I was pouring my first coffee in Hong Kong when the Oracle news hit my feed. March 31st. 6:00 AM Pacific Time. Thirty thousand people—eighteen percent of their global workforce—woke up to an automated email. Not a meeting. Not a conversation. Just a subject line that essentially said: "Access revoked. Don't come in."
The internet exploded with the usual horror. How could they? The inhumanity! The robots are taking over!
But I sat there staring at the numbers, and the numbers didn't add up to a tragedy. They added to a heist.
Oracle isn't dying. They're swimming in cash—$3.7 billion in net profit last quarter, up 21% year-over-year, with half a trillion dollars in future contracts signed. This wasn't a desperation move. It was a liquidation. Thirty thousand human beings were converted into liquidity so Larry Ellison could buy more Nvidia chips. The people were the budget line item; the GPUs were the strategy.
Welcome to the greatest PR laundering scheme in modern corporate history.
The "AI Efficiency" Theater
If you've been watching the layoff announcements this quarter—Block, Amazon, Meta, the bloodletting at big tech—you've noticed the script is suspiciously identical. Every CEO blames "AI." "We're leveraging AI to optimize our workforce." "Automation is allowing us to streamline headcount."
Jack Dorsey at Block was particularly shameless. He fired 4,000 people, claimed it was because of "AI tools," and watched his stock pop 24% in a week. The market loves this narrative. "AI efficiency" is Wall Street's favorite pornography right now—capital deployment without the messy burden of payroll.
But look closer at Block's math. They went from 5,400 employees pre-pandemic to over 10,000 during the boom. The 4,000 "AI-driven cuts" just brought them back to baseline. They didn't achieve magical productivity gains; they corrected their own hiring bloat. It was so clumsy that three weeks later, they quietly rehired some of those same engineers at 25% higher salaries because their core systems started breaking. As one comedian noted: "They accidentally sledgehammered the load-bearing walls."
A Harvard Business Review survey of 1,006 executives recently exposed the grift: 89% of these layoffs were planned months before AI became a credible factor. Fifty-nine percent admitted they used "AI" as cover for poor financial planning. It's not automation replacing humans. It's humans being sacrificed to pay for automation infrastructure, dressed up in futuristic language so the stock chart stays green.
The Iceberg: How Jobs Actually Die
But here's the part that actually keeps me up at night. The thirty thousand people at Oracle? They're the visible tip. The horror story isn't the mass layoff—it's the invisible drowning happening below the waterline.
MIT calls it the "Iceberg Project." I call it Attrition by Algorithm. While CEOs manufacture spectacle for Wall Street, the real job destruction is silent, polite, and mathematically brutal. It happens in three ways that never trigger HR alerts or viral LinkedIn posts:
The Replacement That Never Was
Picture this: A senior marketing analyst at a Fortune 500 retires after twenty years. It's a Tuesday. Historically, her boss would open a requisition that day—post the job, interview candidates, hire a junior replacement within sixty days.
In 2026, the boss looks at the budget, then looks at ChatGPT Enterprise. "Let's see if the remaining two team members can absorb Sarah's reporting duties using the new AI tools. We'll revisit the hire in Q3."
They absorb it. Of course they do. The AI writes the first drafts of the reports; the humans edit and verify. The Q3 review gets pushed to Q4. Then Q1 of next year. Eventually, the requisition simply expires. Sarah's job doesn't move to a robot. It evaporates into the atmosphere, distributed across tired humans and tireless algorithms. No announcement. No severance. Just a permanent ghost position on the org chart.
The Growth Mirage
I see this with our own clients. A wealth management firm hits a revenue spike—up 20% year-over-year. In 2019, they'd immediately hire ten new associates to handle the volume. More clients = more humans, simple algebra.
Today, they buy twenty Claude licenses and give their existing team "productivity targets." The revenue scales. The headcount stays flat. The CFO calls it "operational leverage." The CEO calls it "AI-enhanced efficiency." But walk through that office at 8 PM. The same people are there, looking gray, processing 40% more client portfolios with the help of agents that don't need sleep.
The new jobs—the ones that should have been created for the new graduates, the career switchers, the hungry twenty-somethings—never materialize. They exist in an alternate timeline that got cancelled.
The Consolidation Trap
The cruelest mechanism is the slow squeeze. A five-person data team loses one member to burnout. Then another to a better offer. The manager, under pressure to "do more with less," deploys an autonomous agent to handle the ETL pipelines and reporting dashboards.
Suddenly, three people are doing the work of five. Then two. Then one hyper-productive senior engineer with an army of Python scripts and agent orchestrations. The others didn't get fired—they got obsoleted through natural selection. Their positions aren't eliminated; they're just not rethought. When the last human leaves, the company realizes they don't need the humans at all anymore.
The Missing Generation
This is where the math gets terrifying in the long term. Stanford researchers just published findings that make my stomach hurt: entry-level software developer and call-center job postings have collapsed by 20% in the last eighteen months. Mid-level and senior roles remain stable. The junior layer—the apprenticeship tier—is being deleted.
Think about what this means structurally. If you're 22 right now, graduating with a CS degree, the "junior developer" rung on the ladder has been sawed off. The companies are using AI to automate the "grunt work"—the bug fixes, the testing, the documentation, the code reviews—that used to teach rookies how to become architects.
So where do the 30-year-old senior architects come from in 2034? If nobody is allowed to do the apprentice work, to learn through the tedious repetition that builds pattern recognition, the entire talent pipeline collapses. We're creating a "missing generation" of expertise. In ten years, we'll have brilliant AI tools operated by aging seniors with no one to mentor, and a chasm where the mid-career professionals should be.
It's not just cruel to the graduates. It's suicidal for the industry.
The Psychology of the Survivors
I want to talk about the people who still have their jobs for a moment, because their suffering is invisible too.
When you survive a layoff—or worse, when you survive the slow attrition where colleagues just... stop being replaced—you don't feel lucky. You feel haunted. Survivor's guilt mixes with acute anxiety. You know you're not safer; you're just later in the queue.
I've watched friends at big tech companies describe their new reality. The 9-to-5 is dead. They work 6 AM to 8 PM now, not because they're workaholics, but because the AI "efficiency" means their team of ten is now a team of six doing the work of twelve. They're managing algorithms that never sleep, which means they never fully clock out. The AI doesn't take vacations, so neither can they, really.
They call it "ghost work"—the human oversight that has to happen around the automation. Monitoring the agent outputs. Fixing the hallucinations. Apologizing to clients when the algorithm screws up. They're not replaced by the AI; they're trapped in a toxic marriage with it.
What To Do In the Rising Tide
The metaphor everyone uses is wrong. This isn't a tsunami hitting suddenly. It's a rising tide that comes up an inch every month. You don't drown dramatically; you just find yourself standing on shrinking sand, wondering when the water reached your chin.
If you're waiting for the 6:00 AM email to know you're in danger, you're watching the wrong horizon. The real threat is the slow realization that your department has quietly shed 30% of its headcount through "natural attrition" over three years, and you're now doing the work of three people while training the agent that will eventually absorb your specific function.
The only viable strategy is to become the Pilot, not the Passenger.
If you are the person who merely executes the work that an AI can now automate—the translation layer, the formatting, the standard analysis—you are standing on sand that is eroding in real-time.
If you are the person who directs the algorithms—who defines which problems are worth solving, who designs the workflows the agents execute, who owns the judgment calls when the AI hallucinates—you are building the boat.
In my own company, I've watched this split happen over the last three years. The people who were excellent at "doing"—writing the code, drafting the copy, crunching the numbers—have had to either migrate up to "deciding" (architecture, strategy, judgment) or they've quietly become obsolete despite still being on the payroll.
The work isn't disappearing. It's stratifying. The algorithmic layer is swallowing the middle—the competent executioners—while the top (vision, definition, ownership) and the bottom (physical presence, emotional labor, high-stakes accountability) remain stubbornly human.
The Uncomfortable Truth
Oracle's 30,000 didn't die because AI got smarter. They died because capital decided that infrastructure was more valuable than institutional knowledge, and "AI" was the perfect alibi to make liquidation look like progress.
But the slower death—the attrition by algorithm, the missing generation, the quiet evaporation of entry-level opportunity—that's the real structural transformation. That's the tide rising while we argue about whether the water is wet.
Don't watch for the email. Watch for the requisitions that never open. Watch for the teams that shrink by attrition. Watch for the moment when your job stops being about solving problems and starts being about babysitting the solver.
That's when you know you're already underwater.
— James, Mercury Technology Solutions, Tokyo, April 2026
