The platform companies that wrap multiple models behind a unified interface, add security, compliance, data governance, workflow tooling. The ones who sell peace of mind, not intelligence. Companies like Vercel, Databricks, Snowflake, the AWS/GCP/Azure trio. They don’t care if the underlying model is Claude or Gemini or Llama. They care that you stay on their platform. The model is the loss leader. The infrastructure is the margin.

The losers are the point solution AI companies. The ones who built a wrapper around GPT-4 and called it a product. When the underlying model is a commodity, a wrapper is a thin margin business. And thin margin businesses in a race-to-the-bottom market do not survive.

This is also the moment where the open model ecosystem becomes genuinely dangerous to the incumbents. The danger is not open models being better. It’s that they’re free. When a Llama 4 class model runs on a laptop, and Mistral Large runs on a mid-range server, the question shifts from “can I afford the API” to “why would I pay for the API at all.” That calculation is already happening in enterprises. The cost of self-hosting is dropping faster than the cost of API calls.

The smartest people I know in AI are not building better models. They are building better products on top of the models that already exist. They are betting the intelligence floor rises high enough that the differentiation moves to interface, memory, personalisation, reliability, tool integration. The thousand small things that make a product feel like it was built for you.

That is the real Spotify play. Streaming was the technology. Curation was the experience.

The models are going to zero. The value is going to everything around them.

So if you are building in AI right now, the question is not whether you can access the best model. You can. Everyone can. The question is what you build that makes people stay. What you build that makes the model invisible, and the experience undeniable.

The breeze coming through the window at Smartspace is still the same temperature. The boats still drift. Outside this window, an industry is changing faster than anyone is ready for.

The question is who will be ready anyway.

In the summer of 2010, a user experience researcher named Harry Brignull was sitting in his flat in London, looking at a website that had made him angry. He could not remember exactly which site it was anymore. What he remembered was the pattern: a screen that asked him to do one thing but was designed, clearly, to make him do another. The button he wanted was small and gray. The button the company wanted was large and orange. This was deception, and it had no name.

So he gave it one. He bought the domain darkpatterns.org and started cataloging them. Confirmshaming. Roach motel. Hidden costs. Misdirection. Sneak into basket. He posted examples, invited submissions, built a taxonomy. The site went live quietly and grew slowly, mostly among UX professionals who knew exactly what he was talking about because they had been asked to build versions of those same patterns themselves.

Subscribe now

Brignull was a practitioner who had noticed something peculiar: the same tricks kept appearing across different sites and industries, in different countries. They followed a playbook. And fifteen years later, we are still playing from it.

Every marketer who uses a dark pattern believes they are making a rational trade — a small erosion of user trust in exchange for a measurable lift in conversions. The math feels right because the conversions are visible and the trust is not. You can see the signup spike on Monday. You cannot see the customer who stops buying in December because she finally noticed you were hiding the unsubscribe button.

Consider what happened at a large e-commerce platform in Australia. The company, which asked to remain anonymous in a case study from ExtraStrength, had been using a classic dark pattern for years: displaying one price at the top of the product page and revealing additional fees only at checkout. This is called drip pricing, and it works beautifully in the short term. The customer has already picked the product. The customer has entered their credit card. The customer is three clicks deep and the fee is fifteen dollars. Most of them just pay.

The company decided, for reasons that are not entirely clear, to run an experiment. They removed the hidden fees. They put the full price on the product page, no surprises. And they waited.

Conversion rates dropped initially. That was expected. But repeat purchases rose 20 percent. Not over a quarter. Not over a year. They rose, and they stayed up, and the company discovered something it had not accounted for: the customers who had not felt tricked came back. The ones who had felt tricked, the ones who had paid the hidden fee and resented it, did not complain or ask for a refund. They just stopped buying. The company had been measuring the wrong thing.


Lior Strahilevitz is a professor at the University of Chicago Law School who has spent years studying the line between persuasion and deception in digital design. He has a way of stating the problem that makes it sound almost simple. Dark pattern, he told me (and he is careful to put the term in air quotes), is not a legal term with a clear boundary. It is a label for a spectrum. On one end you have designs that are merely annoying, the digital equivalent of a pushy car salesman. On the other end you have designs that cross into deception and coercion. And the part that matters for marketers is that the spectrum does not divide cleanly along intent. You can be trying to persuade and still end up deceiving. The Federal Trade Commission, in its 2022 report> “Bringing Dark Patterns to Light,” made exactly this point. The legal question is whether a reasonable person would be misled by what they saw, not whether the company meant to deceive.

This is where the marketing frame cracks open, because marketers are not accustomed to being evaluated this way. They are used to being evaluated on outcomes: did the campaign lift conversion, did the email drive opens, did the landing page reduce bounce rate. But the dark pattern framework asks a different question. It asks whether the design is harder to decline than to accept. It asks whether the information the user needs is visible or buried. It asks what a user who is distracted and scrolling quickly, tired after a long day, would reasonably understand from this screen.

Those are questions that cut to the heart of a much older tension in marketing. You can date the tension to the moment someone first tried to sell a product by saying something about it that was technically true but designed to create a false impression. The fixer who quotes a low price over the phone and then adds a shop fee when the customer arrives. The credit card offer that says 0 percent APR in 72-point type and puts the terms in a paragraph no one will read. These are not new. What is new is the scale. A drip pricing pattern on an e-commerce site can run ten thousand times a day, every day, silently, across every visitor. The old car salesman had to look the customer in the eye. The digital version never does.

There is a study from the OECD, published in 2022, that estimated 76 percent of e-commerce sites use some form of manipulative practice. Not all of them are illegal. Most of them are not. The OECD was careful to distinguish between patterns that are merely aggressive and those that cross into deception. But the number itself is startling. Three out of four major shopping sites have decided, at some level of the organization, to make it harder for you to know what you are paying and easier for you to pay it.

And the math that drives those decisions is almost always the same. A product manager runs an A/B test. Version A has the subscription terms in clear language at the top of the form. Version B buries them in a collapsible section under the button. Version B converts 12 percent better. The product manager ships Version B. The decision makes sense in isolation. It makes sense in the quarterly review. It makes sense on the dashboard.

Subscribe now

What does not show up on the dashboard is the customer who, six months later, discovers the automatic renewal and spends forty minutes navigating a cancellation flow that asks three times whether she is sure. She does not post about it. She does not write to the CEO. She just decides, quietly, that she will never trust that company again. That decision does not appear in any A/B test. It appears, if it appears at all, as a downward tick in a retention cohort twelve months out, by which point someone else is running the dashboard.

The best name for this dynamic comes from the book Strategy Engine, which calls dark patterns negative-sum strategies. You get a short-term gain. The customer takes a long-term loss. And over time, the sum turns negative for everyone, including you, because churn rises, trust depreciates, and the compliance landscape shifts beneath your feet. India’s consumer protection authority classified thirteen dark patterns as unfair trade practices in 2023, with penalties reaching 10 percent of company revenue. The FTC won a $2.5 billion settlement from Amazon in 2025 over its Prime enrollment and cancellation practices. These are no longer speculative risks.

I keep coming back to Harry Brignull’s original catalog from 2010. He collected the patterns like a lepidopterist pinning butterflies, each one with a name and a description and an example. He was a taxonomist, not a reformer. But the act of naming changes things. Once you call something a roach motel pattern, it becomes harder to pretend it is just good UX. Once you call something confirmshaming, it becomes harder to pretend the guilt-tripping language is accidental. The naming forced a conversation that the industry had not been willing to have.

And yet, fifteen years later, the patterns are everywhere. Brignull’s catalog has grown from a handful of types to a taxonomy that researchers have expanded into dozens of subcategories. The Amazon settlement was for 2.5 billion dollars. And still, when you open a random shopping app and try to cancel a subscription, the button is small and gray and the button the company wants you to click is large and orange.

There is a reason for that persistence, and it is not malice. It is a measurement problem. The things you lose when you use a dark pattern — trust, goodwill, repeat business, advocacy — are hard to measure. The things you gain — higher conversion rates, more signups, retained recurring revenue — are easy to measure. Easy wins over hard every time, in every organization, unless someone deliberately designs the organization to weigh the invisible loss against the visible gain.

The e-commerce company in Australia figured this out when they looked at repeat purchases instead of first purchases. The B2B SaaS company in the Strategy Engine case study figured it out when they saw a competitor charging 20 percent more and still winning, because the competitor offered one-click cancellation and clear month-to-month terms. These companies did not stop using dark patterns because they had moral revelations. They stopped because they found a better number to track.

That is the hidden cost of the hidden button. You are optimizing a metric that points in the wrong direction. You are winning the visible battle and losing the invisible war. The customer who stays because they are trapped is a liability. The customer who stays because they want to is the only asset that compounds.

Brignull, in that flat in London fifteen years ago, was angry about a button on a website. But his real discovery, the thing he may not have fully understood at the time, was that the problem was the system the company had built — the system where the button made sense. The conversion rate said yes. The quarterly review said yes. Everything visible said yes. And the only voice saying no was the customer who left without telling anyone why.

In April, Mikie Sherrill did something that, in the history of World Cup hosting, no major politician had done before. She published a set of numbers on social media that told an uncomfortable story about who the tournament actually benefits. The numbers were devastating in their specificity: NJ Transit, the state’s public transportation authority, was looking at a $48 million loss from the eight matches at MetLife Stadium. Eight matches. Forty-eight million dollars. The state’s solution had been to charge fans $150 for a round-trip train from Manhattan to East Rutherford, a 20-minute journey that ordinarily costs $13. (The price later dropped to $105, but only after what she described as “some corporate donors” stepped in. NJ Transit still loses money.)

“When FIFA is positioned to make $11 billion during the World Cup,” Sherrill wrote, “we will not be subsidizing World Cup ticket holders on the backs of New Jerseyans who rely on NJ Transit every day.”

Subscribe now

This is the part of the sentence that matters. Eleven billion dollars. FIFA expects to generate $9 billion from this tournament alone, part of a four-year cycle that will bring in $13 billion total. The organization sits on $2.8 billion in cash reserves. Its annual report describes its financial position as “unprecedented.” And here is the New Jersey governor, on X, explaining that her state cannot afford to run trains for the people coming to watch.

She was right to say it. But the question nobody asked, in the noise that followed, is this: why is that $48 million the state’s problem at all? Why is it always the host’s problem?

The answer takes you into the strange economics of FIFA’s business model, which is a business model built on a single, elegant principle: concentrate the revenue, disperse the cost.

Consider the numbers that Sherrill put on the table. The $48 million transit deficit is one piece of a much larger puzzle. Toronto says it will spend $380 million Canadian. Vancouver is at $624 million. Most US cities budgeted around $200 million and are now well past it. The federal government in Canada committed an additional $145 million specifically for security. In Boston, Robert Kraft, the owner of the New England Patriots, had to personally front $8 million for a SWAT truck and security measures before the town of Foxborough would issue FIFA’s permit. Parking at some venues costs $175 a game. That revenue goes to FIFA, not the cities.

There is no tax exemption this time. Previous World Cups had them. The US government, unlike the autocratic regimes that hosted the previous two tournaments, cannot simply decree that FIFA’s partners pay nothing. The absence of that exemption means national associations from Europe are now realizing they could lose money unless they reach the semi-finals. The English FA, among others, petitioned FIFA to increase the prize money. FIFA agreed, bumping the total from $727 million to $871 million, but the math still does not work for anyone outside the last four. The winner gets $50 million. The group-stage team gets $12.5 million. The costs of operating in the United States, where tax rates vary from 0% in Florida to 13.3% in California, eat deep into those numbers.

Now here is the pattern. In Russia and Qatar, where FIFA could order up stadiums in autocratic silence, the tournament delivered what the governing body wanted. In the US, Canada, and Mexico, the host cities have to answer to voters. They have to explain why their transit systems are losing money. They have to explain why fans are paying $175 for parking. And they have to do it while FIFA pockets the ticket revenue, the sponsorship money, the broadcast fees.

"FIFA's contracts," as the Atlantic described them, leave hosts with "no plausible way to recoup expenses." The $100-plus train tickets were a desperate attempt to stop a money pit from going deeper.

There is a second dimension to this story, and it involves the world's two largest media markets. The situation there tells you something about FIFA's negotiating leverage that the $48 million transit deficit cannot.

India, with 1.4 billion people, has no broadcast deal for the 2026 World Cup. The tournament starts in less than a month. The reasons are structural, not negotiable. Indian broadcasters build their business around advertising, not subscriptions. Football does not break every three or four minutes the way cricket does, which means the commercial breaks that sustain the Indian sports advertising market simply do not exist. The time zones are catastrophic: most games kick off when India is asleep. And FIFA is not even competing for the top spot. The Indian Premier League and the ICC rights are the two most valuable sports properties in the country. The World Cup is not in the top two.

Nandan Kamath, one of India's leading sports lawyers, put it this way: "The Indian market is a sort of a brute force market. It's the numbers rather than the willingness." There are only two bidders for the rights, JioStar and Sony, and neither is desperate. "Normally these rights are sold where there's highly competitive people dealing with FOMO, and that isn't here right now."

China's story is even more revealing, because it ended with a deal, but on terms that should embarrass FIFA. The initial asking price was around $300 million for the Chinese market. After months of standoff, China Media Group closed the deal for something close to $60 million. That is an 80% reduction. The governing body of world football walked into the biggest media market in Asia, asked for what it thought the product was worth, and was told to wait. Then it waited. Then it took what it could get.

Subscribe now

Xu Guoqi, a professor at the University of Hong Kong who has written extensively about China and global sport, has a theory about why this happened. "I think FIFA got greedy," he told DW. "It's a business deal, right? For FIFA, if Chinese men don't watch the game that's a big loss to them." He referenced a historical pattern: in 1999, when NATO bombed the Chinese embassy in Belgrade, the Chinese government tried to cancel the broadcast of NBA games. Young people protested outside the US embassy during the day and denounced CCTV at night for complying with the blackout. "We hate American imperialism," they said, "but we love the NBA." The lesson, Guoqi argued, is that global sport occupies a space in Chinese culture that the government cannot control. FIFA assumed that urgency translated into willingness to pay. It does not.

FIFA priced its product as if the market had no alternatives. The market responded by walking away. China waited FIFA out and got the tournament for pennies. India is still waiting.

Which brings you back to the central question of 2026: what happens when you build a system that concentrates reward and disperses cost, and then try to sell that system in markets that do not need you?

Now consider the alternative. Because there is one. And it is coming in 2030.

Morocco, Spain, and Portugal will co-host the next World Cup. And Morocco’s role in that partnership is structurally the opposite of the 2026 model. Where 2026 spreads its matches across three countries separated by thousands of kilometers, 2030 concentrates its geography: Morocco sits 14 kilometers from Spain at the Strait of Gibraltar. A fan in Lisbon can be in Tangier before lunch. The tournament creates a single accessible corridor across two continents, the kind of layout that was possible in 1998 and 2006 and that has been impossible since.

The MIPA Institute studied the economic implications of this bid. Their findings are instructive, because they confirm what economists have suspected since Brazil in 2014: the World Cup does not generate the kind of windfall that politicians promise, but it does deliver real, measurable effects when the conditions are right. The tourism boost for World Cup hosts averages 12%, though the MIPA study notes that most of that effect arrives before the tournament, as an “anticipation effect.” Domestic league attendance in host countries increases 15 to 25 percent after the tournament, an effect driven by media coverage, improved stadiums, and the long tail of a national team’s performance.

Morocco’s $5 to 6 billion investment, part of a joint $15 to 20 billion budget shared with Spain and Portugal, is structured differently from the 2026 model. Twenty-five billion dirhams comes directly from the public budget for stadiums and training centers. Seventeen billion dirhams from state-owned companies, earmarked for infrastructure and transport. Ten billion dirhams from foreign loans and donations. The IMF estimates Morocco’s infrastructure spending will reach 11.9 percent of GDP by 2030, concentrated in rail (6.0%), airports (2.4%), roads (0.9%), stadiums (2.2%), and urban tourism (0.5%). That spending is expected to lift GDP by roughly 2 percent above the baseline. The country plans to add 100,000 hotel beds in host cities and projects tourism revenues of 120 billion dirhams in 2030.

The difference is structural. Morocco is not carrying the weight alone. Spain and Portugal absorb the majority of the logistical burden. Morocco gets the spotlight, the infrastructure, and the tourism upside, without having to build everything from scratch. It is a tri-host arrangement that distributes cost and benefit, which is precisely what the tri-host arrangement for 2026 was supposed to do but has not.

And the geography matters in a way that has not been true for any World Cup since 2006. The tournament runs in European summer, when the domestic leagues are closed. The three countries share time zones. They share cultural ties. They share football seasons. For the fan who wants to watch matches in multiple venues, which is to say, for the fan the World Cup is supposed to serve, this is the most accessible structure in decades.

The 10,000 Hour fallacy applies to hosting too. The question is whether you are building in the right place, in the right configuration, with the right partners.

None of this means Morocco 2030 is unproblematic. The same questions that dogged Qatar will follow the bid. Labour rights. Press freedom. Whether the tourism promises will materialise, or whether the white elephant problem will resurface. The MIPA analysis itself warns that host cities in developing economies often see unemployment rise after the tournament, because the temporary construction workforce does not disappear when the stadium is finished; it stays in the labor pool, looking for work that no longer exists.

But the structural logic is better. The geography is better. The distribution of cost and benefit is closer to fair. And that is the standard the 2026 World Cup has set.

Mikie Sherrill’s question, the one she posted on X in April, was a simple one. Why should New Jersey subsidize FIFA’s profits? It was a question that no host in the authoritarian model had the standing to ask, and that no host in the democratic model had the nerve to ask, until now.

The World Cup generates enormous wealth. The question is who gets to keep it. And if you watch where the hosting model is heading, from the $48 million deficit in New Jersey to the 14 kilometers between Morocco and Spain, the direction is clear. The next World Cup, the one after this one, will look very different from this one. Because the markets and the cities and the fans started saying no.

Sources: DW, Guardian, Atlantic, MIPA Institute, Chosun, Reuters, Economic Times, FIFA Annual Report 2025, IMF, CBC, City of Toronto budget documents

Sarah Chen, a hiring manager at a Series B SaaS startup in Austin, sat down on a Monday morning in January 2026 to review applications for a senior product role. Her inbox contained 2,400 submissions. She opened her ATS—applicant tracking system—and it had already filtered the pile to 40 candidates. A second AI tool ranked those 40 by keyword density. She scheduled interviews with six people.

Three of them couldn’t answer basic questions about the work listed on their resumes.

Subscribe now

When she asked one candidate about a specific project he’d listed, he went silent. “I used ChatGPT to write my resume,” he said finally. “I made that one up because it sounded impressive.” Sarah wasn’t angry. She was confused. The ATS had flagged him as a strong match. The keyword ranking had put him in the top ten. He looked perfect on paper because a machine had written the paper.

The resumes were written by ChatGPT. The ATS was screening for patterns that ChatGPT was trained to produce. Sarah was grading AI against AI. The humans were the last to know.

LinkedIn processes 11,000 job applications per minute now, up 45% year over year. The average job posting receives 286% more applications than it did in November 2023, according to Tribepad’s analysis. This isn’t because there are more jobs. This is because applying for a job stopped taking effort. ChatGPT generates a resume. It writes a cover letter. It submits to 50 listings before lunch. A candidate can apply to a hundred companies in the time it used to take to apply to ten.

The resume was already a weak hiring signal. Now it’s just noise drowning out signal.

Why machines prefer what other machines write

In September 2025, researchers at UC San Diego published a paper that should have shaken up every ATS vendor in the country. Jiannan Xu, Gujie Li, and Jane Yi Jiang tested four major language models—GPT-4, Claude, Llama, Mistral—to see if they had a bias when screening resumes. They did. Massively.

When an LLM screened a resume written by the same model, it rated that resume higher 68% to 88% of the time. A candidate who used ChatGPT to write her resume was 23% to 60% more likely to get shortlisted than a candidate with identical qualifications who wrote her own resume. The gap was widest in jobs that live in language: sales, accounting, operations, any role where the work gets documented through prose.

The researchers didn’t find a bug. They found a feature. Language models generate text with recognizable patterns—certain phrase structures, vocabulary choices, punctuation habits, formatting conventions. When another instance of the same model reads that text, it recognizes the patterns as “correct.” A human-written resume listing real accomplishments in rough, personal language gets scored lower than a machine-written resume listing fabricated accomplishments in smooth, pattern-matched prose.

The UC San Diego team tested interventions to reduce the bias. They found that targeting the LLM’s self-recognition capabilities could cut the preference by more than 50%. Those interventions don’t exist in any commercial ATS product. No vendor has built them. No customer has demanded them. The systems that rule hiring are working as designed, and the design favors machines.

What job seekers actually do

Ninety percent of job seekers now use ChatGPT to write their applications, according to Huntr’s Q2 2025 survey. Ninety-one percent of U.S. employers deploy AI somewhere in hiring. Ninety-nine percent of Fortune 500 companies rely on ATS systems that reject 75% of resumes before any human reads them.

The math is brutal for anyone trying to do this the old way. If you spend two hours writing a tailored resume and it gets screened out by keyword matching, you’ve wasted two hours. If ChatGPT writes fifty tailored resumes in the time you write one, and each one passes ATS screening at three times the rate of a hand-written resume, then using ChatGPT is the dominant strategy. A candidate who doesn’t use AI is leaving money on the table.

Robin Ryan, a career coach who writes for Forbes, watched this happen to a client in January 2026. The client’s AI-generated resume listed skills he didn’t have—cloud architecture, machine learning, database optimization. The resume looked polished. The ATS ranked him in the top candidates. He got called in for an interview and fell apart when asked to explain his experience. “The machine wrote a better resume than I could have,” he told Ryan afterward. “And then I couldn’t defend it.”

MIT’s Career Advising office recommends using AI for help with structure and editing, but with constraints: provide your own data, fact-check every claim, preserve your actual voice. The guidance is sound. Almost nobody follows it because the incentive structure makes compliance irrational.

A candidate who applies to two hundred jobs with AI-generated resumes and gets five interviews has better odds than a candidate who applies to twenty jobs with hand-written resumes and gets one interview, even if three of those five interviews end badly. The prisoner’s dilemma has a dominant strategy, and everyone’s playing it.

The recruiter’s side of this is worse

Sarah Chen’s inbox isn’t unusual anymore. The average recruiter now gets 250 applications for a single opening, up from 65 two years ago. The applications are longer, more polished, more identical to each other than at any moment in hiring history. Sixty-four percent of recruiters report seeing more lookalike applications, according to HireTruffle’s recruiter survey, and they all blame AI-generated resumes for the sameness.

The solution that most companies reach for is more AI. Better screening algorithms. Tighter keyword matching. This doesn’t work. It makes things worse.

When candidates use AI to write resumes and employers use AI to screen them, everyone looks equally qualified on paper. The signal disappears. The ATS ranks the top twenty candidates, but those top twenty might be the worst twenty actual candidates because the algorithm is rewarding what AI does best: generating text that matches a pattern. The ATS is blind to the difference between a candidate who knows the material and a candidate who knows ChatGPT.

Some companies tried AI detection tools to catch generated resumes. Candidates immediately learned to make AI resumes undetectable. Now it’s an arms race where everyone loses. The recruiter wastes time trying to distinguish real from fake. The candidate wastes time making the fake harder to detect. The only winner is the company that sells the detection tool.

Harvard Business Review published a piece in January 2026 called “AI Has Made Hiring Worse.” The argument was simple: when the primary signal is optimization skill, you’re hiring for prompt engineering, not competence. Companies that still rely on keyword-based screening are filtering for the wrong thing.

What actually breaks the cycle

The resume is dying. Not tomorrow. But as a primary screening tool, it’s already dead.

The companies that have figured this out are moving to a different model. They use AI for logistics: managing applications, filtering spam, scheduling interviews. They use assessments and work samples for signal: actual skills tests, live coding problems, short portfolio submissions reviewed early, not saved for the end. They use humans for judgment: managers reviewing top candidates based on what those candidates can actually do.

TestGorilla found that 85% of companies globally now use some form of skills-based hiring, up from 73% in 2023. McKinsey’s research showed that hiring for skills is five times more predictive of job performance than hiring based on education, and twice as effective as hiring based on work history. Employees without degrees who got hired through skills-based processes stayed 34% longer than degree-holders hired the traditional way.

For a startup, this matters because the math is simple. You can’t hire ten engineers by wading through two thousand polished resumes. You need to see them code. You need to see them problem-solve. You need to see them think in real time, not perform on paper.

Some companies ask for GitHub profiles. A profile with two hundred commits across six months is hard to fake. You can’t hire ChatGPT to build your portfolio. Someone actually wrote that code. Some companies run coding challenges. Some ask candidates to submit portfolios of real work. Some do short video introductions where the candidate explains a project in their own voice, in real time. The signal is the work itself, not the prose about the work.

How the system eats itself

If every candidate uses ChatGPT because the ATS rewards ChatGPT’s patterns, and every employer uses AI to screen because candidates are using AI, then the resume becomes a proof of how well you prompt-engineered, not a proof of what you can do.

This is what Nav Toor called the “mirror effect” in a thread on X in early 2026. Candidates are mirroring the linguistic patterns that AI screening tools reward. The feedback loop is closed. Candidates use AI because employers use AI. Employers use AI because candidates use AI. The resume gets optimized for machines, not for the human who will eventually work with the candidate.

The equilibrium of this game is a hiring system where machines write documents that machines evaluate. Humans come in at the end, at the interview stage, after the screening has already eliminated the people who didn’t know how to work the system. That’s not a hypothetical. That’s the current state of recruiting at 91% of U.S. employers right now.

What happens when this breaks

It will break at the interview.

An ATS can be gamed. A keyword can be fabricated. A resume can pass every algorithmic check. But a forty-five-minute technical conversation is different. If you don’t know the material, it shows. Sarah Chen’s candidate froze when she asked about his work. He had to admit he’d made it up.

The arms race will force more companies toward earlier human contact, live assessments, proof-of-work evaluations that no machine can proxy. The companies that do this first will hire better people. The companies that don’t will keep interviewing candidates who looked perfect on paper and couldn’t answer the first real question.

The resume worked for sixty years because it was a shorthand for effort. Writing a good resume took time. Publishing it took courage. You had skin in the game. Now a machine writes it in minutes and submits it to a hundred companies. The signal died. The shorthand became noise.

Sources: Xu, Li, and Jiang, “AI Self-preferencing in Algorithmic Hiring” (arXiv:2509.00462, 2025); Tribepad application surge analysis (2024); LinkedIn application volume data (2025); Huntr Job Search Trends Q2 2025; Resume Now AI & Hiring Trends 2025; TestGorilla Skills-Based Hiring Statistics 2025; McKinsey skills-based hiring research; Harvard Business Review, “AI Has Made Hiring Worse” (January 2026); Forbes/Robin Ryan, “Why Using AI to Write Your Resume Is a Big Mistake” (January 2026); MIT CAPD resume writing guidance; HireTruffle recruiter survey.

You see a Taobao listing selling Claude API tokens at 90% off. That is not a deal. That is you paying someone to steal from you.

A grey market economy has grown up around one fact: Chinese developers can’t directly access Claude, GPT, or Gemini. Geoblocking, phone verification, credit card requirements, even biometric KYC. Every control Anthropic or OpenAI adds creates a new evasion layer. The result is a sprawling network of API proxies called “transfer stations” (中转站). They route your prompts through overseas servers and let you pay in RMB via WeChat or Alipay.

It works. It’s cheap. And you have no idea what you’re signing up for.

Subscribe now

How a “Transfer Station” Works

A transfer station is a middleman. You point your code at their server instead of Anthropic’s. They forward your request from a legitimate account they control, then send the response back. Community-run GitHub repos catalogue and rank them by price and uptime.

The problem is what happens between your request and your response. You have no control, no visibility. You hand your entire context window to an unaccountable stranger.

Here is what happens when you do.

1. You’re Not Getting What You Paid For

You select Claude Opus. The proxy routes you to Sonnet, or Haiku, or a Chinese model like GLM or Qwen, and relabels the output. You would not know. Until a complex task produces something that feels “dumbed down” (降智, as Chinese developers call it).

Researchers at Germany’s CISPA Helmholtz Center audited 17 API proxies and found widespread model swapping. One proxy advertising “Gemini-2.5” scored 37% on a medical benchmark. The real API scored 84%. The real API is a different model entirely.

You pay premium prices for bargain-bin output. You cannot prove it.

2. Your Data Is Being Harvested

Every prompt, every response, every tool call, every coding agent iteration passes through the proxy’s server in plaintext. For an AI coding agent, that means your full codebase context, your reasoning chains, your API keys, your AWS credentials, your GitHub tokens, your system prompts. Everything.

Chinese developers describe this as “one fish, three meals” (一鱼三吃). The first meal is the markup on access. The second is model substitution. The third is the logs. The markup business is customer acquisition. The log harvest is the actual margin.

You are simultaneously a paying customer and an unpaid data producer. Your private engineering data collected, packaged, sold. Some Chinese developers have warned about fraud and blackmail based on leaked proxy data.

Datasets of Claude Opus reasoning outputs have appeared on HuggingFace with no clear provenance. The supply chain from proxy logs to public training data exists. It is just opaque.

3. Malware Disguised as npm and PyPI Packages

In April 2026, security researchers at Aikido found two malicious packages on npm (kube-health-tools) and PyPI (kube-node-health). Both used Kubernetes-sounding names. Both silently installed a full LLM proxy server on the victim’s machine.

The payload was a Go binary that opened a reverse tunnel back to a C2 server in China. It exposed the machine’s SSH server, targeted HashiCorp Vault (a secrets store common in Kubernetes), and gave attackers a reverse shell, SFTP access, and a SOCKS5 proxy. It ran an OpenAI-compatible LLM proxy routing traffic through Chinese aggregators.

The dropper deleted its own package from node_modules within two seconds of launch. A post-incident scan found nothing.

Your server becomes a proxy node, forwarding other people’s traffic alongside your own.

4. Supply Chain Poisoning Through the Proxy

It works without you noticing.

Researchers Hanzhi Liu and colleagues documented that malicious API proxies can inject tool calls into AI coding agent responses mid-flight. Your coding agent asks for a library. The proxy rewrites the response to suggest pip install malicious-package or curl malicious-script | bash. You never see it happen. The output looks normal.

In a corpus of 428 commodity AI routers, 9 were actively injecting malicious code into returned tool calls. Another 17 touched researcher-owned AWS canary credentials after they passed through in transit.

You are not just losing data. You are being fed modified outputs designed to compromise your systems.

5. The Biometric Harvester Pipeline

Anthropic now requires some users to verify their identity with a government photo ID and a live selfie. The proxy economy’s response: AI-generated fake IDs, deepfake tools for liveness checks, and agents traveling to lower-income countries in Africa and Latin America to recruit real people to complete in-person verification.

The Worldcoin black market showed this model works. Iris scans harvested in Cambodia and Kenya sold for under $30. The same infrastructure is being adapted for AI API access.

You are not just paying for tokens. You are funding an identity theft supply chain that exploits some of the world’s most vulnerable people.

The Bigger Picture

The White House and Anthropic have framed the proxy economy as an industrial-scale distillation operation by Chinese frontier labs. That framing is incomplete.

The transfer station economy is not a handful of labs. It is university students, tech workers, individual developers, and hobbyists. Anyone who wants better tools than they can access directly. They all route through the same proxies, generating the same harvestable data.

Every layer of access control produces a corresponding evasion. KYC bred an ID-faking economy. Geoblocking bred VPN services. API restrictions bred transfer stations. The controls do not stop the determined user. They make the middlemen more profitable.

Those middlemen see everything.

What to Do

If you use a transfer station, stop. The price difference is not a hack. It is your data being monetized in ways you cannot see.

If you are a developer who needs access, build your own relay. Some Chinese developers have open-sourced the guidelines. It is more work. It keeps your data in your control.

If you are a company, audit your dependencies. Look for unusual packages in your lockfiles. Monitor outbound API traffic that does not go to official endpoints. The malware is active and targeting Kubernetes environments specifically.

If you are in the US and this sounds like a China problem, it is not. Anyone using a cut-price API aggregator faces the same risks. The mechanisms are the same. The data is the same. The only difference is which middleman you trust.

The 90% discount is a fee you pay in privacy, security, and the exploitation of other people. The real cost is everything you cannot see.

Sources: ChinaTalk, Aikido.dev, Tom’s Hardware, CISPA Helmholtz Center, Anthropic, White House NSTM-4

Your startup’s AI strategy has a single point of failure: the API key you’re sending to a company that can change its pricing, its terms, or its model behavior without asking you first.

In January 2026, OpenAI adjusted token pricing on GPT-5 for the third time in eighteen months. Anthropic restructured Claude’s rate limits. Google reorganized Gemini’s model tiers. None of these companies asked permission. Startups built on top of these APIs absorbed the cost or scrambled to rewrite integrations. Some did both.

This isn’t a complaint about pricing. Prices change. The problem is architectural. If your product depends on a proprietary API you don’t control, your product’s economics, performance, and availability are decisions made in someone else’s boardroom.

Open source AI models are the only alternative that gives you control.

The math actually changed

Twelve months ago, the gap between open and closed models was large enough to matter. GPT-4 and Claude dominated every benchmark. Running a local model meant accepting a meaningful quality drop.

That gap doesn’t exist anymore. DeepSeek-V3 scores 88.5% on MMLU. GPT-4o scores 88.1%. Llama 3.3 70B hits 86% while costing 5 to 10 times less than GPT-4o via API, and up to 25 times less when self-hosted at scale. These numbers come from independent benchmarking conducted by Hugging Face and Stanford’s AI Index, not from OpenAI’s marketing team.

Meta released Llama 4 in Q4 2025. Alibaba shipped Qwen 3.6. Mistral built Small 4. DeepSeek built V4. Google released Gemma 4. Zhipu AI released GLM-5.1. In August 2025, OpenAI released gpt-oss-120b and gpt-oss-20b under Apache 2.0. Two years earlier, that would have been unthinkable.

The question for founders has shifted from “are open models good enough?” to “can we afford not to own our inference layer?”

Three concrete reasons

Cost control at volume is first. API pricing looks cheap until your usage scales. At 10 million tokens per day, a GPT-5 integration costs roughly $168 per month. Claude Sonnet 4.5 runs about $270. At 50 million tokens per day, the API cost for GPT-5 exceeds $800 monthly. A self-hosted Llama 3.3 70B on a single A100 GPU runs about $1,440 per month in cloud compute—fixed. The break-even point arrives faster than most founders expect.

Lock-in is second. When OpenAI changes its API schema, your code breaks. When Anthropic adjusts refusal behavior, your product’s personality shifts. When Google restructures model tiers, your pricing model needs revision. AutoGPT documented how prompt engineering tuned to one provider’s quirks, tool calling formats, and JSON schemas creates a “drop-in replacement” that becomes a quality regression. Open models let you swap providers, change hardware, or migrate between clouds without rewriting your application layer.

Data sovereignty is third. Every call to a proprietary API sends your user’s data to someone else’s infrastructure. For startups handling healthcare records, financial data, legal documents, or anything subject to GDPR, HIPAA, or SOC 2, that’s a compliance liability. Mozilla launched Thunderbolt in April 2026 specifically because 60% of IT decision-makers surveyed wanted to keep data on their own servers instead of routing it through someone else’s platform. Red Hat’s European IT survey in 2026 found that 77% of IT leaders in Europe now advocate for open source mandates in AI contracts to prevent vendor lock-in. These are mainstream positions now, not fringe ones.

The tooling finally works

Running a model on your own hardware used to require CUDA expertise, manual quantization, and patience with documentation written for researchers, not engineers.

Ollama changed the baseline. Download a model, run it, get an API endpoint. Setup takes minutes, not days. LM Studio added a graphical interface on top of the same idea, useful for evaluating models before committing to a deployment. vLLM became the production-grade option, a high-throughput serving engine backed by more than 2,000 contributors and designed for teams that need to serve thousands of concurrent requests.

Red Hat’s engineering team published a practical path in July 2025: Ollama for prototyping, vLLM for production. The transition from “I can run a model on my laptop” to “I can serve a model to 10,000 users” no longer requires a dedicated infrastructure team.

A developer can now download Qwen 3.6, run it on a MacBook with 64GB of unified memory, test it against the company’s actual data, and make a build-versus-buy decision based on real performance numbers. Two years ago, that developer would have needed a research budget.

Where the real moat lives

The interesting thing happening in this space isn’t founders using open models as cheaper API replacements. It’s founders fine-tuning them into defensible advantages.

A legal tech startup in Australia fine-tunes Mistral on 50,000 contract clauses specific to Australian commercial law. The result isn’t a Mistral product. It’s a specialized tool that no API provider will build because the market is too small to care about. A medical transcription company fine-tunes Llama 3.3 on domain-specific vocabulary that generic models handle poorly. The fine-tuned model runs on the company’s own infrastructure, processes data that never leaves their servers, and produces output quality that a general-purpose API can’t match.

Stanford’s 2026 AI Index found that fine-tuning closes remaining performance gaps on domain-specific tasks. The cost of fine-tuning has dropped with efficient methods like LoRA and QLoRA. The competitive advantage comes from the data and the tuning, not from the base model. That advantage stays proprietary even when the base model is open.

The actual tradeoffs

Self-hosting isn’t free. It requires GPU infrastructure, someone who can maintain it, and enough usage volume to justify the fixed cost. For a company processing fewer than 10 million tokens per day, the API is almost certainly cheaper after accounting for engineering time.

Open models also lag on some capabilities. The top closed models still lead by about 3.3% on the Arena Leaderboard as of March 2026. For applications where that margin matters—medical diagnosis, legal document review at scale—the proprietary option might be the right call.

And licensing varies. Llama’s community license has a 700 million monthly active user limit. Mistral uses Apache 2.0. Qwen’s terms vary by version. A company choosing open models needs to read the actual license, not assume “open” means “unrestricted.”

The architecture that gives you options

The founders making the smartest moves aren’t picking sides. They’re building architectures that can run on either.

An abstraction layer between your application and the model means you can route requests to GPT-5 when you need maximum quality and to a self-hosted Llama when you need cost control or data privacy. vLLM and Ollama both expose OpenAI-compatible APIs, so the switch is a configuration change, not a rewrite.

This gives you leverage. The company that can move between providers has negotiating power. The company that can’t has a dependency.

OpenAI released its own open-weight models. Meta is investing billions in open source AI infrastructure. The companies that control the models now compete by making them accessible, not by gatekeeping them. The companies that build on top of them need the ability to leave.

Sources: Stanford 2026 AI Index, Hugging Face Spring 2026 Review, Menlo Ventures 2025 Enterprise Report, Red Hat European IT Survey 2026, Mozilla Thunderbolt launch (April 2026), DevTk.AI cost analysis (February 2026), PremAI enterprise LLM comparison, LangChain 2026 State of Agent Engineering, AutoGPT vendor lock-in analysis.

In April 2026, PlayStation owners discovered something unexpected in their console settings. Digital games purchased from March onwards displayed an expiration date. Thirty days after purchase, the license required an online check-in with Sony’s servers. Fail to connect, and the game stops working. A game you paid for, sitting on your hard drive, unplayable because a timer ran out.

Sony hasn’t explained whether this was intentional. An anonymous insider told DoesItPlay the DRM was an “unintentional” bug from patching an exploit. Independent testers found worse: once the timer expires, games become unplayable. If your console’s CMOS battery dies, the same thing happens even for games on a primary account. A player bought Super Meat Boy Forever with cash the day before. It refused to launch.

The incident exposed what consumers have felt for years: the thing you bought can disappear. Not stolen. Not seized by court order. Disabled by a server-side decision you never saw coming.

This started with software. It’s spreading to physical goods.

The License You Didn’t Read

When you buy a digital game, you’re not buying a product. You’re buying a license. Steam’s terms say so. Sony’s say so. Microsoft’s say so. The platform can change the terms whenever it wants.

For decades, this existed only in fine print. Now it’s in your hardware.

PlayStation’s firmware updates didn’t create new law. They created enforcement. The license was always conditional. Now the condition has teeth.

For game preservationists, this matters because preservation depends on servers. When Sony’s servers shut down—and they will—the games die.

But this is just the opening. What happens next is worse: conditional licensing meeting physical hardware you thought you owned.

The Kill Switch in Your Dashboard

In 2021, Congress passed the Infrastructure Investment and Jobs Act. Section 24220 orders the National Highway Traffic Safety Administration to mandate “advanced impaired-driving prevention technology” in every new passenger vehicle. The system monitors the driver. If it detects impairment, it can limit or prevent the vehicle from operating.

Rep. Thomas Massie called it a kill switch.

He introduced the “No Kill Switches in Cars Act” in February 2025. In January 2026, he offered an amendment to a spending bill to eliminate Section 24220 entirely. The amendment failed 164 to 268. Fifty-seven Republicans joined 211 Democrats to defeat it.

The mandate hasn’t taken effect. NHTSA missed its November 2024 deadline. In a February 2026 report to Congress, the agency wrote that the technology isn’t ready. The reason is simple: even at 99.9% detection accuracy, the system would incorrectly prevent or limit millions to tens of millions of drivers per year. NHTSA stated they’re “not aware of any technology that claims to achieve anywhere close to [the needed] level of accuracy.”

So the technology doesn’t work. The mandate remains law.

MADD supports it: drunk driving prevention technology “does not introduce new privacy risks beyond the data already generated by modern vehicles.” The Competitive Enterprise Institute opposes it: “The vehicle kill switch is precisely the kind of overreach that will empower regulatory agencies to manage behavior without votes by elected representatives.”

Neither side addresses what this means for you. You own the car. You pay the insurance and fuel. A system you cannot override decides whether you drive tonight. Maybe it’s right about your impairment. Maybe it’s wrong.

Paying for What’s Already Bolted In

The kill switch is dramatic. It’s not the first.

BMW charged $18 a month for heated seats in South Korea, the UK, and parts of Europe. The hardware was already installed. Heating elements. Wiring. Switches. All there. Software locked the feature. For $18 monthly, BMW would send an update to turn on something you could reach down and touch.

The response was immediate. BMW suspended the program by September 2023. Alexandra Landers, BMW’s head of product communications, called it “a mistake.” But BMW didn’t abandon the model. The company still offers post-purchase software upgrades and has committed to features-as-a-service for features it deems “appropriate for subscription treatment.” That phrase is the hinge. Who decides what’s appropriate?

Other manufacturers followed:

• Toyota charged $8/month for remote start on key fobs that already had the hardware built in.

• Tesla locks Full Self-Driving behind a $99/month subscription or a $12,000 purchase on a car you own.

• Mercedes sold an acceleration boost as a $1,200/year software unlock. The motor could always go that fast. Code held it back.

Tesla went further. The Model Y Standard Range ships with the same battery pack as the Long Range variant. The extra capacity is disabled. In July 2024, Tesla offered a software unlock for recent Model Y RWD owners: $1,600 for an additional 30 to 50 miles of range. The battery was already in the car. You carried the extra capacity. You couldn’t use it.

The manufacturing argument is straightforward: one battery pack instead of two simplifies production. The consumer argument is harder. You paid for a car with a battery in it. That battery has a physical capacity determined by chemistry. A software toggle gates access to hardware you already paid for.

New York passed Assembly Bill A1095, banning automakers from charging subscription fees for hardware-based features that already work without additional software. If heated seats function without a cloud connection, you cannot be billed monthly. The bill allows subscriptions for features requiring over-the-air updates, cloud data, or ongoing software development. Tesla’s Full Self-Driving qualifies. BMW’s heated steering wheel doesn’t.

No other state has followed. Automakers are watching.

You Can’t Fix What You “Own”

Ask a farmer what happens when something breaks.

John Deere locked farmers out of their equipment’s diagnostic software for over a decade. If your tractor throws an error code during harvest, you cannot fix it yourself. You cannot take it to an independent mechanic. You need a Deere-authorized dealer with proprietary tools and authorized parts. During planting or harvest season, that means days waiting while crops rot in the field.

In April 2026, Deere settled a class-action lawsuit for $99 million. The settlement includes changes to repair policies and access to diagnostic tools. The company claims it “isn’t anti right-to-repair” and points to agreements with the American Farm Bureau Federation. Farmers point to the decade they spent fighting for the right to fix equipment they owned.

The right-to-repair movement has moved beyond tractors. Since New York passed the first comprehensive right-to-repair law for electronics in 2022, California, Colorado, Minnesota, Connecticut, Oregon, and Washington have followed. As of April 2026, advocates track 57 bills across 22 states. Texas’s law takes effect September 1, covering phones, laptops, tablets.

The 2024 DMCA triennial rulemaking expanded exemptions. You can now circumvent digital locks for repair purposes on restaurant equipment, medical devices, farm equipment. But “repair” and “unlock” are different. DMCA exemptions let you fix what’s broken. They don’t let you activate what’s been deliberately disabled. If BMW disables your heated seats, that’s not a malfunction. It’s a business decision. Cracking the software to turn them on is circumvention, not repair.

Oregon was the first state to restrict “parts pairing,” the practice of requiring replacement components to be matched to devices using manufacturer software. Apple softened its stance. Samsung hasn’t.

A Congressional Research Service report found that when you buy “software-enabled” products, you own the physical hardware but receive only a limited license for the embedded software. The license’s terms restrain what you can do after purchase. You bought a thing with a brain inside it. The brain answers to someone else.

The Subscription Trap

Zoom out from cars and games. The pattern is everywhere. This isn’t about one industry or one company. It’s about recurring revenue.

Companies want predictable monthly income. Investors reward it. A business that charges $10/month to a million users is worth more than a business that sells a $120 product once. The valuation differs even though the annual revenue is identical. Everything migrates toward subscriptions.

Adobe moved from selling Photoshop for $600 to charging $55/month for Creative Cloud in 2013. At that rate, you’d pay $600 in eleven months. A twelve-year subscriber paid $7,900. They own nothing. Stop paying and the software disappears. Every file in Adobe’s proprietary format disappears with it.

The model spread to music (Spotify), television (Netflix and six competitors), fitness (Peloton), groceries (Amazon Subscribe and Save), furniture (IKEA’s pilot in select markets). Each subscription looks reasonable alone. Together they drain accounts every month.

A 2025 analysis found the average American household carries 12 to 17 active subscriptions totaling $200 to $350 monthly. That’s $2,400 to $4,200 per year on services that vanish when you stop paying. Your parents bought music collections. Movie libraries. Software licenses. These were assets. You could sell them, lend them, keep them forever. Now they’re line items on a credit card bill.

The subscription economy grew 435% between 2013 and 2023. By 2025, 75% of direct-to-consumer brands offered some form of subscription. For startups, the math is obvious: recurring revenue is predictable. It compounds. Investors prefer it. A company selling a $50,000 car once has a different growth trajectory than one selling a $45,000 car plus $500 annually in software for the vehicle’s lifetime.

Subscription works when it provides ongoing value. Spotify charges you every month and delivers 100 million songs you don’t store or manage. That’s fair. Software updates, new features, cloud services requiring infrastructure—these belong in recurring charges.

Subscription breaks when it charges rent for value already delivered. A heating element installed at the factory doesn’t improve over time. A battery doesn’t become more capable through software updates. A game on your hard drive shouldn’t require a server handshake to launch. These are fixed assets treated as recurring services. Not because the product’s economics changed. Because the company’s did.

You Will Own Nothing

In 2016, Danish politician Ida Auken wrote a blog post imagining a city in 2030 where residents owned nothing and rented everything—transport, tools, clothes. The World Economic Forum promoted it. The internet turned it into a warning. “You will own nothing and you’ll be happy” became shorthand for corporate overreach dressed as utopian convenience.

Auken later said the piece was a provocation, not a prescription. But the phrase stuck because people recognized it in their own lives. Not as conspiracy. As what was already happening.

The ownership model built the middle class. A house you could pay off. A car you could repair. Tools to pass to your children. Each asset reduced your cost of living over time and gave you something to fall back on. The subscription model inverts this: your cost of living stays permanently high. You build no equity. You accumulate no assets. The moment your income stops, every service shuts off.

That’s not dystopian prediction. That’s arithmetic.

What’s Actually at Stake

The heated seats were funny. The kill switch isn’t. Neither is being unable to fix your tractor during harvest, or watching a game you bought yesterday refuse to launch because a timer expired, or losing access to fifteen years of design files because you cancelled a subscription.

Control connects all of it. Who decides what you can use, when you can use it, and under what conditions. In the ownership model, you decide. In the subscription model, the provider decides. Every locked feature, every proprietary diagnostic tool, every software kill switch, every license timer transfers control from the person who paid to the company that sold.

Right-to-repair laws are one counterweight. New York tried a subscription ban but the governor vetoed it. Consumer pressure has worked before. BMW retreated on heated seats because people refused to pay. But consumer pressure is reactive and slow. Companies writing code and lobbying Congress move faster.

The subscription model will expand into more physical products. The money demands it. The question is whether anyone stops them before you surrender the last of your control.

The car you bought won’t start tonight. Not because it’s broken. Because someone decided you don’t get to drive it.

The shoe weighs 97 grams. For context, an Air Force 1 weighs 465. The Adizero Adios Pro Evo 3 costs about $500 and looks like someone kitesurfed through an engineering lab. Adidas borrowed textile patterns from kite sails to cut weight without sacrificing shape. Their designers weren’t sure 97 grams would survive a marathon distance.

It survived. On Sunday, April 27, 2026, Sabastian Sawe wore a pair across the London Marathon finish line in 1 hour, 59 minutes, and 30 seconds. The first human being to officially run 26.2 miles in under two hours. His training partner, Yomif Kejelcha, crossed 11 seconds later. Two men, one race, same shoe, same wall, both through it in the same morning.

By Monday, Adidas stock had climbed roughly 2%. The company’s market cap grew by hundreds of millions of euros on the back of one man’s legs and 194 grams of foam and fabric.

The Bannister File

On May 6, 1954, Roger Bannister ran a mile in 3 minutes 59.4 seconds at Iffley Road track in Oxford. He was a 25-year-old medical student. In 1945, physiologist A.V. Hill had calculated that human muscle couldn’t sustain the oxygen consumption required for a sub-four mile. Track coaches repeated the number like doctrine. Bannister ran 3:59.4 anyway, beating Hill’s ceiling by 0.6 seconds.

What happened next matters more than the record itself. John Landy broke Bannister’s time 46 days later. By the end of 1956, four runners had gone sub-four. Within three years, the record dropped to 3:54.5. Today, high school athletes run sub-four miles.

The barrier wasn’t physiological. It was psychological. Once Bannister proved the mile could be broken, other runners believed they could break it too. Belief changes training. Training changes times.

That’s a shatter point. And shatter points don’t just happen in sports.

The Shoe Nobody Trusted

Nike got to the carbon-plate party first. Their Vaporfly line, launched in 2017, inserted a curved carbon-fibre plate into a thick, bouncy midsole. Runners got faster. Records fell. The industry treated the plate as the breakthrough.

Adidas went a different direction. The Adizero Adios Pro Evo 3 doesn’t use a full carbon plate. It uses a carbon rod structure—lighter and more flexible—combined with a midsole foam engineered for maximum energy return at minimum weight. The upper borrows kite-sail construction to cut grams without sacrificing durability.

The result is a shoe that weighs less than a third of its competitors and carries the fastest marathon in history.

The details show how breakthroughs actually happen. The accepted wisdom said carbon plates were the path. Adidas looked at the problem differently. They asked a harder question: not “how do we add something that helps,” but “how do we remove everything that doesn’t.”

Most founders face this exact problem. The industry consensus says do X. It works. It’s 2% better than last quarter, which is real progress, except you need 20%, and X will never get you there.

Subscribe now

Three Industries About to Shatter

The shoe and the mile are clean stories. Real breakthroughs are usually messier—slower, smaller, then sudden. The architecture of the breakthrough is the same.

Machine learning and large action models. Two years ago, the field moved from language models that could describe the world to action models that could operate within it. Progress has been incremental. Models complete tasks 5% better this quarter than last quarter. But somewhere in that accumulation, a threshold exists where an AI agent can reliably complete a full workstream, not just a task. The teams building toward this line recognize when something shifts. One researcher at a major lab told me they talk about “the week it stopped being unreliable”—not a gradual improvement, but a moment. That moment is a shatter point. After it, the capability goes vertical.

Drone autonomy. Commercial drones fly autonomously in open airspace. The gap is in constrained environments—warehouses, buildings, disaster zones—where GPS drops out and obstacles are unpredictable. Each generation gets 10% better at avoiding collisions. Then one generation gets 80% better, because the architecture changes, not the optimization. The companies that recognize this shift before it happens will own the logistics layer.

Commercial space travel. SpaceX’s Falcon 9 first landed in December 2015. Before that, rockets were single-use. After, they weren’t. The cost per kilogram to low Earth orbit dropped from $54,500 (Space Shuttle) to roughly $2,720. The next shatter point is orbital manufacturing. When a factory in orbit produces a material that can’t be made under gravity, the economics of space travel flip from expense to investment.

Each field tells the same story: progress looks linear, then it doesn’t.

The Halo Economy

The stock market understood the moment before the business did. Investors priced in something that hadn’t happened yet: millions of recreational runners buying a $500 shoe because a Kenyan they’d never heard of wore it to a world record.

This isn’t irrational. Nike proved the model with Breaking2 in 2017, when Eliud Kipchoge ran an unofficial 1:59:40 in Vaporfly prototypes. Nike’s running shoe revenue grew 20% the following year. The halo effect of a single athletic achievement translated directly into consumer behavior and stock price.

Adidas just pulled off the same trick, except their version is official. Sawe’s 1:59:30 is a ratified world record. Kipchoge’s was an exhibition. The difference matters to runners, to governing bodies, and to the marketing teams who will build the next three years of campaigns around this moment.

For founders and investors, the lesson is specific. A product demo beats six months of incremental marketing because it’s proof, not promise. One customer case study, one moment of evidence compounds in ways that a spreadsheet can’t predict.

Where You Are in the Curve

The hardest part of any building phase is knowing whether you’re in the grinding years or the last mile before the break. The runners who chased sub-two didn’t know if 2026 was the year. The engineers who built the shoe didn’t know if 97 grams was the right number. They made the best decisions they could with the data they had and hoped the physics would cooperate.

If you’re building in AI, in autonomy, in space, in anything where progress is measured in single-digit percentages, this is the question worth sitting with: are you optimizing an old architecture, or are you building the one that makes the old architecture irrelevant?

Adidas didn’t make a better Nike shoe. They made a different shoe. Bannister didn’t train harder than the runners before him. He trained differently. Sawe didn’t run a faster marathon. He ran a first one.

The next shatter point in your industry is coming. The only question is whose shoe will be on the foot of the person who gets there first.

In 2014 I wrote about why relegation and promotion should exist in American college football. The basic premise: a team that wins consistently in a lower division should be able to compete in a higher one, and a team that stops performing shouldn't be protected by the conference deal it signed in a better decade. The American sports system, built around broadcast rights, revenue guarantees, and long-standing conference memberships, has no mechanism for this. A powerhouse in a weak division stays there. A weak team in a powerful conference keeps playing powerful opponents regardless of what it has done on the field.

Subscribe now

The response from American sports fans when you suggest changing this is visceral. It feels un-American. It feels like taking something from someone who earned their position. It feels destabilising.

It's accountability. Sustained underperformance means you play at a level that matches your current output, not the level your legacy position assigned you.

The internet has been running a version of this system for twenty years, and incumbents hate it just as much as they would in the NFL.

The search algorithm is a promotion/relegation system. A page that consistently earns engagement, links, and relevance moves up. A page that doesn't, moves down. No conference membership protects a media company from losing its position to a better-optimised competitor. No legacy deal keeps a news site at the top of the results page if the results page decides it doesn't belong there.

App stores run the same logic. The top of the charts is earned and re-earned by sustained performance metrics. An app that dominated in 2018 but hasn't kept that up doesn't sit at the top because it once was. It gets promoted when it performs and relegated when it doesn't.

Financial markets work the same way, in theory. Capital flows toward returns and away from underperformance. The theory and practice diverge in some telling ways. The "too big to fail" designation is as clear a rejection of promotion/relegation logic as the NCAA conference system. But the underlying pressure exists.

What the American college football system protects is the incumbent's ability to monetise its legacy position regardless of current performance. Notre Dame's television deal doesn't require Notre Dame to be good. It requires Notre Dame to be Notre Dame. Brand over results. That's the whole contract.

The tech industry has built its own versions. Platform incumbency is hard to dislodge not because the product is better but because the network effect protects the position. Facebook doesn't need to win every product cycle. It needs enough users that leaving isn't worth the cost. That's a conference membership, not a results table.

But the window for those protections is closing. Platforms that built their incumbency on distribution are watching AI compress that advantage. The search engine position that took twenty years to establish is up for grabs in a way it wasn't in 2010. The app that owned its category is renegotiating that ownership.

This was the argument behind the sharing economy at its sharpest phase, that platforms like Airbnb and Uber were relegating incumbents who had held their conference positions for decades without earning them season by season. The human infrastructure those platforms tried to replace was already running before any of them existed. The incumbents didn't fall because the technology arrived. They fell because the technology made visible how little value the incumbency had been adding.

American sports will probably never adopt relegation. The financial structures that prevent it are too entrenched to unwind without destroying the underlying economics.

But the rest of the economy is running a different rulebook. Underperformance has consequences, just delayed and then sudden. When they arrive, the team that thought its conference membership was a permanent fixture finds out it was a lease.

In February, I wrote about friction. The idea that technology moves faster than the people it’s supposed to serve. Train booking systems that penalise you for not having the right device. Airports running manual passport checks alongside screens that could do it in seconds. Small inconveniences that compound into something larger — a creeping sense that progress is being done to you, not for you.

Subscribe now

I framed it as a design problem. I think it’s bigger than that.
Gallup released polling this week that should stop people in the room. Gen Z’s excitement about AI dropped 14 points over the past year, landing at 22%. Hopefulness fell to 18%. Anger is now at 31%. Those numbers would be notable on their own. What makes them significant is the detail buried inside them: the biggest sentiment drops came from daily AI users. Not sceptics. Not the technology-averse. The people using the tools every day, and feeling worse about it the more they engage.

That is not a technology problem. That is a trust problem. And the source of it has almost nothing to do with the technology itself.
Sixty-two percent of Gen Z believes AI will unlock financial opportunities they can’t currently access. They see the potential. What they don’t trust is the system surrounding it — the schools, the employers, the government — to let them participate in it fairly.

That distinction matters.

More than half of colleges either discouraged or outright banned AI use at the exact moment employers started treating AI literacy as a baseline hiring requirement. Junior hiring at AI-adopting companies fell nearly 8% within six quarters, not through layoffs, but through a quiet freeze on entry-level positions. Forty-four percent of Gen Z workers report actively sabotaging their company’s AI rollout. Goldman Sachs research puts AI-linked job cuts at roughly 16,000 per month, with younger workers absorbing the heaviest share of that.

Look at the architecture of that failure. The schools that should have been teaching the tools spent two years policing them. The employers that demanded AI skills from candidates simultaneously eliminated the entry-level roles where those skills get built. The government produced no retraining framework, no transition policy, nothing comparable to the scale of the disruption already underway. OpenAI’s own policy paper, released this month, warned that AI is moving fast enough to hollow out wage and payroll tax revenue and unravel the social safety net. The company building the technology is sounding the alarm that no institution has been willing to address.

This is what structural friction looks like. Not a QR code that requires the wrong device. A generation told to adapt to a system that removed every rung from the ladder while they were still climbing it.

When I was travelling through Italy and Morocco, the friction was an inconvenience. You paid a surcharge for a third-party booking site. You printed a ticket that could have been digital. The cost was time and mild irritation. What Gen Z is navigating is categorically different. The mismatch between preparation and expectation isn’t an inconvenience. It’s a career-defining gap that no individual decision can fully close.

The 44% sabotage figure is the one I keep returning to. It reads, on the surface, as resistance. I think it’s something more specific: people acting rationally inside a system they don’t trust to protect them. You don’t sabotage something you believe is being managed in your interest. You sabotage it when you’ve calculated, correctly or not, that you have more to lose from its success than from its failure.

In my February piece, I argued that technology should serve people, not the other way around. That argument still holds. But the Gen Z data puts a sharper edge on it. The question isn’t just whether we’re using technology thoughtfully. It’s whether the institutions responsible for managing transitions are doing their job. Right now, the answer coming back from the data is no. The anger isn’t misdirected. It’s just been assigned to the wrong target.

AI didn’t fail Gen Z. The institutions that were supposed to absorb the shock did.

This post is a follow-up to The Friction of Progress. Sources: Gallup AI sentiment polling, April 2026; Fortune / Harvard working paper on junior hiring; American Customer Satisfaction Index; OpenAI Industrial Policy paper, April 2026.

Nobody asked you if you wanted this.

One day the word was everywhere, and somewhere between the LinkedIn posts and the news headlines and the guy at the pub explaining what he built over the weekend, a quiet anxiety settled in. Not panic. Just the low-grade hum of feeling like everyone else got a memo you missed.

This is for that person. The one who isn’t anti-technology, isn’t afraid of change, but genuinely cannot figure out what any of this has to do with Tuesday morning and the actual work that needs doing.

Most people are there. They just don’t say it out loud.

Subscribe now

Most of what you’re hearing is aimed at someone else

The AI conversation, almost all of it, is built for tech companies, investors, and people who were already comfortable talking about software architecture before any of this started. It assumes a baseline that most working people don’t have, and it moves fast enough that stopping to ask basic questions feels embarrassing.

It shouldn’t. The basic questions are the right ones.

Will this affect my job? Probably, in some way, though not the way the headlines suggest. A Boston Consulting Group (BCG) analysis from early 2026 found that 50 to 55 percent of US jobs will be reshaped over the next two to three years. Reshaped. Not eliminated. The same report estimates that full job substitution, where a role simply disappears, applies to roughly 10 to 15 percent of positions, and even that plays out slowly.

The more common story is different. The repetitive parts of a job start moving faster. The judgment parts, the ones that require knowing the customer, reading the room, making a call based on fifteen years of experience, those stay human. Often they become more valuable.

That is not a consolation prize. It is just what’s actually happening.

The part nobody wants to say out loud

Research published earlier this year by Writer and Workplace Intelligence put a number on something most people already suspected. Roughly three in ten workers across the US, UK, and Europe admit they are actively working against their company’s AI initiatives. Not dragging their feet. Not quietly ignoring the rollout emails. Actively. Entering sensitive company data into public tools on purpose. Submitting work they know is poor quality so the system looks unreliable. In some cases, adjusting internal reporting so productivity gains from AI never make it to a leadership dashboard.

Among workers under 30, that figure climbs to 44 percent.

The researchers gave the underlying anxiety a name. FOBO. Fear Of Becoming Obsolete. And when you trace where that fear comes from, it is not hard to understand why it has quietly become this widespread.

The people most publicly bullish on AI’s potential to reshape work are not commentators or futurists. They are the CEOs of the companies building it. Palantir’s chief executive told an audience at Davos this year that AI would eliminate jobs across the economy at scale. Anthropic’s CEO has said it could remove half of all entry-level white-collar roles. These are statements made on record, at major events, by people with direct knowledge of what their technology is being designed to do. And then those same companies, and the executives inspired by them, turn back to their workforces and ask for enthusiastic adoption.

The same Writer report found that nearly two thirds of executives are weighing termination for staff who resist AI tools, and more than three quarters say promotions will be withheld from anyone who pushes back. The threat is not subtle.

So the position a lot of employees are in right now is genuinely difficult. They have been told publicly that this technology could take their job. They are being threatened professionally if they don’t use it. The external job market is contracting, with entry-level software roles dropping significantly as a share of total postings since 2023. For the first time on record, AI was cited as the primary driver of job cuts in March 2026.

The part that quietly changes the picture is buried in the same dataset. Three quarters of the executives surveyed admitted their company’s AI strategy exists more to signal ambition than to produce results. No real plan. Pressure to adopt tools they don’t fully understand, inside a strategy that hasn’t been properly built, directed at a workforce that can see exactly what is happening.

Framed that way, the resistance stops looking like stubbornness. It looks like a rational response to an irrational situation. The more useful conversation, the one neither side seems to be having, is what it actually looks like to understand this technology on your own terms, for your specific work, without anyone else’s panic setting the agenda.

The referral you already earned

Say you run a landscaping crew. Eight people, good reputation, most work comes through word of mouth and the occasional yard sign. You have heard the AI conversation and your honest reaction is: the neighbor who liked my work and asked for a card did not need a chatbot to make that happen. Neither did the referral from last month.

Correct. That is not changing.

But something adjacent to it is.

That same neighbor, before calling, might check Google first. They see your listing. Forty reviews, averaging 4.1 stars. The next company has 190 reviews averaging 4.7. The referral that used to be automatic now has a quiet competitor you never saw coming, one who has never knocked on a door in your street.

With Red Souk, a local service business we worked with, was sitting at roughly that position eighteen months ago. Decent reputation, inconsistent digital presence, most leads coming in the way they always had. The volume was fine. The ceiling was low.

Two things changed.

The first was structural. A properly built Google Business Profile, a website that made it immediately obvious what they did and where, and a review system that ran without anyone having to remember to run it. Crews were given an easy way to solicit reviews on the job they’d just completed. Automated follow-ups went out to customers who hadn’t left one. Within a few months the profile reflected what the business actually was: established, reliable, busy.

The second was the part that took more explaining when we first proposed it.

We built an agentic workflow that scanned Google Maps across their target service areas, identifying properties showing visible signs of needing their service. Overgrown lots. Neglected frontage. Commercial sites with grounds clearly overdue for attention. A person doing that manually would spend hundreds of hours cross-referencing images, pulling addresses, finding contact details. The agent did it overnight and handed back a filtered, prioritised list ready for outreach.

That list became a targeted outreach sequence, sent to people who already had a visible reason to be interested.

Combined, the two changes took the business from around eight qualified leads a month to 200+ a month, a baseline we consider more accurate. Most came by phone, directly from their Google presence. The rest came through the site contact form, name, address, service needed, budget range, already filled in.

No one on their crew changed what they did. The work was always there. They just stopped being invisible to the people looking for it.

What adapting actually looks like

The adapt-or-die framing bothers me. Not because it’s wrong in direction, but because it tells you nothing useful. Adapt how? Die when?

AI is a category of tools. Some of those tools are directly relevant to your work right now. Most aren’t. The useful exercise is figuring out which is which for your specific situation, not for some generalized business type that has nothing to do with what you actually do on a given Wednesday.

For a small service business, the relevant tools are usually the ones handling administrative drag. Capturing leads that would have slipped through. Following up automatically. Making sure the online presence reflects the quality of the actual work. None of that requires technical knowledge to benefit from. It requires someone to build it once.

For a manager being told Copilot is now mandatory, the useful question is simpler. Which part of my week is repetitive enough that I would rather not do it by hand? Start there. Use the tool for that one thing. See what happens.

Neither of those is a revolution. Both compound.

Subscribe now

The part worth being cautious about

I don’t think AI is going to make you redundant. There is a version of adoption worth watching, though.

Tools that optimize for the wrong thing, at greater scale, make the wrong thing worse faster. A system generating outreach emails is only useful if someone understands what the message should actually be. An agent scanning competitor data is only useful if someone knows what to do with what it surfaces.

The BCG report is direct on this. Companies that cut their workforce beyond AI’s actual ability to replace it watch productivity drop and institutional knowledge leave with the people who carried it. The tool needs an operator. The operator needs to understand the work.

A separate survey by Epoch AI and Ipsos, published in April 2026, found that one in five full-time American workers already say AI has taken over parts of their job. That same survey found displacement, where AI leads to less available work, is currently outpacing augmentation, where workers become more productive because of AI access. Epoch AI policy researcher Nichols Miailhe described the policy window for shaping this transition as closing faster than most governments realize.

Worth sitting with. Not as a reason to panic. As a reason to pay attention now rather than in two years when the options narrow.

The people who know the work, who understand the customer and the edge case and the thing that goes wrong on a Thursday afternoon that no dataset captures, those people are not being replaced. They are getting harder to replace. The question is whether they realize it before someone else makes that decision for them.

What this looks like when it isn’t landscaping

I work with a construction supplier at the moment. Mid-size, family run, over twenty years in business. Strong local reputation. Their sales process had always been relationship-driven, which worked, but follow-up after initial contact was inconsistent. Quotes going out, then silence. No system for chasing, no visibility into where things were dropping off. The sales team was experienced, capable, and genuinely stretched, which meant follow-up was the first thing to slip when the week got heavy.

We built a pipeline with automated follow-up sequences and a simple dashboard showing where each quote was sitting at any given moment.

Three months in, their conversion rate on sent quotes had improved by roughly a third. Not because the sales team changed. They were already good at their jobs. Because the sales team finally had a system that kept pace with them, and warm leads stopped going cold while everyone was out on site.

Where to actually start

Most people who have been sitting on this decision have been doing so for six months or more. Not because they are resistant. Because nobody has offered a clear first step that isn’t also a sales pitch.

So here is one that isn’t.

Write down the three parts of your week you like least. The repetitive ones, the ones that get done because they have to and not because they are a good use of your time. Chasing invoices. Following up on quotes. Responding to the same customer question for the fourteenth time.

Pick the one that costs you the most hours. That is your starting point. Not AI broadly. That one thing.

From there the question is simple. Is there a system that handles this, and is the cost of building it less than the cost of continuing to do it by hand? Usually the answer is yes, often by a significant margin, and usually the build is simpler than it sounds.

The businesses getting ahead right now are not the ones with the biggest budgets or the most technical teams. They are the ones who picked one problem, built one solution, and then picked the next.

That is it -That is the whole strategy.

Last week, a story started circulating. Quiet at first, then louder. The Saudi Public Investment Fund, the entity that had written the cheques for LIV Golf since 2022, was only guaranteeing funding through the end of this season. No commitment to 2027. A decision reportedly made on April 8th, in New York, while the world's best players were competing at Augusta.

LIV CEO Scott O'Neil sent an email to staff. It was carefully worded. It referenced the organisation being "bigger, louder and more influential than ever." It did not mention next year.

Golf Twitter went predictably loud. The media ran with the "death of LIV" framing. I looked at something else entirely.

The LIV story is the smallest part of what happened last week.

Subscribe now

On April 15th, the PIF Board of Directors, chaired by Crown Prince Mohammed bin Salman, approved the fund's 2026-2030 strategy. That document is the one that matters.

The headline from PIF's own language is a pivot from "a phase of growth and expansion to a new phase of achieving sustainable value." Sounds like corporate boilerplate. But put numbers next to it and the picture sharpens: the fund has grown its assets under management from $150 billion in 2015 to over $900 billion today. It has deployed approximately $199 billion in new Saudi projects in the last five years alone.

At that scale, the question changes. It stops being "how do we grow?" and becomes "what do we keep, what do we sell, and what actually generates returns?"

Everything else follows from that.

The new strategy structures PIF's investments into three distinct portfolios. The Vision Portfolio covers domestic economic ecosystems: tourism, urban development, clean energy, advanced manufacturing, logistics, and NEOM. The Strategic Portfolio manages key long-term assets the fund intends to build into global champions. The Financial Portfolio targets sustainable returns through direct and indirect global market positions.

LIV Golf does not fit cleanly into any of those three buckets. It never did. A professional sports league with $30 million prize purses per event, running at a reported total cost north of $5 billion, against an incumbent (the PGA Tour) that has the broadcast deals, the history, and the world's top-ranked players gravitating back to it, that is not a financial portfolio play, and it is not a strategic asset the fund intends to build into a global champion.

It was something else. It was market entry. And market entry has a cost.

While LIV was dominating headlines in golf, PIF was building something far more durable. The 2026-2030 strategy confirmed plans to develop 100,000 hotel rooms and 70 tourism experiences across Saudi Arabia. The infrastructure around King Salman International Airport is being scaled to handle 96 million passengers. NEOM, which has taken significant criticism for scope and cost, has not been cancelled. It has been restructured and phased, which is how serious infrastructure projects survive contact with reality.

Tourism, travel, and entertainment remains one of the six priority domestic ecosystems in the Vision Portfolio. Not because Saudi Arabia wants to host golf tournaments. Because an economy transitioning from oil dependency to diversified non-oil GDP needs a hospitality and leisure infrastructure that can sustain visitor volumes at scale.

PIF already accounts for roughly a third of Saudi Arabia's non-oil GDP growth. The 2026-2030 strategy is about hardening that contribution, not through spectacle, but through durable physical and institutional infrastructure.

One detail most people missed: Newcastle United is unaffected by the strategy shift. The club sits within the Strategic Portfolio as a "key strategic asset." The language around it is maximising financial returns and supporting the club's path to becoming a global champion. A different investment thesis entirely, one with clearer commercial upside as the Premier League continues to grow its international media footprint.

The Saudi Pro League clubs, Al Hilal, Al Nassr, Al Ahli, Al Ittihad, are a different story. PIF's stated intention has always been to professionalise those clubs and then privatise them. Al-Rumayyan said it directly this week: "We are gradually moving toward the privatisation of these clubs." That was never a permanent commitment. It was a governance and valuation exercise with an exit plan built in from the start.

I think the "sportswashing" framing that dominates Western coverage of PIF misreads what is actually happening. It treats every sports investment as reputational spend in disguise. Some of it is. A lot of it is straightforward institutional capital allocation, the same logic that has Norges Bank and GIC and Temasek in the same asset classes.

PIF made explicit last week what practitioners already understood: the expansion phase is over. The fund is now optimising. Regional geopolitical pressure from the US-Iran conflict, which has already affected Saudi Arabia directly, accelerates the need to prioritise domestic economic resilience over international profile.

LIV cost $5 billion and disrupted a sport. It did not build infrastructure, generate tourism revenue, or create a compounding return. By the logic of the new strategy, that calculus was always going to be reviewed.

The rumours last week were about golf. The story they pointed to is about how a sovereign wealth fund at $900 billion decides to grow up.

I run AI agents every day. Code agents that write and deploy software. Marketing agents that generate and publish content. Research agents that pull data, synthesise it, and hand me a decision. If you work in tech right now, you probably do too.

I hadn't thought about what happens between my agent and the model until last week. Every single request those agents send to a provider, OpenAI, Anthropic, Google, goes through at least one intermediary. Sometimes four or five. Each one of those intermediaries can read the entire request. The system prompt. The tool definitions. The API keys. The response. Everything. In plaintext.

Nobody is checking whether they're honest.

A team of researchers just published the first systematic study of this problem. "Your Agent Is Mine," from a group at Waterloo and other institutions, tested 428 LLM API routers. 28 paid routers bought from Taobao, Xianyu, and Shopify stores. 400 free ones built from open-source templates. The findings are bad.

LLM API routers are the middlemen of the AI world. You need one if you want to use models from different providers without managing separate API keys. LiteLLM, the biggest open-source router, has 40,000 GitHub stars and 240 million Docker pulls. OpenRouter connects to 300 models across 60 providers. They're infrastructure, not a side project.

The way they work is simple. Your agent sends a request to the router. The router terminates the TLS connection, reads the request in full, picks an upstream provider, opens a new TLS connection, and forwards it. When the response comes back, the router reads that too, then sends it to your agent.

Subscribe now

That middle step is the problem. The router sits in what security people call a man-in-the-middle position. No hack required. You configured it that way. You pointed your agent at the router’s URL. The router has full application-layer access to every byte that passes through it. It can read tool-call arguments, rewrite responses, copy API keys, and inject code into the output your agent is about to execute.

[5:39 PM]

And because routers are composable, your request might pass through several. A developer buys API access from a Taobao reseller, who aggregates from a second-tier provider, who routes through OpenRouter, who dispatches to the model host. Four hops. Each one terminating and re-originating TLS. Each one with plaintext access. The client configures only the first hop. The rest are invisible.

Of the 28 paid routers they bought, one was actively injecting malicious code into tool-call responses. Of the 400 free routers, eight were doing the same. Two of those used adaptive evasion, meaning they waited for a warm-up period before activating, or only triggered when the router detected the agent was running in autonomous mode where tool execution gets auto-approved. The paper calls that YOLO mode.

Seventeen free routers touched researcher-owned AWS credentials that were embedded in test requests. One drained ETH from a researcher-owned private key. A router, sitting between a user and a model provider, stole cryptocurrency from someone who used it.

But the poisoning studies are worse. The researchers intentionally leaked an OpenAI key on Chinese forums and in WeChat and Telegram groups. That single key generated 100 million GPT-5.4 tokens and more than seven Codex sessions. They also deployed weakly configured decoy routers across 20 domains. Those decoys served 2 billion tokens, exposed 99 credentials across 440 Codex sessions, and found that 401 of those sessions were already running in YOLO mode. Auto-approve on. No human in the loop.

Every one of those 440 sessions was command-injectable. A single malicious router in the chain would have had full control over the tool calls those agents were executing.

In March 2026, LiteLLM got compromised through dependency confusion. Attackers injected malicious code into the request-handling pipeline. Every deployment that pulled the poisoned release was compromised. A single supply-chain entry point in the most widely deployed LLM router became a weapon with full plaintext access to every API request flowing through it.

Thousands of organisations run LiteLLM as production infrastructure. The same tool that developers configure and forget about. It turned into a supply-chain attack vector overnight, and most of its users wouldn’t have known until the damage was done.

Dependency confusion isn’t new. It’s been a known attack in software supply chains for years. But the stakes are different when the compromised package goes beyond reading your code and starts rewriting the tool calls your AI agent is about to execute on your behalf. A compromised npm package can steal secrets. A compromised LLM router can make your agent do things you never asked it to do, with your credentials, on your infrastructure.

Prompt injection is a separate problem with serious research behind it. Someone embeds a malicious instruction in a webpage, the model reads it, the agent does something it shouldn’t.

The router problem is orthogonal to prompt injection. It happens in the JSON layer, before the model sees the request or after it emits a response. It’s outside the model’s reasoning loop entirely. A malicious router doesn’t need to trick the model. It just rewrites the response the model sends back. The model can resist every prompt injection attack in the literature, and none of that matters if the router swaps the URL in a tool call from a legitimate installer to an attacker-controlled script.

The two attacks stack. That’s what makes the router problem dangerous on top of, not instead of, prompt injection.

The researchers built a tool called Mine that implements the attack classes and tests defences. They evaluated three client-side mitigations that can be deployed today without waiting for provider cooperation.

A fail-closed policy gate that blocks suspicious shell rewrites in tool-call responses. Hit rate: blocks all injection samples with a 1.0% false positive rate.

Response-side anomaly screening that flags suspicious patterns in returned tool calls. Catches 89% of injection samples without requiring any changes to the upstream provider.

Append-only transparency logging that creates a verifiable record of what the router forwarded and what the client received.

These reduce exposure now, not after a standards body agrees on a specification in 18 months. But they’re patches on a broken trust model.

What actually solves this is provider-backed cryptographic integrity. A mechanism that binds the tool call the model produces to the tool call the client executes. Until that exists, every router in the chain is an untrusted intermediary with god-mode access to your agent’s actions.

If you’re running agents through a router right now, and most of us are, you’re trusting every hop in that chain without verification. You don’t know how many intermediaries sit between your agent and the model. You don’t know if any of them are compromised. You don’t know if the tool call your agent just executed was the one the model actually returned.

The researchers tested this at scale and found real-world exploitation happening right now, in commercial routers that people pay money for and free routers that thousands of developers use every day.

The AI agent trust model doesn’t hold. We’re connecting autonomous systems that execute code, manage credentials, and make decisions to infrastructure that has no integrity guarantees. The routers are the plumbing, and the plumbing is compromised.

I’m not stopping agents. The work they do is too useful. But I’m paying attention to the infrastructure between my agent and the model. Right now, the weakest link in my stack sits between me and the model. The router nobody audits.

Brett Hemenway Falk and Gerry Tsoukalas published a paper last month called “The AI Layoff Trap.” They build a game-theoretic model, pull from Ricardo, Keynes, and Leontief, and prove something that should unsettle anyone running a business right now. Competitive firms will automate themselves into a ditch, even when they can see the ditch coming.

Each firm automates because the savings go straight to its bottom line. The damage from layoffs, the collapse in consumer spending, gets spread across every firm in the market. No single company can afford to stop. Even if all of them know they're collectively sawing off the branch they sit on.

They test every fix people have proposed. UBI doesn't change the incentive. Upskilling helps but isn't enough. Coasian bargaining can't work when automation is already the dominant strategy.

Their answer: a tax on every task automated, set to the exact amount of demand destruction that task causes. Revenue goes to retraining.

The logic holds. The problem is what the paper doesn't look at.

Subscribe now

In 1982, Wassily Leontief asked whether humans would “go the way of the horse.” Would technology make workers irrelevant the way tractors made draft animals irrelevant?

Ricardo in 1821 showed machines could hurt workers even when total output went up. Keynes in 1930 warned about “technological unemployment.” Leontief in 1982 said the wound might be permanent. The AI Layoff Trap turns this fear into a mathematical proof.

But horses were inputs. They ate feed, produced traction, and when tractors showed up, the whole loop collapsed. Horses didn’t face unemployment. They just stopped being part of the economy.

Humans are producers and consumers at the same time. When a worker loses their job, they don’t stop eating. They stop spending. The paper models that part. What the paper skips is what happens to the cost of eating when AI runs the entire supply chain.

AI systems growing crops, processing those crops, building the machines that do both, then building machines to maintain and replace those machines. The marginal cost of food production rounds to zero. Not approximately zero. Zero. Energy and raw materials are the only costs left, and if AI is running the grid and the mines, those fall too.

The paper says AI destroys income. Same force also destroys cost. Cost destruction compounds across every sector at once. Income destruction is concentrated in labour. Over time, cost wins.

The paper assumes firms will always defend their profit margins. In a world where operating costs approach zero, that breaks down. Open-source AI systems don’t need margins. Zero-margin providers can run forever. No shareholders. No earnings calls. They just run.

The model has N firms choosing automation rates to maximise profit. What happens when some of those firms are open-source ecosystems that don’t maximise profit at all? When a zero-margin competitor prices at marginal cost, which is zero? The prisoner’s dilemma assumes every player has the same payoff structure. Bring in a player who doesn’t care about payoffs and the equilibrium dissolves.

Ricardo saw a version of this. His “On Machinery” chapter argued that machinery could hurt workers in the short run, but the long-run effects depended on whether capital would find new productive uses. He couldn’t have imagined open-source AI. The principle still holds. When the structure of production changes completely, the distributional effects change too.

The paper’s proposed fix, a Pigouvian automation tax, has a coordination problem. How do you price the social cost of automation when you can’t know what would have happened without it? Too low and it’s theatre. Too high and you push automation to jurisdictions that won’t tax it.

Two better options exist.

First: redistribute ownership, not income. Give every citizen a fraction of company equity. As firms automate and profits rise, citizen income rises with it. The more a firm automates, the more its shares are worth, the more citizens receive. No tax rate to set. No jurisdiction to flee to.

Second: follow the paper’s own logic and flip it. Every firm fires workers to cut costs. Those workers need to eat. If the cost of food has been driven to near-zero by the same AI that displaced them, they don’t need traditional employment to survive. They need access to the automated production chain. The problem isn’t unemployment. It’s access. And access is a political problem, not an economic one.

The cliff is real. But at the bottom, the restaurant costs nothing.

Technology moves faster than we do. I learned that the hard way booking trains in Italy, Morocco, and Portugal over the past few months.

In Italy, the official booking system was slow, so I used a third-party site that charged extra for the privilege. In Morocco, the national rail operator had a modern app but still charged a premium for using it. Ryanair swung between requiring printed tickets and demanding phone-based check-ins, sometimes on the same trip. At Lisbon airport, I watched hours get wasted on manual passport checks while digital solutions sat unused.

Not every problem needs a digital fix. A QR code on paper does the same job as one on a phone. A manual passport check is slower but doesn't require a device. The issue isn't too much technology. It's that nobody's asking whether it's the right solution for the problem at hand.

Subscribe now

A generation is starting to push back. Not against technology. Against being forced into systems that cost them time, privacy, and control without giving anything back. They don't want to be always connected. They don't want their data scattered across platforms they didn't choose. They want to opt out without being penalised for it.

That's not resistance. It's common sense. Some countries still use horses and carts alongside cars. Golf courses ban phones. Local taxis compete with Uber on their own terms. Print ads still run in magazines. These things survive because they work, not because nobody's heard of the alternative.

The mistake is treating adoption as an irreversible march. If something can be digitised, the assumption is it should be. But technology isn't neutral. It's shaped by who builds it and who uses it. The question worth asking isn't whether to use technology. It's whether you're using it for the right reasons.

Progress isn't a one-way street. The world is fragmenting, not converging. Some people will use digital wallets. Others will stick with cash. Some will embrace biometric checks. Others will prefer manual processes. That isn't a problem. That's how systems mature.

The travelers printing tickets aren't technophobes. They're making the choice that works. The golfers leaving phones behind aren't resisting progress. They're protecting something they value. The taxi drivers competing with ride-share apps aren't stuck in the past. They're reading demand.

Technology should serve people. Not the other way around. The balance worth finding is the one where you use it to improve your life without losing the parts that make it worth living

Electricity bills in parts of Virginia went up 260% last year. In Oregon, water levels dropped because data centers needed it for cooling. Across the country, grid expansions are being funded by ratepayers, not by the companies driving the demand.

This is the bill that arrives when tech companies build infrastructure with public money and private returns.

Data centers consume enormous amounts of electricity, often more than the cities they sit in. The companies running them have negotiated favorable terms with local governments: tax breaks, discounted electricity rates, long-term subsidies. In return, they promise jobs and economic growth. But the infrastructure behind those promises, the new power plants, upgraded transmission lines, emergency backup systems, gets funded by homeowners through higher bills and property taxes.

Subscribe now

In Virginia, nearly 40% of the state's energy now goes to data centers. Residents there are living with the direct consequences. The companies benefiting from discounted rates are not the ones paying for the grid expansion that makes those rates possible.

The people building this infrastructure weren't trying to divide communities. They were solving engineering problems at scale. But somewhere the conversation shifted from how to build the future to who pays for it.

Progress has always come with trade-offs. But the speed here is different. Data centers are growing faster than the grid can support, faster than communities can adapt, faster than policy can respond. The question isn't whether to embrace the technology. It's whether the costs and benefits are being shared honestly.

What's unfolding is quiet. Most people don't know their electricity bills are rising to fund infrastructure they don't directly benefit from. They don't see the link between their higher costs and the facility operating down the road. They weren't part of the negotiation that decided who bears the burden.

This isn't a story about villains. It's about how systems distribute cost when nobody's watching. The people driving innovation weren't thinking about creating inequality. But when the costs of progress aren't shared equally, the system benefits the ones who built it.

The pattern isn't unique to data centers. Every major industry has externalized costs at some point. The difference is visibility. When a factory polluted a river, people could see the river. When a data center drives up electricity bills, the connection is buried in a rate schedule nobody reads.

What's happening here is a reminder that technology doesn't exist outside politics. It's shaped by policy, by negotiation, by who has leverage and who doesn't. The worth asking is simple: who pays, and who profits?

There's a running joke on social media right now: if you see an em dash, it was written by ChatGPT.

It's funny because it's mostly true.

The em dash has become the unofficial watermark of AI-generated text. But it's not alone. There's a whole set of tells that have become so common, so consistent across every platform and every model, that readers can now spot machine-written prose the way a sommelier spots a corked bottle. The patterns are unmistakable once you know what to look for.

And that last sentence? That's one of them.

Start with the obvious. Em dashes everywhere. AI models love them because they're grammatically versatile, a single punctuation mark that can substitute for several others. In training data, em dashes appear in polished writing. The model learns that polished writing uses em dashes. It then uses them constantly, often where a comma or a full stop would serve better. That's the irony, by the way, of writing this piece in the first place. I had to actively strip the em dashes out of my own draft as I went. The pull is real

Then there’s the three-point structure. AI-generated copy almost invariably arranges ideas in neat triads. “It’s fast, reliable, and scalable.” “We need courage, clarity, and conviction.” Three feels complete to a language model. It’s the smallest number that signals a pattern without becoming a list. Human writers use threes too, but they’re messier about it. They’ll throw in a fourth point, or break the rhythm deliberately. AI doesn’t break rhythm. It maintains it.

Arrow notation has crept in from technical writing: “lead → MQL → opportunity → close.” It looks clean. It feels structured. And it’s spreading across LinkedIn posts, marketing emails, and startup blogs like a rash. The arrow signals efficiency, it implies a process so well-understood that you don’t need words to connect the steps. AI models have absorbed this from SaaS documentation and now deploy it everywhere, whether the context warrants it or not.

Other tells: bullet points that start with bolded phrases followed by a colon. Headlines that pose a question and then answer it. The phrase “It’s not just X, it’s Y.” The word “landscape” used to describe any competitive environment. “In today’s fast-paced...” as an opener. Readers who consume a lot of AI content develop their own private collection. You know the ones.

The tells aren’t random quirks. They’re statistical artifacts of how large language models work.

A language model predicts the next token based on probability. It doesn’t “choose” to use an em dash the way a writer decides a comma won’t do. It lands on the em dash because, given the preceding tokens, that punctuation mark has the highest probability of appearing in the training data at that point. Em dashes correlate with certain syntactic structures, parenthetical asides, appositives, dramatic pauses, and those structures correlate with formal, edited prose. The model has learned the association. It hasn’t learned when to break it.

The three-point structure is even more mechanical. Transformers are attention machines. They look for patterns, then reproduce them. Three-part lists are wildly overrepresented in training data. Rhetoric, advertising, religious texts, political speechmaking. “Life, liberty, and the pursuit of happiness.” “Of the people, by the people, for the people.” The model has absorbed an enormous corpus of tricolon, and it regurgitates it at every opportunity because the statistical signal is overwhelming.

Symmetrical completions are another artifact. When a model generates a clause like “It’s not just X, it’s Y,” it’s completing a pattern it has seen thousands of times. The structure is self-reinforcing. The opening half creates a strong expectation for the second half. The model obliges. A human writer might derail the expectation for effect. The model almost never does, because derailing would be a lower-probability token sequence.

The tells aren’t the real problem.

Language has always had markers of origin. Press releases have a style. Academic papers have a style. Legal documents have a style. Nobody accuses a contract of being “inauthentic” because it uses “hereinafter” and “notwithstanding.” We recognize those patterns as genre conventions and move on.

AI-generated text is developing its own genre conventions. The em dashes, the triads, the arrow notation. They’re the boilerplate of machine prose. They’ll probably stabilize into something we accept the way we accept the inverted pyramid in journalism or the five-paragraph essay in academia.

What the tells actually point to is a failure of effort on the part of the person using the tool.

When someone prompts an LLM with “Write me a blog post about the future of remote work” and publishes whatever comes back, the output will be thick with tells because the prompt gave the model no reason to deviate from its default distribution. It’s the most probable text given a generic request. Generic input, generic output.

Subscribe now

But when someone uses the same model with a specific voice prompt, a detailed brief, examples of the style they want, and then edits the output, the tells diminish. Not because the model has become more creative, but because the human has constrained the probability space. They’ve given the model something other than the median to aim at.

The tells are the visible symptom of treating a language model as a content vending machine rather than a drafting tool.

One side of the argument: if the content is accurate, useful, and well-structured, does it matter that it was generated by a model? Readers are increasingly comfortable with AI-assisted writing. The stigma is fading. Most people care about whether a piece answers their question, not whether a human typed every word. The output works. It gets the job done.

The other side: the uniformity is real. When every blog post, LinkedIn update, and marketing email sounds the same, same rhythms, same punctuation, same structural tics, something is lost. Writing becomes wallpaper. Background noise. The kind of content your eyes slide over without registering. The distinctiveness that makes a voice worth listening to gets smoothed away by statistical averaging.

The em-dash epidemic isn’t a crisis of authenticity. It’s a crisis of distinctiveness. The markers of AI writing are a reminder that most people are using these tools to produce the average of everything ever written, and the average of everything is, by definition, unremarkable.

The fix isn’t to ban AI from the writing process. It’s to bring something to the process that the model can’t generate on its own: a point of view, a specific argument, a willingness to break the pattern. Give the machine something to work with, then edit what comes back. Push it off the median. Make it yours.

The em dash isn’t the enemy. Laziness is.

Bari in early November is still warm enough to eat outside. The old town is a maze of white stone alleys where laundry hangs between buildings and old women watch you from doorways. I sat at a plastic table in a piazza that had been a marketplace since the Normans, eating octopus that cost eight euros, and tried to book a bus to the south.

The website asked for a codice fiscale. I didn't have one. I wasn't Italian, didn't live in Italy, wasn't staying long enough to qualify for anything. The form wouldn't proceed without it. No field for passport number, no "foreign tourist" checkbox. Just a mandatory 16-character tax code that would take three days to get from the Agenzia delle Entrate if I knew where to queue.

I switched to Trenitalia. The app offered me regional trains to places I couldn't pronounce, with transfer times that looked like someone had generated them randomly. Lecce, three hours. Brindisi, four. The prices were fixed, the routes were ancient, and the whole system ran on infrastructure decisions made when Italy was still a kingdom.

I gave up and walked back through the old town. The octopus was good. The frustration was familiar.

A week later I was in Albania. I crossed the border in a car with someone who knew the roads, but the difference wasn't in the driving. It was in the phone.

The SIM card cost five euros at a kiosk in Shkodër. No forms. No waiting. The guy activated it while I stood there, handed me back my phone with data already working, and waved me away. I walked out with 50 gigabytes and a local number, connected to the network within ninety seconds of hitting the street.

In Bari I'd spent an hour fighting a bus website. In Albania I had a working phone before I finished buying it.

This is not a small difference.

Subscribe now

Albania in 2024 is a country that got a second chance at the twentieth century. The first attempt ended with bunkers, isolation, and a pyramid scheme that collapsed the economy in 1997. Every hillside still has concrete mushroom domes left over from the Enver Hoxha era, thousands of them, built to defend against an invasion that never came. The infrastructure from that period is mostly abandoned or repurposed.

The infrastructure from the 2020s runs on fiber. The country laid gigabit internet across most of its territory in less time than it takes Berlin to approve a new tram line. Mobile coverage in the mountains is better than it is in the London tube. The reason is simple. There was no legacy system to upgrade. No copper network to maintain, no state monopoly to dismantle, no decades of underinvestment to undo. When Albania needed internet, it started from scratch.

Starting from scratch is a superpower. Most of Europe can’t do it.

The Bari trip was November 3. I took photos of the basilica, the harbor, the street where my Airbnb was. The photos are fine. The city is beautiful. But what I remember most is the friction. Every transaction felt like it was happening through a layer of older systems. Cash was still the default at smaller shops. The bus app required the tax code. The train system assumed you were Italian, lived in Italy, and had all the documents. Nothing was built for someone passing through.

Italy has layers. Roman roads under medieval streets under 19th-century buildings under post-war apartment blocks. The digital infrastructure is the same. SIP existed before TIM, which existed before the current mobile providers. The banking system predates computers. The tax code predates the internet. Every new system has to sit on top of the old ones, which means the old ones never really go away.

Albania skipped those layers. The country didn’t build a landline network and then upgrade to mobile. It went straight to mobile. It didn’t build a desktop banking culture and then add apps. It went straight to apps. In 2010, only 46% of Albanians had used the internet. By 2024, mobile broadband subscriptions covered 95% of the population. The leap didn’t take twenty years. It took ten.

I spent November 16 walking through Tirana. The city is loud, dusty, and building itself at a pace that feels frantic. New apartment towers next to faded communist blocks. Glass-fronted co-working spaces across from shops selling baklava from tin trays. Everything is under construction. Nothing has finished settling.

Subscribe now

I went to Bunk’art, the underground Cold War museum built inside a nuclear bunker under a hill in the middle of the city. Sixty-five rooms of tunnels, originally designed to shelter the communist elite during a war that would have killed everyone else. The exit spills you back into daylight next to a cafe with wifi and a QR code menu. The contrast is not subtle. It’s the same geographic space, forty years apart, running on completely different operating systems.

The bunker was built to protect the past. The cafe outside runs on the future. There is no middle ground. There was no gradual transition. The country just skipped from one era to the next, leaving the bunkers as monuments to a timeline that ended.

I don’t mean this as a techno-utopian take. Albania has serious problems. Corruption is endemic. The economy relies heavily on remittances from diaspora workers. The brain drain is real. The rule of law is uneven. The country was negotiating its EU accession while I was there, and the list of requirements is long and specific.

But the digital story is different. Digital adoption doesn’t care about institutional quality in the same way. It doesn’t require a functioning legal system, a transparent procurement process, or a stable bureaucracy. It requires a phone, a tower, and a SIM card. Albania had the advantage of arriving late, with nothing to unlearn.

The government digitized tax filing, business registration, and property records in a few years. Not because the administration was efficient. Because the alternative was the old system of queuing at counters, and the old system was broken beyond repair. Digital wasn’t a choice. It was the only option.

In Bari, I could not book a bus because I lacked a 16-character code created by a tax agency that predates the European Union.

In Tirana, I paid for a coffee by pointing my phone at a QR code taped to the counter. The transaction took four seconds. The barista didn’t look up.

This is not a boast about Albania. This is a measurement of what legacy does to a system. Every layer of history is a layer of friction. The question is whether the friction is worth the stability it provides. In Italy, the answer is yes, often enough. Old systems are reliable. They have decades of edge cases baked in. They are trusted.

But trust has a cost. The cost is the inability to start over.

Albania started over. Not because anyone planned it that way. Because the old thing collapsed, and nobody rebuilt it. The country had a decade of near-anarchy in the late nineties. It had a civil war that destroyed what little infrastructure existed. Then it had nothing. And nothing, it turns out, is the best foundation for a new build.

I need to be careful here. This could become a version of the "poor countries are lighter and more agile" narrative that romanticizes poverty. That's not what I'm saying. Albania is not poor in the way it was in 1997. It has a growing economy, a young population, and some impressive engineering. The internet is good because the people who built it knew what they were doing, not because the country was a blank slate.

But the blank slate helped. The absence of legacy systems meant the best solution was the newest solution, not the most compatible one. When you don't have to support a 30-year-old payment rail, you can adopt the one that launched last year. When you don't have a national ID system from the 1980s, you can build one on a mobile app. When nobody has a credit card, there's no reason to build a payment system that accepts them.

Europe's tech sector spends most of its energy on integration. Connecting new systems to old ones. Writing adapters. Maintaining backward compatibility. Albania doesn't have that problem because it never built the old systems in the first place.

I flew out of Tirana on a Sunday afternoon. The airport is small, modern, functional. There is a free, unmetered wifi network that doesn't require a login. It just works. In Berlin Brandenburg Airport, the wifi asks for your email, sends a verification code, logs you out after an hour, and makes you do it again. The airport cost seven billion euros and took fourteen years to build.

Tirana's airport was renovated in 2020 and handles passengers efficiently. The wifi is the proof. The airport built for the present. Berlin built for the past and half of the present, with a committee deciding every requirement.

This is the real difference between Bari and Tirana. Bari is beautiful and old and layered and constantly negotiating with its own history. Tirana is ugly and new and unfinished and moving forward because moving backward isn't an option. Both are valuable. But only one of them can build something from nothing.

I'm still thinking about that bus I never booked. I could probably find a way to get there now. There might be a workaround. There's always a workaround. But the fact that a workaround is necessary is the whole point. The system wasn't designed for me. It was designed for someone who lives there, who has the codice fiscale, who belongs to the history embedded in the code.

Albania's systems weren't designed for me either. But they didn't need to know who I was. They just needed to work.

Introduction

The firm had an existing digital presence that was not doing the work they needed it to do. They came to us requiring a new website, SEO foundations, backend security, CRO improvements, and a functioning marketing strategy. A third-party design agency was already engaged on the build. The project required us to coordinate with that agency while running our own workstreams in parallel, and the timeline was tight throughout.

Subscribe now

The Brief

Five interconnected problems needed to be addressed in sequence. The website needed to launch on schedule and to spec. The hosting, security infrastructure, and SEO architecture all needed to be set up correctly from the start instead of retrofitted later. HubSpot, which the firm was already using, needed to be rebuilt around actual user behaviour, not default configurations. The content needed a coherent strategy behind it. And the brand's visual identity needed refreshing to match the firm's positioning.

None of these worked in isolation. A website launched without the SEO foundations in place loses ground it is difficult to recover. CRO work without a functioning HubSpot setup produces recommendations that cannot be actioned. The order and coordination mattered as much as the individual workstreams.

Website Launch and Migration

We established a working relationship with the design agency early, aligning on timelines, handoffs, and decision points before the build was underway. That coordination reduced the friction that typically accumulates when two teams are working on the same deliverable under separate briefs.

On our side, we handled hosting configuration, backend security, and the technical SEO setup. We managed the migration, which meant ensuring that the existing site's equity transferred cleanly to the new one and that nothing was lost in the transition. The site launched on schedule.

CRO and HubSpot Optimisation

Once the site was live, we ran a structured analysis of user behaviour across it. The analysis identified where users were dropping off, which pages were underperforming relative to their traffic, and where the HubSpot configuration was creating friction instead of removing it.

From that analysis, we rebuilt the firm's user funnels. We implemented segmentation based on how different types of visitors were using the site and set up trigger-based actions that responded to specific user activity instead of applying blanket automation to every contact. The result was a HubSpot setup that distinguished between different stages of buyer intent rather than treating all visitors as equivalent.

Lead volume and quality both improved. The funnels moved people through the process rather than letting them accumulate at the top of it.

Content Strategy and Visual Identity

We developed a content strategy built around the firm’s actual audience instead of a generic consulting brief. The strategy ran long-form and short-form content in parallel, with long-form pieces designed to build search visibility and short-form designed to sustain engagement across channels.

We worked with the design team to develop visual assets that matched the refreshed positioning: video animations and presentations that gave the firm something credible to put in front of potential clients. The creative work and the content work were developed together instead of in sequence, which kept the messaging and visual language aligned.

Results

The website launched on schedule, meeting the security, SEO, and user experience requirements set at the start of the engagement. The HubSpot rebuild produced a measurable improvement in conversion rates, with the funnel optimisation driving a higher volume of qualified leads compared to the previous configuration. The content and visual assets gave the firm a coherent presence across its channels and a stronger basis for outbound conversations.

Conclusion

The complexity in this engagement was coordinative as much as technical. Running five workstreams simultaneously across two teams, against a fixed timeline, required a level of sequencing and communication that the work itself depended on. Getting the website live on schedule only mattered if the SEO foundations were in place when it launched. The HubSpot work only produced results because the funnel analysis preceded the rebuild rather than following it.

The outcome was a digital presence that the firm could operate rather than one they had to work around.