My new manager had been with the company for six weeks. I had been there fifteen years.
He asked what kind of work I did. I told him: configuration, which is coding that happens outside the code. SQL queries. Functional analysis documents. Business analysis, Excel, test cases, requirements. I tried not to go into too much detail, and still, with every item, I watched the concern grow on his face.
It was not the concern of a man hearing a colleague describe failure. I am good at these things. It took me fifteen years to get good at them. It was something else: a man sorting my expertise into two columns and finding the safe column empty.
Everything I described was cognitive work. The drafting, the analysis, the specifying, the translating of business intent into structured documents. This was the work people like me had been told to move toward when machines took over the manual labour.
He meant it kindly. He was, in his way, trying to save me.
“Danny,” he said, leaning in with the quiet weight of someone sharing an inescapable truth, “you’re turning fifty. With everything changing around us, and every company demanding brand-new skills, the only question that matters is this: how do you remain relevant?”
He didn’t wait for me to answer. He had a question of his own: did I use AI?
I did. Quietly, for emails, and to polish meeting minutes. I didn’t say so. I said something vague about watching the technology, and he heard what he needed to hear: not embracing it.
“It’s not about whether AI will replace you,” he continued. “Although it might. That’s just a matter of when. What matters is that if you don’t use AI, you’ll be replaced by the guy with five years of experience who does. When a tiger is chasing you, do you still run, even though you can’t outrun the tiger? . . . Yes, you run! Because you only have to run faster than the other guy.” He looked at me for a pulse of agreement.
I nodded. What else does one in my seat do?
I didn’t hear a joke and a way out; I heard a warning. The tiger is coming for everyone. His advice offered no escape, only a ranking of who gets eaten first.
I would like to say my interest in AI began that day as curiosity. It began as fear. Everything I have learned about these tools, I learned running from that tiger, and this essay is what I saw on the way.
Apparently, your skills now have a half-life.
Less than five years, runs a figure repeated for over a decade. In some technical fields, as little as two and a half. It is a rule of thumb, not a measurement.
What, exactly, has lost half its value? Your knowledge? Your competence? Your salary? Does a software engineer wake up to find that half of what she knew yesterday has evaporated? Is there a professional Geiger counter for the radioactive decay of employability?
Probably not. But the metaphor touches a nerve. Something is expiring, and increasingly that something is us. Not our biological existence, but our economic one: the version of ourselves the world once considered useful, qualified and experienced. The person we spent decades becoming.
We used to assume that person would grow more valuable with time. Now we are advised to check the expiry date.
The Promise
There was once a remarkably simple proposition at the centre of modern life.
You went to school. You learned a trade or earned a degree. You found an employer, worked hard, accumulated experience, and built a life: a home, a mortgage, holidays, children who might enjoy more opportunity than you had.
The arrangement was never universal. Women, minorities, casual labourers and much of the developing world were often excluded from the golden age of stable employment. But that does not make the promise imaginary. For millions of workers in the postwar twentieth century, it was real enough to organise a society around. Banks lent against it. Governments designed pensions around it. Families planned thirty years ahead on the assumption that tomorrow would resemble today.
There was even a phrase for it: lifetime employment.
In the West, we hear those words now as an archaeological curiosity. If you want a pay rise, you leave. The career ladder is climbed between companies, not within them. Notice what that does to everything else. An employee who expects to depart in three years is a resource to be extracted, not cultivated. A company that cannot retain people has no reason to develop them. The worker responds in kind. Each side treats the other as temporary, and each behaves accordingly.
But consider a country where the phrase is not a curiosity.
In Japan, shūshin koyō, employment for life, was a working system with visible machinery. Large firms hired graduates not for a role but as members: the company hired the person first and assigned duties afterwards, on the assumption that the duties would change many times over the decades. Pay rose with seniority rather than performance, held below the young worker's productivity and pushed above it later in life, a bond that made departure costly for the worker and dismissal expensive for the firm. When the economy shifted, the company did not shed the worker. It reassigned, retrained and seconded, moving people between divisions while the storm passed.
Loyalty, in other words, was not a cultural gift the Japanese worker brought to the corporation. It was what the system paid them to produce. The bargain was engineered, not inherited.
And that engineered bargain performed the function the West has forgotten: it made a changing economy survivable for the individual. The worker did not need to predict the future, because the institution had agreed to absorb it on their behalf. The company could afford to train, because it would keep what it trained. The worker could afford to commit, because commitment would be repaid.
The system was always narrower than the legend, and it is now visibly fraying. But it proves the point the West has forgotten. Stability was never a natural property of employment. It was something a society built, and could build.
And what had been built could be dismantled.
The Renegotiation
The bargain did not collapse overnight. It was renegotiated under a succession of reasonable names: Competition. Efficiency. Restructuring. Globalisation. Shareholder value. Flexibility.
Each sounded like progress. Often, each delivered it. Goods became cheaper, supply chains faster, choices wider than our grandparents could imagine.
But beneath the language, the foundation was moving. The employer was quietly shedding responsibility for the continuity of the worker's life. Factories closed. Departments vanished. Occupations migrated across oceans. A profitable company could eliminate ten thousand jobs simply because a more profitable arrangement had become possible. The worker might have done everything right. It no longer mattered.
Loyalty, once a mutual obligation, became a one-sided risk.
The vocabulary adapted to mask the shift. Workers became human resources, then human capital, then talent. And when the organisation no longer needed them, they were made redundant.
Redundant. An extraordinary word to apply to a human being. Not incompetent. Not dishonest. Not unwilling. Simply no longer required.
You could be exceptional at something the economy had quietly decided it no longer needed.
Japan was not spared. Decades later, in 2019, the chairman of Keidanren, its largest business federation, who also chaired Hitachi, said publicly that Japan had reached the limits of thinking about business in terms of lifetime employment. The dismantling travels; only the arrival times differ.
The Freedom to Be Insecure
Eventually, this instability was rebranded as liberation: the gig economy.
For knowledge workers, the pitch was irresistible. The largest freelance platforms now list millions of registered developers, designers and writers. No employer. No office. No fixed hours. Be your own boss.
Look at the machinery underneath. On the biggest of these platforms, a freelancer pays simply to apply for work, whether or not the bid succeeds. A cut of up to fifteen per cent is taken from every payment, at a rate that varies with the contract and with supply and demand. The worker's first recurring cost, before any income arrives, is the permission to ask for income.
Some prosper in this arena, the way A-list actors prosper in an ocean of aspiring actors. Autonomy is a real virtue. But freedom is a hollow gift when it arrives without security. To a developer with savings and a reputation, the platform is opportunity. To a developer living contract to contract, the same arrangement is structural dread, with a subscription fee for the dread itself.
The gig economy did not invent precarious labour. What it did was subtler: it took the offloading of economic risk and expressed it as software. One party gained the freedom to purchase labour only when needed. The other gained the freedom to pay for the right to be considered.
The Education That Never Ends
Education underwent its own quiet mutation.
For generations, it was an investment made early in life: a foundational period of study and sacrifice. You were preparing to become someone: an engineer, a teacher, an accountant. A person with a profession.
Then the language shifted. Your degree was no longer enough. You needed continuing professional development. Then upskilling. Then lifelong learning.
Who could object? It sounds noble, even Socratic. But listen to the word again: lifelong. The lifetime commitment did not disappear. It changed hands. Once the institution owed the worker a lifetime; now the worker owes the market one. Learning stopped being an invitation to flourish and became a maintenance requirement.
We can date the shift in vocabulary, because the vocabulary is not ours. The framework that sorts skills by decay, perishable, semi-durable, durable, appears in IBM's skills-transformation training material, borrowed from a trade magazine: an actuarial table for human capability. Perishable skills have a half-life under two and a half years. It is repeated by a company that sells training.
Sit with that word. Not "evolving." Not "due for a refresh." Perishable: the adjective we reserve for food. Human capabilities, sorted by the logic of a cold-chain manual.
And increasingly the worker finances the updates, after hours, on weekends, between jobs. Should they fail to keep pace, the diagnosis is swift: They failed to adapt. They failed to remain relevant.
Notice what vanishes from that story: the corporation that restructured, the technology that disrupted the field, the economic model that devalued the old qualification. All recede. What remains is an individual who should simply have learned something else.
The Half-Life of a Human Being
Imagine taking the five-year half-life literally. A person spends four years on a degree. By graduation, some of it is already approaching its expiry date. They enter the workforce, build judgment, learn how organisations really function.
I took a version of that path. Fifteen years in, what had I accumulated? Configuration of a product only my company uses. SQL queries against databases only my company built. Functional analysis in a house style only my company reads. Each took years to deepen. And each shares a property that only became visible when a new manager looked at me with growing concern: it is knowledge of a particular place. The judgment underneath transfers. The syntax does not.
This is the arithmetic nobody states plainly. Your judgment compounds. Your company-specific knowledge compounds too, but only while the company values the world in which it was acquired. A fifteen-year veteran and a five-year veteran can look nearly interchangeable on the open market. The extra decade bought depth in one room, and the room can be closed.
Or the room can stay open while the foundations are replaced.
I did not have to leave my company to watch the ground dissolve. When I began, software lived on physical servers, in client data centres, against relational databases we knew intimately. When something broke, we could trace the transaction and follow the failure through a system we understood. Then came virtualisation, cloud, containers, microservices, Kubernetes. Each wave brought new tools, new specialists and a new vocabulary of competence.
Yet the business beneath it barely changed. A lease calculation was still a lease calculation. An end-of-month reconciliation still had to balance, down to the cent. I understood why the software existed and what it had to accomplish. The new specialists understood how to deploy and orchestrate what it ran on. Both forms of knowledge mattered. Only one carried the scent of the future.
A company can retain its customers, increase its productivity and secure its future while steadily devaluing the expertise of the people who built its past. The company survives by changing. The worker must survive the company's changes.
And when companies across an industry make the same transitions, changing employers offers no refuge. The ground is moving everywhere.
You do not have to become incompetent to become obsolete. You do not even have to change jobs.
Which makes the metaphor revealing. What is expiring is not the skill. It is the market's willingness to pay for it. My manager's face, scanning my expertise item by item, was not measuring what I knew. It was estimating what could still be sold.
A person can retain every ounce of intellect and every year of hard-earned experience and still become economically obsolete. Nothing inside the human being need have deteriorated. Something in the system has shifted.
Yet the system diagnoses the human as outdated, and calls it a skills gap.
The Machine That Learns Faster
Then artificial intelligence arrives.
At first, the familiar advice is dusted off: Learn to prompt. Develop AI literacy. Future-proof your career.
The evidence for what these tools can do has already been gathered. In a controlled experiment by researchers at Microsoft, GitHub and MIT, professional developers were asked to build a web server. Those with an AI coding assistant finished in seventy-one minutes. Those without took two hours and forty-one. Fifty-five per cent less time on a single, well-defined task.
Notice who the subjects were: freelancers recruited from a gig platform. The study measuring how much less human labour a task requires was staffed by the workers whose bidding sustains that labour.
The comfortable reading is that technology evolves, humans adapt, the diligent prosper. We told ourselves that story through mechanisation, electrification, computing and the internet. And the gains are not always real. A 2025 trial of sixteen experienced developers working on their own codebases found them slower with early-2025 AI tools, while they believed they were faster. The researchers, METR, have since said developers are probably more sped up now, and that their newer data is too distorted to say by how much. The perception of speed is now sold by the hour. The reality will be settled later, by someone else's budget.
Either way, the direction is not in doubt. A carpenter with a power saw is a faster carpenter. But when software performs substantial parts of the cognitive work itself, the drafting, the coding, the analysis, the specification, the relationship changes.
The old instruction was to move toward what machines could not do. When machines took physical labour, develop your mind. When software automated administration, learn to analyse and design. Each transition offered another destination. Now AI is moving into the territory to which workers were directed. The next skill is no longer a refuge. It is another capability the machine is learning. And the worker is still expected to pay for acquiring it.
My manager was reading my job description against the machine's capability list and finding the overlaps. He was not wrong. He was right about the tiger.
Learning the tool may make you twice as productive, only for your employer to realise they need half as many people. You followed the rules. You adapted. You upskilled. And your reward may be redundancy.
Consider how fast the new skills themselves expire. In 2023, prompt engineering was declared the job of the future, with salaries advertised at up to $335,000. Within two years it was obsolete as a standalone role: not because the skill became useless, but because everyone acquired it and then the software began doing it unassisted. The skill of operating the last generation of AI had a shorter shelf life than the milk in my refrigerator.
How many times can a worker be asked to climb a ladder whose rungs dissolve faster than they can be mounted? And who decided that climbing indefinitely was the purpose of human life?
The Great Transfer
Look across the last half-century and the pattern becomes clear. It is not simple decline. It is a transfer of risk.
Employers who once expected to develop workers now seek workers already developed. Experience that once promised security must compete with recency. Education, once a launching pad, is now a subscription service.
And the transfer did not stop at training. It moved on to the work of being wanted.
There was a time when an employer knew their people, vouched for them, promoted them and carried their reputation. Today the individual is both the product and its brand manager. You keep a profile on a platform with more than a billion members, in a digital supermarket where static inventory goes stale. An unoptimised profile sinks to the bottom shelf. An inactive worker becomes an invisible one. It is no longer enough to be competent; you must stay at eye level.
Every imperative of the modern workplace sounds empowering alone:
Be agile. Be resilient. Be entrepreneurial. Be visible. Be authentic. Own your career. Build the brand of you.
Together they form a remarkable mandate. The economy reserves the right to change its tastes, while the worker pays the equivalent of slotting fees, the premium stores charge for eye-level position: the courses, the photographs, the constant posting, the endless redesign of their own wrapper, all to keep the right to be considered, in a store they will never own.
If you succeed, you earn another brief cycle on the shelf. If you fail to sell, the expiration is declared your own fault.
We have converted structural instability into a moral obligation.
And here is the deeper loss. A person permanently repackaging themselves for eye-level visibility is not building a life. They are waiting to be picked, indefinitely, by an algorithmic inventory manager that scans the packaging so the buyer doesn't have to.
The economy did not stop changing; it never had. What changed was who was expected to absorb the cost of going out of date, and how much of themselves they were expected to spend to stay on display.
The Problem With Remaining Relevant
Recall the words of my manager: remain Relevant.
Relevant to what?
Relevance is never an absolute quality; it is a relationship. A tool is relevant to a job. A skill is relevant to a task. A worker is relevant to a market. When we demand that people remain relevant, we quietly erase the second half of the equation. The instruction conceals a philosophy of human worth: You must adapt to the system. The system owes nothing to you.
The employer need not remain relevant to the life you are trying to build. The economy need not remain relevant to human flourishing. The technology need not remain relevant to the wellbeing of the society that built it.
Human beings are adaptable by nature; it is among our greatest strengths. But when adaptation becomes the principal condition of survival, it becomes a mechanism for shifting institutional failure onto individual shoulders. At what point does adaptability stop being a virtue and become submission?
Technological progress was supposed to reduce the burden of necessity. Machines that perform our work were supposed to make life more secure. Instead, the more capable our tools become, the more often human beings are asked to prove they are worth employing.
The Expiry Date
Whether the five-year half-life is mathematically sound or merely promotional is almost beside the point. What matters is how readily we recognise the world it describes.
The idea that a skilled professional must reinvent their economic identity every few years would once have sounded like the failure of the institutions meant to provide continuity. Today we put it in a corporate brochure and call it career agility. We treat endless self-reinvention as a law of nature rather than a consequence of how we have organised economic life.
An economy can grow more productive without making its people more secure. A corporation can grow vastly richer while making its workforce more fragile. These are not paradoxes. They are outcomes of choices about who receives the benefits of progress and who bears its costs.
The knowledge may remain. The experience may deepen. The person may be more capable than ever. The market has simply decided to value something else. The expiry date is not something we possess. It is something the economy assigns to us.
Now return to the room where I nodded.
My manager's advice was sound by the rules of the game. Run. Run faster than the other guy. What neither of us said is that nobody asked who decided that outrunning each other was the only response, or why the one thing we may not do is ask the system to share the risk.
It has been done. The Japanese worker did not have to outrun anyone; the institution absorbed the tiger. That arrangement was imperfect, narrow, and is now fraying. But it was built by people, on purpose, and what can be built can be built again.
I started paying attention to AI out of fear, and I am still afraid. But the fear has changed shape. It is no longer about how quickly the models learn. It is how quietly we have conceded the premise: that the answer to a faster tiger is a faster human.
What does the economy owe the people whose lives it organises?
What does technological progress owe the society that makes it possible?
Human beings are not perishable goods.
The economy was supposed to be the tool.