In 2005, Nicholas Negroponte stood up at the World Economic Forum in Davos and promised a laptop for every child on Earth, priced at $100. Two decades later, the same utopian pitch is being made about artificial intelligence in classrooms — and the stakes, and the skepticism, are identical.
The $100 Laptop Dream: What OLPC Set Out to Do
When the One Laptop Per Child initiative launched, it carried a seductive logic: if you gave every child in the developing world a connected computer, you would unlock education, economic mobility, and a generation of problem-solvers. The XO-1, the rugged green-and-white machine designed for the project, was engineered for dust, heat, and sunlight-readable screens. It was meant to be handed out like textbooks — not sold at retail, not rolled out in pilot classrooms, but distributed en masse to children who had never touched a keyboard.
The ambition was intoxicating precisely because it was simple. Silicon Valley executives and philanthropists believed hardware was the binding constraint. Remove it, and learning would follow. Negroponte framed the project as an educational moonshot, not a gadget launch, and the press largely treated it that way. Governments from Peru to Uruguay to Rwanda signed on. The rhetoric was global, but the plan was hardware-first.
That framing is worth sitting with, because it recurs every time a new technology promises to fix schooling. The device was never the point — the point was a theory of change in which access to a tool, by itself, produces outcomes. As the reported history of the initiative shows, the project's founders felt that getting a computer into every child's hands was the way to fix everything.
Why the OLPC Initiative Fell Short of Its Promise
The numbers that ground the digital divide today were already visible during the OLPC era, and they have not moved as much as the optimism implied. UNESCO and World Bank data consistently show that a large share of children in low- and middle-income countries still lack reliable internet access or a device at home. Connectivity is not a switch you flip; it requires electricity, maintenance, repair ecosystems, teacher training, and content in local languages — none of which ship in a box.
Read next Laika's Wildwood: Stop-Motion Fantasy at TIFF 2026Independent researchers who studied OLPC deployments found a pattern that should sound familiar. In Peru, which received hundreds of thousands of XO-1 units in one of the largest rollouts, studies documented limited measurable gains in math and reading. In Uruguay, the Plan Ceibal program got machines into children's hands at scale, but the learning effects were mixed and heavily dependent on how teachers used the devices. In Nigeria and elsewhere, pilots ran into familiar obstacles: broken units, missing chargers, insufficient bandwidth, and teachers who had never been trained to integrate the hardware into instruction.
The market rejected the core promise too. At scale, the "$100 laptop" cost closer to $200, and low-cost commercial netbooks and later tablets undercut the philanthropic model. Governments that had signed letters of intent found the economics harder to justify. The Verge's reporting on why the initiative never stood a chance captures the essential failure: the project treated a pedagogical and political problem as an engineering one.
That is the central lesson. A device delivered is not a device used. A device used is not a learning outcome achieved.
AI Is Making the Same Grand Bet on Education
Now consider the current wave. Districts and ministries are being sold AI tutors, adaptive learning platforms, and chatbots that promise personalized instruction for every student at a fraction of the cost of human tutoring. The pitch is structurally identical to OLPC's: a technology is framed as the lever that finally closes achievement gaps, and access to it becomes the goal.
The rhetoric has shifted from hardware to software, but the underlying bet has not. Where Negroponte promised a laptop per child, today's vendors promise an AI tutor per child — and the implied theory of change remains access-first. Get the tool to the student, and the learning follows.
Education researchers who lived through the OLPC era are watching with a mixture of recognition and unease. Scholars at the Brookings Institution and critics associated with MIT Media Lab have long argued that technology in education succeeds or fails based on pedagogy, teacher capacity, and institutional support — not on the sophistication of the tool. An AI tutor with no curriculum alignment, no trained teacher to supervise it, and no reliable internet connection is the XO-1 in a different costume.
What History Teaches Us About Tech-First Education Reform
The recurring pattern across decades of ed-tech reform is what policy analysts call the "access trap": measuring success by deployment rather than by learning. Peru could count laptops distributed. Uruguay could count students reached. Those metrics looked like victory in press releases and failure in independent assessments.
There are concrete lessons buried in those deployments. Where devices were paired with sustained teacher professional development, local-language content, and reliable infrastructure, outcomes improved modestly. Where they were dropped into classrooms as standalone solutions, outcomes were flat or negative. The variable was never the device. It was the system around it.
OLPC also underestimated the maintenance burden. Laptops break. Batteries die. Chargers disappear. In regions without repair supply chains, a broken laptop is a permanent loss. AI systems carry a comparable hidden cost: they require connectivity, electricity, data, content moderation, and ongoing technical support — all of which are scarce exactly where the need is greatest.
And then there is the problem of evidence. Rigorous evaluations of ed-tech interventions repeatedly find that effects are small, context-dependent, and hard to replicate at scale. That is not an argument against technology. It is an argument against treating technology as a shortcut past the hard work of building educational capacity.
Can AI Succeed Where the $100 Laptop Could Not?
There are genuine differences this time. AI can adapt to a student's level in real time, provide immediate feedback, and operate in multiple languages without a human translator. In well-resourced settings with strong connectivity and trained teachers, AI tutoring has shown promise in narrow domains like early literacy and math practice. Those are real advantages, not marketing.
But the advantages only materialize under conditions that OLPC's target markets lacked. If a school has no reliable electricity or internet, an AI tutor is unusable. If teachers are not trained to integrate it, it becomes a distraction. If the content is not aligned to the local curriculum, it is irrelevant. If the system is proprietary and expensive, it recreates the cost problem that killed the $100 laptop.
The honest answer to whether AI can succeed is: it depends — on the same things it always depended on. Technology can amplify good instruction. It rarely substitutes for it. The OLPC experience suggests that when reformers skip the institutional groundwork and lead with the tool, the tool becomes the scapegoat for a failure of planning.
The Real Question: Who Gets to Define Educational Progress?
The deepest issue raised by OLPC was never technical. It was philosophical. Who decides what counts as progress in education? Is it the number of devices delivered, the number of students logged in, the amount of time spent on a platform — or is it whether children actually learn to read, reason, and think?
Two decades after Davos, the metrics have changed but the question has not. AI vendors will report engagement, usage, and completion rates. Those are deployment metrics dressed as learning metrics. The OLPC era taught us that such numbers can be impressive and meaningless at the same time.
The lesson is not that technology has no place in education. It is that the order matters. Pedagogy first, institutions second, technology third. Get that sequence right, and a laptop — or an AI tutor — can help. Get it wrong, and we will be writing the same postmortem in 2046, about a tool that promised to save education and delivered a warehouse of unused hardware instead.
Source: The Verge
