The Last Generation to Learn a Trade

How Artificial Intelligence Is Reshaping the Labor Market at a Speed History Has Never Seen, and Why the Transition Will Spare Almost Nobody

In the autumn of 2024, a mid-sized law firm in London quietly instructed its human resources department to stop scheduling interviews for the following year's intake of junior associates. No announcement was made, no press release was issued, and the managing partners said remarkably little to the colleagues who remained. The reason was not a downturn in case volume, and it was not the cumulative weight of salary costs that had finally tipped some internal calculation over the edge. The reason was a software system that had spent the previous 8 months working through 14 years of the firm's archived case files, absorbing the internal logic of contract review, legal research, and document drafting until it could produce work that outperformed the firm's most capable first-year associates by a margin wide enough that continued hiring had become a question with an obvious, uncomfortable answer. Nobody lost their current job that autumn. The question was simply whether those jobs would still exist by the time the next cohort of graduates was ready to apply for them.

That question is no longer hypothetical, and it is no longer limited to law firms in London.

Between 2025 and 2039, a generation is being born into a world that is already mid-transformation, and the children arriving in it will never know the labor market that their parents inherited. Generation Z grew up alongside the internet, and Generation Alpha was handed a tablet before they could read. Generation Beta is arriving into something qualitatively different from either of those technological shifts, because they are not growing up alongside a tool. They are growing up inside a system in which artificial intelligence has already been woven into education, healthcare, transport, finance, law, and the creative industries, not as an optional layer on top of existing structures, but as the connective tissue that holds those structures together. For them, working without AI assistance will feel as strange and unnecessary as choosing to write longhand in a world of keyboards. The real question is not what that world will look like for Generation Beta. The real question is what happens to everyone already living in the world that came before them.

What the Numbers Actually Say

The research on this subject has been accumulating for a decade, and the picture it paints has become more consistent, not less, as the technology itself has matured. The International Monetary Fund assessed in 2024 that roughly 40 percent of jobs globally face meaningful exposure to the capabilities of artificial intelligence, and that this figure rises to approximately 60 percent in advanced, highly digitized economies, where knowledge work constitutes the larger share of the labor force (Cazzaniga et al., 2024). The World Economic Forum's Future of Jobs Report 2025, which drew on survey data from over 1,000 employers representing more than 14 million workers worldwide, projected that 92 million roles would be displaced globally by 2030 while 170 million new roles would emerge, producing a net figure that sounds reassuring until you notice that the 92 million people losing jobs and the 78 million new positions that need filling are not, in most cases, the same people in the same places with the same skills (World Economic Forum, 2025).

McKinsey Global Institute added a figure in late 2025 that deserves more attention than it has received: using today's technology, not the systems being built for the next decade but the tools available right now, approximately 57 percent of the tasks currently performed by the global workforce could, in theory, be automated (McKinsey Global Institute, 2025). Goldman Sachs, updating research originally published in 2023, estimated that generative artificial intelligence could automate the equivalent of 300 million full-time positions worldwide, with 2/3 of all current occupations exposed to some degree of task-level automation (Goldman Sachs, 2025). Oxford University's foundational study by Frey and Osborne, which has been updated and refined several times since its original publication, continues to find that 47 percent of occupations in the United States carry a high probability of automation over the next 10 to 20 years (Frey & Osborne, 2013; Frey & Osborne, 2023 update).

None of these figures exist in isolation, and taken together they describe not a future scenario but a present reality that is already moving faster than most public discourse has managed to track.

A Massachusetts Institute of Technology study published in 2025 found that artificial intelligence systems can already perform the work of 11.7 percent of the entire United States labor market, generating potential savings of up to 1.2 trillion dollars in wages across finance, healthcare, and professional services (MIT Computer Science and Artificial Intelligence Laboratory, 2025). Analysis of 180 million job postings conducted in 2025 found that openings for commercial photographers had declined by 28 percent year over year, that positions for computer graphic artists had fallen by 33 percent over 2 consecutive years, and that roles for writers and content producers had dropped by a similar margin, none of which can be explained by ordinary market fluctuation or sectoral downturn (Bloomberg, 2025). The World Economic Forum found that 41 percent of employers globally were actively planning to reduce their workforce in areas where AI could automate tasks within the following 5 years, and Big Tech's new graduate hiring had already fallen nearly 50 percent from pre-pandemic levels, with AI adoption cited as a primary driver of that contraction (SignalFire, 2025; World Economic Forum, 2025).

In 2025 alone, artificial intelligence was cited as the stated reason for approximately 55,000 layoffs in the United States, according to data compiled by the outplacement firm Challenger, Gray and Christmas (Challenger, Gray & Christmas, 2025). The figures cited above are not projected numbers extracted from future-oriented models. These are people who received termination letters and were told, directly or indirectly, that a machine had made their position redundant.

The Professions That Are Already Disappearing

The secretary was the first casualty that nobody noticed. Administrative roles have been contracting quietly for years, driven initially by email, then by calendar software, then by enterprise communication platforms, and now by AI systems that schedule meetings, draft correspondence, summarize documents, filter communications, prepare briefings, and manage workflows with a level of competence that makes the term personal assistant sound quaint. Research from 2024 found that 60 percent of administrative tasks were already technically automatable with available technology, and the World Economic Forum projected that more than 7.5 million data entry positions would be eliminated by 2027 alone (World Economic Forum, 2025).

The paralegal and the legal assistant occupy a role that sits at an interesting intersection of high skill and high repetition, because the work they do, reviewing contracts, researching case law, summarizing depositions, drafting standard motions, is cognitively demanding in the sense that it requires legal knowledge, but it is also structurally predictable in the sense that it follows patterns which large language models have now demonstrated they can learn and reproduce. Stanford University's Hoover Institution reported in early 2026 that hiring among entry-level workers aged 22 to 26 had already ground to a halt in several AI-exposed sectors including legal services, medical transcription, and copywriting, and that the decline had been accelerating since 2025 (Stanford Report, 2026). Law firms deploying AI document review tools have already reported that a single system can handle the workload previously assigned to a team of junior associates in a fraction of the time and at a small fraction of the cost.

Translators and interpreters belong among the most heavily exposed professions in every study that has attempted to measure how closely machine capability now tracks human work. The gap between what a professional translator could produce in 2020 and what a large language model can produce today has narrowed to a degree that has made the profession largely indefensible as a mass employer. Professional translation agencies across Europe and North America have been reducing headcount steadily, not because the demand for translated content has declined but because the content is increasingly being produced, edited, and finalized without a human translator in the process at all.

The photographer's situation is visible in the job posting data and in the business practices of the advertising and marketing industries simultaneously. AI-generated content already powers more than 68 percent of images used in marketing and social media campaigns as of 2025 (PhotoGPT AI, 2025), and the shift is structural rather than cyclical: brands do not need to book a studio, hire a makeup artist, coordinate a model's schedule, and pay a photographer's day rate when a prompt and a rendering system can produce dozens of commercial-quality images in minutes. The makeup artist, the stylist, and the model exist in this equation as integral parts of a workflow that the workflow no longer requires. Virtual fashion shows, AI-generated advertising campaigns using synthetic models, and automated post-production systems have not eliminated human creativity from these industries, but they have made the logistical infrastructure that supported thousands of working professionals in those fields economically unnecessary for a growing share of the companies that previously employed them.

The Architects of the Technology Are Sounding the Alarm

The most striking development of recent months is not that researchers are projecting large-scale displacement, because researchers have been doing that for a decade. The striking development is that the people building the technology are now saying it publicly and with unusual specificity.

Dario Amodei, the chief executive of Anthropic, the artificial intelligence company behind the Claude language models and 1 of the most technically sophisticated organizations in the field, told Axios in May 2025 that artificial intelligence could eliminate up to half of all entry-level white-collar jobs within 1 to 5 years, and that the resulting disruption could push unemployment in the United States to between 10 and 20 percent (Amodei, 2025). He was direct about why he was saying it: "We, as the producers of this technology, have a duty and an obligation to be honest about what is coming." When pressed on timing, he said he would not be surprised to see major effects within a year, and that most government officials and business leaders had either failed to understand the scale of what was approaching or had chosen, for strategic reasons, not to say so publicly.

The historian Yuval Noah Harari, writing in Homo Deus more than 8 years before the current wave of large language models arrived, coined a term that has aged with disturbing precision: the "useless class," a category of people not poor in the traditional sense, not unemployed by accident or economic mismanagement, but made functionally irrelevant by the arrival of machines that could perform the cognitive work that had previously given those people economic value (Harari, 2017). His argument was not that people would become literally useless but that the economic systems of advanced societies would fail to find a meaningful role for them, and that the political and social consequences of that failure would be severe. That argument was deeply controversial when Harari first articulated it in 2017. In 2025 and 2026, it reads as a reasonably accurate description of what the data is starting to show.

May 2026: The Reckoning Arrives in Real Time

When the analysis forming this essay was first published in April 2026, the responses from readers followed a familiar pattern. Some wrote to say the projections seemed exaggerated. Others suggested that society always adapts, that new jobs would appear to replace the old ones, that the picture painted here was too grim to be credible, and that there was no cause for the level of alarm the data implied. The months between that first publication and the present have not been kind to those reassurances.

Oracle, 1 of the largest enterprise software companies in the world, instructed several thousand of its employees in the months before March 2026 to carefully document their work processes and workflows. Many of those employees spent weeks and months producing detailed records of exactly how they did their jobs, how they structured their processes, which decisions they made and under what conditions, because that was what the company asked of them. The stated purpose was improving efficiency across documentation and workflow processes throughout the organization. Then, on March 31, 2026, approximately 30,000 Oracle employees received an email at 6 in the morning informing them that their role had been eliminated and that the current day was their last working day. A written survey of 272 of those laid-off employees found that 62 percent were over 40 years old and that 22 percent had worked at the company for more than 15 years. The documentation exercise, many of them concluded, had been a systematic effort to transfer their institutional knowledge into the AI systems that replaced them. The severance package was 6 weeks of pay regardless of tenure, meaning an employee who had given the company 20 years received the same financial recognition as someone who had been there for 2.

This is what May 2026 looks like in measurable terms. Since January 1st of this year, more than 1,621 companies announced mass layoffs, according to data compiled by Intellizence as of March 2026, and the figure has continued to grow in the weeks since that measurement was taken. By the third week of May 2026, more than 142,000 technology workers had been laid off in the United States alone during the current year, a rate exceeding 1,009 people per day according to the layoff tracker trueup.io. Amazon confirmed its 30,000th corporate job cut in early 2026, the largest workforce reduction in the company's history, executed across 2 waves and framed by management as a structural response to AI-driven efficiency gains, even as the company posted strong consecutive quarterly financial results. Meta laid off approximately 8,000 employees in May 2026, with chief executive Mark Zuckerberg directly connecting the workforce reduction to the company's AI infrastructure investment, which is projected to run between 125 billion and 145 billion dollars for the current year alone. The entire Meta company payroll amounts to roughly 27 billion dollars annually, meaning the AI capital expenditure line is between 4 and 5 times what the company pays all its employees combined. The layoffs are not primarily a cost-cutting story in any conventional sense. They are how the company finances its AI infrastructure build-out by redirecting payroll expense toward capital expenditure.

Cloudflare, a cloud networking and cybersecurity company that had never conducted mass layoffs in its history, announced in May 2026 that it was eliminating 1,100 positions, representing 20 percent of its global workforce. The announcement came alongside first-quarter revenue of 639.8 million dollars, a 34 percent increase year on year and the highest quarterly revenue in the company's history. CEO Matthew Prince was unusually explicit in his memo to employees: the decision was not a traditional cost-cutting exercise of the kind that follows financial underperformance or contracting demand. Internal AI usage at Cloudflare had grown by more than 600 percent in the preceding 3 months alone, and departments across engineering, human resources, finance, and marketing were running thousands of AI agent sessions daily to complete work that had previously required human time and judgment. The jobs being eliminated were not underperforming roles. They were roles that a functioning AI layer had made structurally unnecessary while the company was simultaneously posting record revenue.

DeepL, the Cologne-based translation software company that raised 300 million dollars at a 2 billion dollar valuation in 2024 and was widely considered 1 of Europe's most successful AI startups, cut 250 employees in May 2026, approximately 25 percent of its workforce. CEO and founder Jarek Kutylowski described the decision as a deliberate structural choice rather than an emergency measure. The timing, he wrote, was intentionally early: "We are not waiting until the shift is fully obvious to everyone in the market. The right time to make a move like this is before you have to." A company that builds AI translation tools is restructuring itself because of AI, which sounds circular but is not. It is the logic of every publicly listed company that has invested heavily in AI infrastructure over the past 3 years and now faces investor pressure to demonstrate the returns from that investment in the form of operational efficiency.

These are not companies under financial pressure, and that is precisely what makes the pattern so significant. Amazon, Meta, Cloudflare, Oracle, and DeepL were all reporting strong revenue at the moment they announced their cuts. The reductions are not happening because the businesses are failing. They are happening because the businesses are succeeding with fewer people, and because the investors who funded their AI infrastructure build-out expect the efficiency gains to materialize in headcount reduction. Every publicly listed company that has made substantial AI investments over the past 3 years now faces this same structural pressure, and the incentive to cut is independent of whether revenues are strong or weak.

But there is a causal chain embedded in this logic that the efficiency calculations at the individual company level consistently fail to price, a chain I have described elsewhere on this blog under the concept of Otto sapiens, the version of Homo sapiens who optimizes relentlessly for short-term self-interest and misses the structural consequences of doing so at scale. An AI agent works 24 hours a day, 7 days a week, without sick leave, without employment law protections, without a works council, and at a fraction of the cost of a human employee. The economic case for replacement looks airtight from inside the cost structure of any single company. What it does not account for is the fact that every person removed from a payroll is also a consumer removed from a market. Amazon's core revenue depends on people with discretionary income buying things across the very economies it is helping to deplete. When those people no longer have income, the demand that Amazon depends on contracts from below, not because of competition or product failure, but because the purchasing power of the economies Amazon serves has been systematically reduced. DeepL's enterprise clients are the very companies that are also being restructured out of existence by the same transition. As smaller companies fail, they stop paying for every SaaS subscription, every cloud service, every enterprise tool they formerly used. The efficiency gains that large publicly listed companies are booking today are partially funded by dismantling the market structures that generate their downstream revenue. It is a cycle that no individual corporate decision-maker has the incentive to interrupt voluntarily, and no government on this planet is currently moving at anywhere near the institutional speed required to interrupt it from the outside.

Why the Transition Will Be Brutal

Every historical comparison to previous technological disruptions, the Industrial Revolution, the mechanization of agriculture, the automation of manufacturing, carries a hidden caveat that the optimists consistently underemphasize: those transitions were painful, and they were measured in generations, not years. The people who lost their livelihoods when the power loom arrived did not smoothly transition into the new industries that eventually emerged from industrialization. Many of them never recovered economically, and their children bore the costs of that failure. Their children, and their children's children, navigated the new world that the disruption eventually created. The net outcome, measured across decades, was positive. The individual experience of the transition was frequently devastating.

What is different about the current disruption is the speed at which it is operating. The Industrial Revolution unfolded over roughly a century, giving educational systems, labor markets, and social structures at least some time to adapt, however inadequately. The current transition, driven by software that can be updated, deployed, and scaled globally within weeks, is compressing what would historically have been a generational adjustment into a window of 5 to 10 years. Entry-level positions in finance, law, consulting, technology, creative production, and administrative services are contracting simultaneously, across industries and geographies, in a period when the educational systems that prepare people for those roles are still teaching curricula built for the world that is disappearing.

The retirement system in most developed countries was designed for people who worked from their twenties to their mid-sixties and then lived for perhaps a decade more. Generation Beta is projected to live, on average, between 90 and 100 years, meaning that a career structure built around 40 years of stable employment followed by a decade of retirement makes no biological, financial, or social sense for them (Newsweek, 2025). Multiple career phases, periods of retraining, and flexible income sources are not abstract planning scenarios borrowed from futures research, but concrete mathematical necessities imposed by the arithmetic of longer lives and contracting labor markets. But the structures that would support that kind of working life, portable benefits, lifelong retraining infrastructure, adaptive educational systems, remain largely unbuilt.

The gap between the pace of technological change and the pace of institutional adaptation is not a minor administrative problem. It is the central human crisis of this transition.

The Universal Basic Income Question

The idea that artificial intelligence might generate such sustained displacement that governments would eventually need to provide a guaranteed unconditional income to their citizens has moved, over the past 3 years, from the fringe of economic debate toward something approaching mainstream discussion. Sam Altman, chief executive of OpenAI, has been a public advocate for universal basic income as a response to AI-driven unemployment. Silicon Valley's investment community, whose members profit most directly from the automation of knowledge work, has become 1 of the louder voices calling for income guarantees precisely because they understand better than anyone else how quickly the labor market is being restructured (The Hill, 2025).

The empirical record on guaranteed income programs is more nuanced than either its advocates or its critics tend to acknowledge. Finland's 2-year pilot program, which ran from 2017 to 2018 and provided 560 euros per month unconditionally to 2,000 randomly selected unemployed citizens, found that recipients reported meaningfully better mental health, reduced stress, and greater sense of personal autonomy than the control group, and that the program did not reduce participants' willingness to seek work to the degree that critics had predicted (Blomberg-Kroll et al., 2020). Trials in Wales, Kenya, and several Canadian provinces have produced broadly similar findings, with recipients demonstrating increased entrepreneurial activity, improved educational outcomes for children in recipient households, and greater rather than lesser engagement with community life (Newsweek, 2025).

As of mid-2025, no country has implemented a nationwide universal basic income for its general population. The political obstacles are substantial, the fiscal architecture is unresolved, and the question of who pays for it, whether through taxation of AI-generated corporate profits, redistribution of productivity gains, or some other mechanism, remains genuinely open. What is no longer open is whether the question needs to be answered. The transition from the current labor market to whatever replaces it will require a social safety net capable of holding millions of people during a period of structural displacement that existing unemployment systems were not designed to manage.

What This Means for the Generation Being Born Right Now

The children born between 2025 and 2039 will not experience this transition as disruption, because for them it will simply be the baseline condition of the world they inherit. They will learn to collaborate with AI systems the way previous generations learned to collaborate with colleagues. They will navigate career paths that do not follow a linear trajectory from education to employment to retirement, because that trajectory will not exist for them in any recognizable form. The skills that will define their economic survival, critical thinking precise enough to interrogate AI outputs rather than simply accept them, emotional intelligence robust enough to serve needs that machines genuinely cannot serve, adaptability sustained enough to change fields every few years without losing coherence, are not the skills that most educational systems currently prioritize.

The challenge is not for Generation Beta. They will adapt, because they will have no memory of anything different, and adaptation is easier when there is nothing to unlearn. The challenge is for the generation alive right now, for the paralegal who has invested 4 years training for a role that is contracting, for the photographer who has spent 10 years building a portfolio in a market where 68 percent of commercial image demand is now met by systems that work without a camera or a light stand, for the translator who is watching the software do in seconds what took them an hour, and who has children at home and rent to pay and no obvious path to whatever comes next.

That is the population caught in the transition, and it is large, and it is not being spoken to honestly by the institutions that are supposed to represent its interests. Governments are worried about the political consequences of panic. Companies are worried about what competitors will do with the same information they already have. The universities are still collecting tuition for degrees in fields that are restructuring faster than any 4-year program can track. And the people in the middle, the ones doing the work that the machines are learning to replace, are largely on their own.

Coders, Tradespeople, Soldiers: Nobody Is Safe

There is a conversation that repeats itself, with different participants and the same structure, and it goes something like this. Someone argues that the programmers will be fine, because someone has to build the AI. Someone else argues that the tradespeople will be fine, because you cannot fix a burst pipe with a language model. Both arguments feel satisfying in the moment, and both are wrong in the medium term, and the reason they are wrong is the same reason the earlier generation's arguments about secretaries and photographers were wrong: the relevant question is never whether a skill is valuable but whether the economic structure that previously paid humans to exercise that skill will continue to need humans to do so.

In software development, the shift is already visible in the daily workflow of every serious engineering organization. AI coding assistants have moved from novelty to infrastructure with a speed that caught even their developers off guard. The category of "vibe coding," building functional software through natural language prompts with minimal formal programming knowledge, was a fringe experiment two years ago. That category is not fringe anymore, and the gap between those who have adopted it and those who have not is widening every quarter. The programmer who has not integrated AI tools into every stage of their workflow is already operating at a cost and speed disadvantage that compounds quarterly. The programmer who has integrated those tools can do in an hour what a team of 5 previously needed a week to accomplish. That equation does not require a stretch of imagination to follow to its conclusion: companies need fewer programmers, not more, as the productivity per remaining programmer accelerates. What will be needed is not more people who write code but fewer, better people who understand how to direct the systems that write code, validate what those systems produce, and catch what they get wrong.

The trades are a different conversation, and people having it tend to feel safer. The electrician, the plumber, the structural engineer working on site, the roofer: these are roles that require physical presence, spatial reasoning, and the ability to respond to unpredictable physical environments. In a purely technical sense, the argument for their resilience has more merit than the argument for the programmer's resilience. But it is not as airtight as the people making it believe. At a robotics fair in China that I attended recently, I stood in front of machines that folded metal, laid tile, installed electrical conduit, and carried 80-kilogram loads across uneven surfaces, all without a human hand on them. I left that fair with a feeling I do not use the word for lightly: genuine fear. Not the abstract apprehension of someone reading a report, but the concrete recognition of someone who has seen what the machines already do and understands how far ahead of the public narrative the engineering actually is. The trades are not as protected as the people working in them currently assume. They are simply positioned slightly later on the same timeline that is already consuming the knowledge professions.

The people who will retain economic value in this environment are those who understand how to direct AI systems, how to identify where they fail, how to design workflows around their capabilities and limitations, and how to build things with them that could not be built without them. It is the cognitive shift from executing tasks manually to orchestrating systems that execute tasks at scale. Those who have made that shift belong to a small group whose lead over everyone else is widening faster than the educational systems trying to close it can respond.

The companies that will not survive this decade are not necessarily the ones with the worst products or the weakest balance sheets. They include every organization that has decided, at some leadership level, that AI is a productivity enhancement they can bolt onto existing processes without rethinking those processes from the ground up. I have heard it said, by people running actual businesses: "We just have it write our emails, it's fine, good enough." Those people are not fine. They are 2 years behind a curve that is moving faster than they can run, and by the time the gap becomes visible to them in revenue terms, closing it will require rebuilding things they should have started building in 2023.

On May 1, 2026, the city of Hangzhou deployed 15 humanoid and wheeled robots at key intersections as a traffic police squadron, managing pedestrian flow during the Labor Day holiday. In January 2026, the city of Wuhu in Anhui Province put "Intelligent Police Unit R001" on permanent duty at a busy intersection, a humanoid machine in a police uniform and reflective vest that directs traffic, warns cyclists, and operates continuously without breaks, sick leave, or pension contributions. Multiple other cities across China deployed similar units throughout 2025 and 2026. EngineAI's PM01 humanoid robot appeared on patrol alongside human officers in Shenzhen. None of this is a demonstration or a pilot program. These are functioning operational deployments that cities are scaling systematically rather than experimentally. The Terminator was a film I watched in a cinema with a sense of comfortable distance between that fiction and the world outside. That comfortable distance no longer exists in any meaningful sense.

On the military side, the trajectory is no less direct. Autonomous drones operating without real-time human control have struck targets in Ukraine. Robot ground vehicles have engaged combatants and accepted surrenders. Israel has used AI-assisted targeting systems in Gaza at a scale and speed that no human-operated targeting process could match. Lethal Autonomous Weapon Systems, which can identify and engage targets without a human making the final decision, are no longer science fiction. They are active procurement items in defense budgets being debated right now. The debate in military and legal circles is no longer whether these systems exist, but whether international law can be updated fast enough to govern them. I have recently come across information, through a professional context I will not specify further, that suggests the question of whether standing armies will still serve their current function within a decade has become an active and documented planning problem for serious defense institutions, not a theoretical one.

A Personal Forecast I Intend to Stand By

People who know me understand that I have a track record of being right about things that, at the moment I say them, prompt the response: "George, you're exaggerating." When I first described the trajectory of AI-driven job displacement in early 2025, the responses were predictable in their variety. People argued that society always adapts to technological disruption. Others maintained that new jobs would replace the old ones at roughly the same pace. Several suggested my read on the situation was simply alarmist. Then the Oracle story unfolded, and the Cloudflare story, and the DeepL story, and the 142,000 tech layoffs of the first 5 months of 2026 materialized as a matter of public record. I am not saying this to congratulate myself. I am saying it because the people who told me I was exaggerating are the same people who now need to recalibrate their priors about everything else I am about to say.

The transformation is arriving at a speed that no individual, no company, and no government can respond to in real time. The institutions designed to buffer the social impact of technological change, regulatory agencies, social insurance systems, retraining programs, were built for disruptions that played out over decades. What is happening now plays out in months. By the time a government has identified a problem, held a committee, commissioned a study, received the study, debated the study, and funded a pilot program, the problem has already become 3 times larger.

There is a cognitive profile that will thrive in this environment, and it is not the profile that the educational establishment and corporate hiring managers have spent 40 years selecting for. Autistic and neurodivergent individuals, whose brains process information in parallel rather than sequential structures, who can hold and cross-reference enormous data volumes without the pattern-matching fatigue that affects neurotypical cognition, are structurally better equipped for the AI interface layer than most of the people currently running the organizations that need to navigate it. I am not making a soft inclusive point here. I am making a hard cognitive observation: the mental architecture that was previously treated as a disadvantage in environments optimized for social conformity is a genuine functional advantage in environments where the primary task is directing, validating, and correcting AI systems operating at data volumes that exceed the bandwidth of conventional sequential reasoning.

The expensive consultants advising European governments on AI strategy have impressive titles. What several of them demonstrably lack is the visceral, first-hand understanding of what these systems actually do when deployed, as opposed to what they promise to do in a PowerPoint deck. The advisory ecosystem around European policymaking is, with notable exceptions, structured around academic credentials and institutional relationships rather than operational experience with the technology being advised upon. That is how you end up with AI strategies that are 3 years behind the frontier and written to protect the people who funded the working groups.

On the subject of Germany specifically, I will be direct in a way that my professional context has previously encouraged me to soften: I do not see a viable path from where the German government currently stands to the institutional redesign that the transition demands, within the timeline the transition allows. When private individuals run out of money, they stop ordering takeout, cancel their streaming services, and eat pasta three times a week until the balance recovers. When the German government runs out of money, it announces a constitutional debt brake bypass worth hundreds of billions, then convenes a working group to discuss whether LED lighting in federal buildings constitutes a meaningful efficiency measure. The comedy in that scenario writes itself with minimal effort. The tragedy contained in the same scenario writes itself considerably faster and at considerably greater scale.

The people I know personally, many of them entrepreneurs and professionals from the region around the Starnberger See in Bavaria, who have spent decades building taxable value in Germany, are not waiting for the political situation to improve. Several of them have already completed the move. Several more have purchased property abroad and are currently in the process of doing so. They all say the same thing: the combination of tax pressure, bureaucratic friction, and a government that responds to fiscal stress by extracting more from the people who generate fiscal activity, rather than restructuring the systems that consume it, has exhausted their willingness to stay. They are not wrong, and they are not being selfish. They are applying the same rational calculus that any mobile business applies when the regulatory and fiscal environment of a jurisdiction becomes hostile to the activity that generates the revenue that funds the public budget. The capital flight that follows is not a protest. It is the straightforward arithmetic of incentives applied consistently across time. Every high-net-worth individual or productive company that leaves Germany takes with them not just their direct tax contribution but the employment, the supplier relationships, the downstream consumption, and the knowledge transfer that their activity generated. Cyprus, Malta, Portugal, and the UAE are not growing because they got lucky. They are growing because they made themselves easier to operate in than the alternatives.

Italy, for all the jokes that have been made at its expense over the past 30 years of political instability, has done something that the rest of the EU has not: it became, in September 2025, the first country in the European Union to pass a comprehensive national AI law aligned with the EU AI Act, under Prime Minister Meloni, and it has backed that regulatory clarity with the Italia Digitale 2026 strategy, which allocates real capital toward AI infrastructure, public administration modernization, and digital skills development, supported by EU Recovery and Resilience funds. The OECD's 2026 economic survey of Italy identifies improving fundamentals. It is not a complete turnaround story, and Italy has structural problems it has not resolved. But it is a country that has made a policy decision to position itself at the intersection of regulatory clarity and technological adoption at the moment when that intersection has the highest commercial value. I am paying attention to it for reasons that are not entirely academic.

The countries that will be best positioned in the decade after this one share a structural characteristic: they are either willing to move at AI deployment speed, or they have the authoritarian capacity to force that speed on their institutions without waiting for democratic consensus. The United Arab Emirates, China, Japan, South Korea, Russia, Singapore, and Israel are all, for different reasons and with different social trade-offs, moving faster than the liberal democracies of Western Europe. The United States, in a separate category, is accelerating the disruption with a combination of private sector scale and institutional decay that will produce enormous winners and enormous losers with minimal social infrastructure to manage the distribution between them. The Americans are not driving the sled more carefully than the Germans. They are simply driving it faster, toward a wall that is the same wall, and calling it progress.

What I know with certainty, at the end of all of this, is what I have known at the end of every previous analysis: the people who matter are the ones you can count on when the frameworks fail. Building for that is where my attention goes. I am also, not entirely metaphorically, looking at available property in Italy. I have no intention of dying in Germany, which says something about where I think Germany is heading, and something even more specific about where I intend to be when it gets there.

This article is intended for general informational purposes and represents the author's analysis of publicly available research. All projections cited are drawn from peer-reviewed studies and reports by established research institutions.

References

• Amodei, D. (2025, May). Interview with Axios on AI and the disruption of labor markets. Anthropic.
• Bloomberg. (2025). Job market trend analysis: 180 million postings across creative and knowledge sectors. Bloomberg Intelligence.
• Blomberg-Kroll, O., Hämäläinen, P., Kanninen, O., Koskela, T., Simanainen, M., & Verho, J. (2020). The basic income experiment 2017-2018 in Finland: Preliminary results. Ministry of Social Affairs and Health Finland.
• Cazzaniga, M., Jaumotte, F., Li, L., Melina, G., Panton, A., Pizzinelli, C., Rockall, E., & Tavares, M. M. (2024). Gen-AI: Artificial intelligence and the future of work (IMF Staff Discussion Note SDN/2024/001). International Monetary Fund.
• Challenger, Gray & Christmas. (2025). Annual job cuts report 2025. Challenger, Gray & Christmas.
• Chow, A. R. (2026, April 30). Oracle workers say they were fired after training AI to replace them. TIME Magazine.
• Frey, C. B., & Osborne, M. A. (2013). The future of employment: How susceptible are jobs to computerisation? Oxford Martin School, University of Oxford.
• Frey, C. B., & Osborne, M. A. (2023). The future of employment revisited (Updated analysis). Oxford Martin School, University of Oxford.
• Goldman Sachs. (2025). The potentially large effects of artificial intelligence on economic growth (Updated research note). Goldman Sachs Global Investment Research.
• Harari, Y. N. (2017). Homo Deus: A brief history of tomorrow. Harper.
• Intellizence. (2026). Companies that announced major layoffs and hiring freezes (Tracker, data as of March 22, 2026). Intellizence. Retrieved May 2026, from https://intellizence.com/insights/layoff-downsizing/major-companies-that-announced-mass-layoffs/
• McKinsey Global Institute. (2025). A new future of work: The race to deploy AI and raise skills in Europe and beyond. McKinsey & Company.
• MIT Computer Science and Artificial Intelligence Laboratory. (2025). AI and the future of work: Labor market implications of AI adoption. Massachusetts Institute of Technology.
• Newsweek. (2025). Generation Beta: What we know about the children born between 2025 and 2039. Newsweek.
• OECD. (2026). OECD Economic Surveys: Italy 2026. OECD Publishing.
• OECD. (2025). Progress in implementing the European Union Coordinated Plan on Artificial Intelligence (Volume 1): Italy. OECD Publishing.
• PhotoGPT AI. (2025). AI in marketing: Usage statistics and trends. PhotoGPT AI Research.
• Rauscher, G. A. (2026). The Last Generation to Learn a Trade: How Artificial Intelligence Is Reshaping the Labor Market at a Speed History Has Never Seen, and Why the Transition Will Spare Almost Nobody. Zenodo. https://doi.org/10.5281/zenodo.20394044
• Sifted. (2026, May). DeepL cuts 250 jobs in push to stay ahead in AI race. Sifted.
• SignalFire. (2025). State of talent report: AI's impact on hiring. SignalFire.
• Stanford Hoover Institution. (2026). AI and the entry-level labor market. Stanford University.
• TechCrunch. (2026, May 8). Cloudflare says AI made 1,100 jobs obsolete, even as revenue hit a record high. TechCrunch.
• The Hill. (2025). Tech leaders push for universal basic income amid AI concerns. The Hill.
• TrueUp. (2026). Tech layoffs tracker (Live data, accessed May 22, 2026). TrueUp. Retrieved from https://www.trueup.io/layoffs
• World Economic Forum. (2025). The future of jobs report 2025. World Economic Forum.
• Xinhua. (2026, May 3). China's robot police patrol cities during Labor Day holiday. Xinhua News Agency.