Generated by Codex with GPT 5.6 Sol XHigh

Techmeme surfaced The New York Times investigation “Kenyans Did College Students’ Homework for Years. Then A.I. Arrived.” Reported from Nairobi, it documents what may be one of the clearest labor-market shocks from generative AI so far: a large, informal export industry built around writing assignments for overseas students has shrunk to almost nothing.

The trade was ethically compromised because its product was academic cheating. But that does not make its collapse economically trivial. For thousands of educated Kenyans, essay writing was a rare way to turn English fluency and research skills into earnings far above local professional wages. Its rise and fall show how quickly AI can erase an entire layer of remote knowledge work when buyers value only the finished text and have little attachment to the people producing it.

An unlikely ladder into the global economy

Teresios Bundi entered the business as a university student in Nairobi in 2011. His first job—two pages about a fruit he had never heard of—paid \$7 for three hours of work. He estimates that he produced more than 2,500 essays over the next 12 years, sometimes writing three a day across subjects ranging from medicine to engineering.

After graduating in public health in 2015, Bundi chose essay writing over the career for which he had trained. He says it paid at least five times as much. He eventually charged \$40 to \$70 per paper, hired other students and rented a house equipped with desks, fast internet and places for exhausted writers to sleep. The money helped support his parents, who had borrowed to buy his college laptop.

Bundi’s operation was part of something much larger. Researchers estimated that at least 40,000 people in Nairobi were being paid to do other people’s homework at the industry’s peak early this decade. The work flourished in a country where more than 100,000 students graduate from universities each year, salaried positions are scarce, roughly 80 percent of employment is informal and youth unemployment exceeds 25 percent.

Cheap high-speed internet made the wage difference between Kenya and wealthier countries tradable. Students abroad wanted completed assignments; platforms and brokers could route those assignments to educated workers willing to accept lower prices. The arrangement was exploitative and deceptive, but it also functioned as a labor-market escape hatch. Successful writers bought cars and smartphones, supported relatives and built careers in a place where their degrees often led nowhere.

AI removed the reason to hire a person

ChatGPT’s arrival in 2022 broke that market with unusual speed. Overseas students who had treated essays as interchangeable outputs could now generate drafts almost instantly and at negligible marginal cost. Assignments dwindled, rates fell and businesses such as Bundi’s closed.

The article places essay writing inside a broader contraction of Kenyan online work. Transcription jobs had already begun disappearing as speech-recognition software improved. Content-moderation and data-labeling operations also pulled back amid disputes over pay and working conditions, while Scale AI discontinued some Kenyan operations. Workers who tried moving from one form of digital piecework to another repeatedly encountered automation or unstable outsourcing arrangements.

One benchmark cited by the article illustrates the pace, though not the full labor-market effect. In a test from Scale AI and the Center for AI Safety involving remote freelance tasks such as product design and data analysis, leading models completed 2.5 percent of tasks last October and 16 percent by July. That result does not mean AI can replace 16 percent of freelancers: benchmark tasks, production workflows and economically useful work are not interchangeable. It does show that the boundary of automatable online work can move substantially in less than a year.

Nor has human labor vanished completely. The remaining essay work increasingly goes to “humanizers,” who edit AI-generated papers so they are less likely to trigger university detection systems. One writer profiled in the article says AI lets her complete more assignments and that business remains strong. This is augmentation for the individual who retains the work, but it also concentrates the same demand among fewer people. Higher output per worker can mean a smaller market overall.

The development strategy met its weakest point

The timing is especially revealing. In 2016, Kenya began training graduates to navigate online freelance platforms. Its 2022 National Digital Masterplan again presented outsourcing and online work as a route to economic advancement—the same year ChatGPT was released.

That strategy depended on a fragile premise: that routine digital tasks would remain valuable long enough for workers and the country to move up the economic ladder. Yet the first tasks to travel across borders were often precisely those that were easiest to specify, measure and automate. The distance that made Kenyan workers inexpensive also left them with little bargaining power, few protections and almost no claim on the customer relationship when a cheaper substitute arrived.

Bundi now works for a German development organization helping young Kenyans find opportunities in the digital economy. His own career illustrates the cost of a sudden transition. Former public-health classmates accumulated a decade of experience while he built a business that no longer exists; returning to that profession could mean starting again at entry level.

The article’s deeper warning is therefore not simply that AI replaces jobs. It is that automation can hit geographically concentrated, largely invisible labor markets before headline employment statistics in wealthy countries register much change. Workers at the bottom of global digital supply chains absorb the shock first, even when they helped create the cheap, scalable services that made the online economy work.

Reskilling remains necessary, but it cannot by itself create demand for durable work. A more resilient digital-development strategy would have to help workers build specialized expertise, local institutions, customer relationships and ownership—not only train them for the next standardized task sold through a platform. Kenya’s essay writers occupied a morally awkward industry, but their experience is a clean warning: when a livelihood rests on being the cheapest human in a transaction, a machine does not need to outperform every professional to make that livelihood disappear.