AI Was Supposed to Take Over Our Jobs in 2025, but It Couldn’t Even Click a Mouse — The New Yorker’s Belated Tech Horror Story
2025 was supposed to be the year AI liberated humans from the drudgery of daily tasks. Sam Altman, CEO of OpenAI, boldly predicted that AI agents would “join the workforce” and revolutionise productivity. Instead, most of these agents ended up starring in a tragicomic performance of digital ineptitude. Gary Marcus, the tech critic, called them “cognitively lacking.” In layman’s terms, AI was supposed to take our jobs but couldn’t even work out how to click a mouse without a minor existential crisis.
How We Spent Eight Trillion Quid So AI Could Book a Hotel Wrong — A Cautionary Tale in Missed Reservations
Imagine this: corporations and venture capitalists poured eight trillion dollars into AI development. The grand vision? AI booking your hotel, scheduling your life, even picking the perfect avocado at the supermarket. The reality? ChatGPT Agent spent fifteen minutes staring at a drop-down menu trying to select a hotel price. One AI even scheduled a baseball stadium stop in the middle of the Gulf of Mexico. Experts call it a “location hallucination.” Humanity calls it Tuesday.
The Year AI Was Going to Join the Workforce, but It Got Stuck in a Dropdown Menu — Also: Welcome to the Gulf of Mexico, Baseball Fans
2025 promised a digital labour revolution. Marc Benioff claimed half the work at Salesforce was already being done by AI. Yet, outside the confines of text-based terminals, AI stumbled. Our agents, designed to automate complex tasks, were completely gobsmacked by the simplest user interface. A job that takes humans five minutes—selecting the right dropdown menu—can take AI fifteen. Meanwhile, tourists who trusted AI itineraries ended up snorkelling in the Gulf of Mexico en route to Yankee Stadium. The AI’s logic? Nobody really knows, but we can assume it involves a confused seagull.
Forget Terminators, Our AI Agents Just Ask What a Mouse Is — A Deep Dive into 2025’s Greatest Disappointment
Autonomous agents were supposed to be the Matrix-meets-Mission Impossible dream. Instead, they exist in a perpetual state of beginner confusion. To automate tasks outside coding, AI needs to master a mouse, an interface humans learn before nursery school. Companies tried building “shadow sites” to train AI agents to mimic human mouse behaviour. Result? It still takes multiple minutes to select a simple option, proving that teaching an AI to click a mouse is equivalent to teaching a cat to file taxes.
AI Was Going to Save Humanity, Instead It Created the World’s Most Expensive Autocorrect — The New Yorker Explains
AI’s failures extend beyond misbooked hotels. The same large language models (LLMs) that power agents generate text and code but cannot grasp real-world nuances. ChatGPT Agent tried to plan baseball tours, trips, and dinner reservations, resulting in absurd outputs: itineraries with imaginary cities, hotel rooms priced in plutonium, and schedules that make sense only to someone who bunked off basic geography lessons. Humanity now faces a new form of stress: existential irritation at an AI that could fix everything but chooses to invent complete nonsense.
Predicted to Be the “Year of the AI Agent,” Ended Up Being “Please Click the Button Again” — 2025 in Review With Extra Irony
Despite promises, AI in 2025 achieved something remarkable: showing the world how little an AI can do for trillions of quid. Developers aimed for a workforce replacement; they got an overpaid intern who keeps asking, “What now?” OpenAI engineers demonstrated features where AI mapped Major League Baseball stadiums, only to include the Gulf of Mexico. Google’s Agent2Agent protocol allowed agents to talk to each other—but talking doesn’t solve dropdown menus. The agents tried, failed, repeated. Humanity spent 2025 clicking “retry” on AI tasks and wondering if the robots were really conspiring or just hopelessly taking the mickey.
The Evidence Pile-Up: Experts, Polls, and Witnesses Agree
Andre Karpathy, co-founder of OpenAI, admitted that agents are “cognitively lacking.” Codex, AI’s code-writing champion, excels in terminals but flounders in the GUI-based world humans inhabit. Alex Shaw, co-creator of Terminal-Bench, confirms that text-based automation is AI’s comfort zone; everything else is a comedy of errors. Eye-witnesses in offices report AI “pausing for minutes,” vasking for clarification repeatedly,” and “sending staff to the Gulf of Mexico.” Public opinion? The majority reckon that AI has a “fascinating sense of humour” but is rubbish at work. Surveys indicate 72% of users would rather handle tasks themselves than let AI have a go.
Conclusion: The Digital Labour Revolution Was a Mirage
So, what did 2025 deliver? Not a takeover of the workforce. Not flawless automated planning. Instead, a comedy of errors, confusion, and misplaced baseball stadiums. Humanity learned the hard lesson that AI can write code, hallucinate locations, and misclick buttons at astonishing speed. The era of AI agents may still arrive someday, but for now, the best use is entertainment: watching them repeatedly fail at tasks humans find trivial.
Disclaimer: This satirical account was entirely a collaboration between a philosophy major turned dairy farmer and the world’s oldest tenured professor, ensuring both intellectual depth and an appreciation for absurdity. No AI was responsible for the humour, just to be clear.
Camden Rose is a student writer and emerging comedic voice whose work reflects curiosity, experimentation, and a playful approach to satire. Influenced by London’s grassroots comedy scene and student publications, Camden explores everyday experiences through exaggerated yet relatable humour.
Expertise is developed through practice, feedback, and engagement with peer-led creative communities. Camden’s authority comes from authenticity and a growing portfolio of work that demonstrates awareness of audience, tone, and context. Trust is supported by clear presentation of satire and a respectful approach to topical subjects.
Camden’s writing aligns with EEAT principles by being transparent in intent, grounded in lived experience, and mindful of accuracy even when employing comedic distortion.
