The Technology to Educate Every Unreached Child Already Exists | Craig Miles

For as long as I can remember, the conversation about education for the world’s unreached children has been about what is missing. Missing infrastructure. Missing teachers. Missing connectivity. Missing funding. The conversation has been framed, consistently and reasonably, as a problem of absence — a gap between what exists and what is needed that is simply too large and too expensive to close through conventional means.

That framing is now out of date.

In 2026, for the first time in history, the technology required to deliver a secondary-education-standard curriculum to every child on Earth — including the 273 million currently without any school access — is not missing. It exists. It is operational, or approaching operational status. It is affordable at scale. And it requires no new physical infrastructure in the communities it would serve.

This is not a prediction about what might be possible in ten years. It is a description of what is available right now, and what has changed in the last three years to make it so. Understanding what changed — and why the timing matters — is the starting point for understanding what ReachED is trying to do.

Three things had to be true simultaneously

The architecture for satellite-delivered education that ReachED is developing requires three components to be available concurrently. For most of the last two decades, at least one of them was not. By 2026, all three are.

The first is connectivity — a mechanism for delivering content to a device in a remote location with no terrestrial network coverage, at a cost that scales to the communities that need it most.

The second is intelligence — a system capable of delivering genuinely personalised, adaptive learning without a teacher present and without a continuous internet connection.

The third is reach — a device in the hands of the learner that can run both of the above.

Let me explain what has happened to each.

Connectivity: LEO direct-to-device satellite is now operational

On 22 April 2026, the FCC granted AST SpaceMobile commercial authorisation to provide direct-to-device broadband connectivity from space — the first time a LEO satellite operator has received full commercial approval to deliver broadband directly to unmodified smartphones using mobile network operator spectrum. Within days, Abel Avellan, AST SpaceMobile’s founder and CEO, described the company as “the only technology positioned to capture the massive direct to device broadband opportunity in full.”

That is a commercial statement. But behind it is a technical reality that has profound implications for education access.

Direct-to-device (D2D) LEO satellite means exactly what it says: a satellite in Low Earth Orbit communicates directly with a standard, unmodified smartphone — no dish, no ground terminal, no specialist hardware. The satellite acts as a cell tower in the sky. A phone that would ordinarily show no signal in a remote rural area connects to it as if connecting to a terrestrial base station.

This is possible because of 3GPP Release 17, published in 2022, which for the first time standardised Non-Terrestrial Network (NTN) integration with the 5G air interface. That standardisation means device manufacturers — Qualcomm, MediaTek, Samsung — can build NTN capability into standard chipsets. A phone manufactured today can, in principle, connect to a LEO satellite without any modification.

AST SpaceMobile is not the only operator. Starlink Direct to Cell — partnered with EE in the UK and T-Mobile in the US — provides SMS, voice and data to standard LTE devices. Lynk Global operates a carrier-agnostic direct-to-device service focused on developing markets. Virgin Media O2 announced a Starlink D2D service for UK customers in 2026. The competitive landscape is developing rapidly, which matters for pricing.

The implication for education access is this: a child in a village in rural Kenya, Sierra Leone, or Myanmar with a smartphone and no mobile coverage can, for the first time, receive data from space — without any infrastructure being built in or near her community. The connectivity gap that has kept 273 million children outside the reach of digital education is, technologically, closing.

Intelligence: edge AI runs offline on a mid-range smartphone

Connectivity alone is not education. A satellite link that delivers a static PDF curriculum is better than nothing but it is not the transformative intervention that changes life outcomes. What makes education genuinely effective — particularly for children learning without a teacher present — is personalisation: understanding where a child is in her learning journey, adapting the next explanation to address her specific misconceptions, sequencing concepts in the order that builds genuine understanding rather than surface familiarity.

Until very recently, that kind of adaptive, personalised learning required a continuous internet connection to access cloud-based AI. That was a fatal constraint for any education system designed to serve children in areas with no connectivity.

In 2026, that constraint has been removed.

Edge AI — the ability to run machine learning models directly on a device, without cloud connectivity — has matured from experimental feature to mainstream capability. Google’s LiteRT (formerly TensorFlow Lite), rebranded and significantly upgraded in March 2026, runs on over four billion devices through Google Play Services. Gemini Nano — Google’s smallest large language model — runs on-device on mid-range Android smartphones. Google AI Edge Eloquent, announced in 2026, brings offline generative AI to standard consumer devices.

The practical implication: a mid-range Android smartphone — a device available for under £100 at 2026 prices — can run a capable AI tutoring system completely offline. It can ask a child a question, assess her response, identify where her understanding is incomplete, offer a different explanation, and sequence the next concept — all without sending a single byte of data to a server.

This is the intelligence layer of the ReachED architecture. Not a replacement for a teacher — nothing is — but a system capable of providing structured, responsive, personalised learning support in the absence of one. And capable of doing so in the twelve languages spoken by the majority of the world’s unreached children, with audio narration for children with low literacy, on a device that fits in a pocket and runs on a battery that lasts a day.

Reach: four billion smartphones, many already in the right hands

The third component — the device itself — is where the ReachED model differs most sharply from previous attempts to address the education access gap.

Previous technology-for-education initiatives have typically required distributing new hardware — laptops, tablets, specialist devices — to the communities they serve. The cost, logistics, and sustainability of this approach have consistently limited scale. Devices break, replacements are expensive, supply chains into remote communities are unreliable, and the initiative depends on a continuous funding stream to maintain the hardware fleet.

The ReachED model does not distribute new hardware. It uses hardware that is already there.

There are an estimated four billion smartphones in active use globally. GSMA data consistently shows that device ownership runs significantly ahead of connectivity in developing markets — people in low-income communities have smartphones, often purchased second-hand, often running older Android versions, often with functional hardware but no useful connectivity. The device gap that previous initiatives tried to solve with hardware distribution does not exist in the way it once did.

The ReachED model targets Android devices from 2015 onwards — a generation of hardware that is now widely available as recycled or refurbished stock, often at near-zero cost in the communities that need them. The application is designed to run on devices with 2GB of RAM, under 50MB of storage footprint, with audio narration to compensate for low screen-reading literacy, and in offline mode as the primary operating state rather than the fallback.

Satellite connectivity — when it arrives via a LEO pass — synchronises the lesson content, uploads the child’s progress data for teacher oversight and credentialling, and downloads the next learning sequence. Between passes, the device works entirely offline. The child’s education does not depend on a continuous connection. It depends on a daily or twice-daily synchronisation window, which LEO constellation density already makes possible across most of the Earth’s surface.

Why this is different from everything that came before

I want to be precise about this, because the history of technology-for-development is littered with initiatives that were genuinely well-intentioned and genuinely failed. One Laptop Per Child. Various satellite internet projects that could not survive without subsidy. E-learning platforms that required reliable broadband to function. The failure modes are well documented and they all share a common characteristic: they required something that the communities they served could not provide for themselves.

The ReachED architecture is designed around what already exists in those communities, not what would need to be introduced. The device is already there. The satellite connectivity is overhead — not requiring permission from a local infrastructure owner, not requiring a cable to be laid or a tower to be built. The AI runs on the device, not in a data centre that needs a connection to reach. The curriculum is cached locally, not streamed from a server.

The dependency structure is different. And that difference is what makes this moment genuinely unlike what has come before.

What this means for the 273 million

None of this is to say that the problem is solved. The technology exists. The architecture is credible. The funding frameworks — Giga, Partner2Connect, the Global Partnership for Education, ECW — exist to support deployment. What remains is the work of systems integration: connecting the satellite operators, the device supply chains, the curriculum developers, the national education ministries, and the funding bodies into a coherent programme that can be piloted, validated, and scaled.

That is the work ReachED exists to do. Not to build new satellites — AST SpaceMobile and Starlink are doing that. Not to build new AI models — Google and others are doing that. Not to write new curricula — learning organisations like Learning Equality and Khan Academy have done that. But to design the system that connects all of these components into something that actually reaches a child in a village in Sierra Leone and gives her the education she should have had from birth.

The technology is ready. The question now is whether the will — institutional, political, financial — matches the moment.

I believe it does. And I am building the evidence for that belief one conversation, one partnership, and one pilot at a time.

If you work in satellite operations, international development, EdTech, curriculum development, or development finance and want to understand more about the ReachED model — or if you represent an organisation that could contribute to making it real — get in touch or connect with me on LinkedIn.

Craig Miles is a wireless communications engineer, TEDx Brayford Pool 2023 speaker, and founder of the ReachED Foundation — an initiative using LEO direct-to-device satellite connectivity, offline edge AI, and recycled smartphones to deliver secondary-education-standard learning to the 273 million children globally without school access. BSc · PGCE · QTS · Former ILS Engineer, Airbus Defence and Space. About Craig →