Long before the age of multi-billion-dollar AI companies promising to disrupt the field of software development, Wendy Liu was learning to code the hard way. As a child in the mid-2000s with unmonitored access to the family computer, she taught herself to build websites from scratch using a basic text editor. The results were never as polished as she imagined, but the painstaking hours of debugging and poring over documentation for projects she later abandoned never felt wasted. She was learning a craft, and in the process discovered a love for a certain way of thinking. That journey carried her through a four-year computer science degree and multiple software development jobs. Liu tells a similar story about becoming a writer: her initial desire to write about the tech industry came from frustration with what she saw as a gap between her own critical understanding of Silicon Valley and the credulous optimism of the discourse. She has since published many thousands of words, with countless more left on the cutting-room floor — but those discarded words, she argues, were never wasted. They were the byproduct of thinking.
Learning to Think
Liu, a writer based in San Francisco and author of the 2020 book Abolish Silicon Valley: How to Liberate Technology from Capitalism, left the tech industry to pursue a master’s degree in inequality at the London School of Economics. She is now working on a novel. Her educational journey was not efficient; she toiled away solo, following her own made-up syllabus, motivated by curiosity and a desire to understand. For her, the writing process is transformational: you start with one idea and end up somewhere different. Writing is more than outputting words, she says — it is about discovering your values and convincing yourself they are worth fighting for.
In both coding and writing, Liu now feels as if she has taken the last helicopter out of Saigon. The fields have been revolutionised by large language models (LLMs). Software development has been deskilled by “vibe-coding”, where AI tools are prompted using natural, conversational language to generate code. Tech companies previously known as great employers are using AI as an excuse for large-scale redundancies. In the UK, job adverts for high-exposure AI occupations fell significantly between 2022 and 2025, with junior developer openings notably in decline — even as salaries for software developers surge, driven by demand for AI skills. Meanwhile, writing has been overwhelmed by AI slop to the point where people have become afraid to use stylistic elements such as em dashes, now seen as a hallmark of AI-generated text. The freelance writing market has seen a sharp drop in demand for certain types of content, with AI replacing human writers for tasks like blog posts and SEO filler. Some copywriters report their income has halved. More than half of UK novelists believe AI will replace their work, and many have lost income and express concerns about copyright infringement.
The Risks of Cognitive Offloading
Liu says she avoids using AI as much as possible. “I am wary of cognitive offloading, as tempting as it can be to turn over certain tasks to a machine so I don’t have to think so much,” she writes. “Thinking is the point. I don’t want to get into the habit of avoiding it purely for the sake of convenience.” Her concern is grounded in research: studies indicate that even a few minutes of AI chatbot usage can have a negative impact on cognition. A growing body of evidence shows a negative correlation between frequent AI tool usage and critical thinking skills, largely due to cognitive offloading. Younger participants in one study showed higher dependence on AI tools and lower critical thinking scores. The reliance can erode essential skills such as memory retention and analytical thinking, potentially leading to an atrophy of internal cognitive abilities.
Liu worries in particular about young people coming of age during the AI boom. She fears the mystique around AI teaches them to see technology as a black box — something foisted upon them, managed by opaque corporations over which they have no control. “What does that do to their relationship with technology, if they see it as something that simply happens to them, whose inner workings cannot be fathomed, much less changed?” she asks. The concept of “cognitive sovereignty” — the authority to direct and take responsibility for one’s own reasoning process — has emerged as a counterweight. Liu argues that huge AI companies are aiming to make intelligence a “utility”, effectively privatising thought. Limiting use of this technology, she says, may be a way to protect cognitive sovereignty: preserving our ability to think, keeping our brains active rather than outsourcing every decision to probabilistic software.
On a collective level, she sees this as a political matter — a way to combat dependence on AI companies raising unprecedented amounts of money with the goal of inserting themselves into every facet of society. The world, she writes, is in the midst of an AI bubble. Trillions of dollars are projected to be spent on data centres. Corporations posting record revenues are initiating mass redundancies to invest more in AI, while remaining employees feel pressure to maximise their own use of AI to stay competitive. People are using AI to write their wedding vows and even falling in love with AI companions, with some research suggesting the brain struggles to distinguish between real and simulated social interactions. In a short span of time, it has become terrifyingly normalised.
The environmental consequences are stark. Revised UK government estimates suggest AI data centres could emit up to 123 million tonnes of carbon dioxide over the next ten years — a substantial increase from earlier projections. Data centres require vast amounts of electricity, often generated by fossil fuels. Water demand is also a growing concern, with projections indicating billions of cubic metres consumed annually for cooling and electricity generation, threatening water security in the UK and globally. Training a single AI model can emit as much carbon as five cars over their lifetimes.
In the UK financial sector, over 75% of firms are now using AI for tasks including processing claims and assessing creditworthiness. But regulators and the government have been criticised for a “wait-and-see” approach, with no AI-specific laws in place. Concerns include increased cybersecurity risks, over-reliance on a few US tech companies, and amplification of “herd behaviour” that could risk a financial crisis. Britons are generally comfortable with AI for operational efficiency such as fraud detection, but only 19% are comfortable receiving AI-generated financial advice, and just 15% believe AI services act in their best interests.
Liu acknowledges that her refusal to engage with AI feels heretical in this atmosphere. “I may believe in my heart that I am right, but I have to live every day surrounded by apparent evidence of my wrongness, the AI billboards looming over me like monuments,” she writes. She knows she is a less efficient coder and writer because she has not learned the latest tooling. In the time it took her to write and rewrite her essay, she notes, she could have prompt-engineered hundreds of books. But she argues that in a world where efficiency and convenience have become vehicles for the advancement of corporate greed, inconvenience and inefficiency may simply be the cost of preserving humanity and building character. She is taking a path she believes will help her become the kind of person she wants to be: someone deeply rooted in the world, who moves with intention and integrity. The trade-offs — including the money she could be making at an AI startup — are worth it, she says. She knows what values she wants to fight for.
