From Internet to Agentnet

For most of modern history, Africa received the future late.

The steam engine. The telegraph. The electric grid. The computer. Each one arrived years, sometimes decades, after it had already rewired another part of the world. We imported finished things, learned to operate them, and were told to be grateful.

I grew up believing that was simply how it worked. That somewhere far away, people invented tomorrow, and the rest of us waited for it to be shipped.

It isn’t how it works anymore. And the next shift, the one already underway, could close the gap for good.

01The waiting room

Think about what innovation meant during the industrial age. A machine was built in Manchester or Pittsburgh. Then it had to be patented, manufactured, financed, shipped, installed and maintained. Every step took time. Every step had a gatekeeper.

If you lived in Accra, Lagos or Nairobi, you didn’t just need the idea. You needed the factory, the ship, the license, the spare parts and the engineer who knew how to fix it. Progress travelled at the speed of cargo.

So we waited. Not because we lacked intelligence. Because we lacked access.

Africa was never short of minds. It was short of access. The internet changed the second part.

02The internet ended the wait

Then the cables came ashore and the phones got cheap.

Something remarkable happened. We didn’t climb the ladder one rung at a time. We skipped rungs. Most homes never got a landline; they went straight to mobile. Kenya’s M-Pesa put money on basic phones in 2007, years before most of the rich world took mobile payments seriously.

For the first time, a teenager in Ho could learn the same thing, on the same day, as a teenager in California. Real time. Same tutorial, same documentation, same open-source code.

But here is the uncomfortable truth. Access is not ownership. We got connected to the internet, and mostly we became its users. The platforms, the standards and the profits were built somewhere else. We consumed the future in real time, but we still didn’t write much of it.

That is the gap I care about. And it is about to become the most important gap in the world.

03From internet to agentnet

The internet connected information. You searched, you read, you decided, you clicked.

What is coming connects action.

AI agents don’t just answer questions. They read the contract, compare the suppliers, file the claim, book the shipment, reconcile the books, watch the network and flag the anomaly. They act on our behalf, all day, at a scale no team of humans could match. Millions of them, talking to each other, negotiating, transacting.

I call this the agentnet. A web not of pages, but of agents.

And this time, the distance between where the future is made and where it is used can be zero. An agent doesn’t need to be shipped. It doesn’t need a port, a factory or a distributor. It needs compute, data, a problem worth solving and someone brave enough to build it.

The industrial age made us wait for the future. The internet let us watch it in real time. The agentnet is the first wave we can build at the same time as everyone else.

I have said it before and I will keep saying it: AI’s current best is the least it will ever be. Every month the models get better, cheaper and more capable. Whoever builds the agents for a market will shape how that market works for the next twenty years.

The question is no longer whether Africa will be part of the agentnet. It is whether we will be its users again, or its architects.

04Our problems are our unfair advantage

People love to list Africa’s problems. Unreliable power. Patchy infrastructure. Fragmented markets. Weak institutions in places. Currency volatility. All real. I have lived inside most of them.

But look at that list again through a builder’s eyes. Every problem is a specification. Every gap is a market nobody has served properly. Every broken system is one you don’t have to fight to replace, because it was never really built.

Where systems are missing, you don’t have to migrate them. You get to design them. Native. Mobile-first. Agent-first. From day one.

And consider what sits underneath those problems:

• The youngest population on earth. The median age on the continent is under twenty. In a few decades, a large share of the world’s working-age people will be African.
• The materials of the future. Africa holds a significant share of the world’s critical minerals, including most of the cobalt that goes into batteries. The hardware of the AI age is partly made of African ground.
• One of the largest single markets ever designed. The African Continental Free Trade Area brings dozens of countries into one trading bloc on paper. The paperwork exists. The software that makes it real mostly doesn’t, yet.
• A strategic position. Between the Americas, Europe, the Middle East and Asia, in time zones that overlap with all of them, with new undersea cables landing on both coasts.

Problems, plus people, plus resources, plus position. That isn’t a charity case. That is the strongest founder market I know of.

Every problem is a specification. Every gap is a market nobody has served. Africa isn’t late. It’s early.

05The surprise is in the informal economy

Here is the part most people skip.

Walk through Makola Market in Accra, Balogun in Lagos or Gikomba in Nairobi. Fortunes change hands every day in cash, in mobile money, in susu savings groups, in credit extended on nothing but a name and a handshake. No balance sheet sees it. No credit bureau scores it. Most banks never reach it.

Economists call it the informal sector. I call it the largest untapped economy on earth, hiding in plain sight.

Think about what that means. The woman who has run the same stall for twenty years, paid every supplier and never missed a susu contribution has a better repayment record than many people who get loans in London. It just isn’t written anywhere a system can read.

The goal is not to drag these people into somebody else’s paperwork. It is to connect them. Make their reputation portable. Make payments seamless across networks and borders. Make savings safe, credit fair and insurance possible. The unbanked and the underbanked are not a problem to be solved. They are customers who have never been properly served.

Africa’s wealth isn’t missing. It’s unconnected. That is where the surprise is.

And this is exactly where the agentnet meets the street. An agent that keeps a trader’s books by voice, in her own language. One that turns years of mobile-money history into a credit score. One that finds her a cheaper supplier two towns away, or reconciles a cooperative’s accounts every night without an accountant.

But none of it works if it only runs in a data centre on another continent. The informal economy runs on cheap phones, patchy signal and power that drops when it likes. That is why I care so much about edge intelligence: running the model on the device, or close to it, so it works offline, answers instantly, costs less and keeps people’s data with them. It is why I am building Kalon.

We need more young minds here. Not only in the glamorous corners of AI, but in the market, the trotro station and the farm gate, where the problems are real and the rewards are enormous. The next great African companies will not be the ones that ignored the informal economy. They will be the ones that finally connected it.

06The age of data

Every AI revolution runs on one raw material. Not chips. Not code. Data.

And Africa is sitting on one of the largest untapped reserves of it on earth. Thousands of languages. Centuries of history, much of it spoken rather than written. Markets, farms, rivers and weather patterns that no model has ever properly seen.

For decades we have paid the price for not having it written down.

We still struggle to predict the rains a farmer plants by. To forecast the flood or the drought before it hits, and to plan the recovery after. To see the next trend coming. Not because we lack intelligence. Because we lack data: long, clean, well-documented history. And much of what does exist sits on shelves, in archives and in silos that never talk to each other.

That last number should keep every African technologist up at night.

The open web is the main diet of today’s large language models. In Common Crawl’s August 2026 crawl, more than two billion pages, English makes up 40.5%. Swahili, spoken by more than 200 million people, makes up 0.0117%. Icelandic, spoken by fewer than 400,000, has almost three times as much. Put twenty-one major African languages together, Hausa, Yoruba, Amharic, Zulu, Somali, Igbo and more, and they still come to about 0.05%.

A rounding error.

Our languages are spoken by hundreds of millions of people. In the data that trains AI, they are a rounding error.

That is why a model that writes English with near perfection still stumbles over Ewe, Twi, Yoruba or Kinyarwanda. There has been real progress, and brilliant people are working on it. But we are not at par yet. And we won’t be until our languages are in the data at a scale that matches the people who speak them.

So the work is twofold. Use our own data to build and train our own models and systems. And contribute it, on our terms, so the models built by the hyperscalers represent us fairly too. Both, so that AI ends up serving all of humanity, not most of it.

This is what I am building: a data infrastructure layer for Africa. One that documents and preserves our history before it is lost. One that turns archives, records, voices and languages into data that is clean, owned and usable. So that we don’t just import the next revolution. We contextualise it. We understand our past through data, read our own trends, and pass our culture and our way of life to the next generation in a form that crosses every border.

And I am not doing it alone. I’m building it with some of the brightest minds I know, backed by some of the best investors in the world.

The industrial age ran on coal. The internet ran on connectivity. The agentnet runs on data. The next frontier is the age of data, and it is waiting for us.

07Brilliant minds, no borders

Here is what excites me most. The tools that once required a lab in Silicon Valley now fit in a laptop, a phone and an internet connection.

A gifted engineer in Tamale can train a model, deploy an agent and sell it to a customer in Singapore before lunch. A student in Kigali can contribute to the same open-source project as an engineer at the biggest companies on earth, and be judged only on the quality of the code.

Solutions built for the hardest conditions travel well. If your product works on a cheap phone, on a weak signal, on a power grid that drops twice a day, in five languages, it will work almost anywhere. Constraint is a forge.

That is why I don’t think of African builders as building “for Africa”. We are building from Africa, for the world. The boundary is a starting point, not a ceiling.

08What I’m building, and why

Everything above isn’t theory to me. It is the reason I build.

I have worked across the full stack of this shift: hardware and software, robotics, blockchain, machine-learning models, and cybersecurity. Not because I couldn’t pick a lane, but because the agentnet doesn’t respect lanes. An agent touches the model, the device, the network, the ledger and the person, all at once. You need to understand all of them to build one you can trust.

That thinking runs through the work I’m doing now:

• OpenAI4Africa.orgOpening frontier AI to the builders, students and institutions who will shape how it is used here.
• Sisu AIAI with the grit the name promises, built for real conditions rather than ideal ones.
• KalonEdge intelligence. Putting AI where the data and the people actually are, on the device and offline when it has to be, not only in a distant data centre.
• Skylake.aiPart of the same bet: intelligence that is useful, accountable and built to scale beyond any one border, bringing the best ideas to life on the agentnet.

And many more that I’ll shed light on here, in time.

Different products. One thesis. The agentnet is coming, and the people closest to the hardest problems should be among those who build it.

09The part nobody wants to talk about

Here is the warning that comes with all this optimism.

An agent that can act on your behalf can also be tricked into acting against you. A hidden instruction in an email. A poisoned document. A compromised tool in the supply chain. When software only showed you information, a mistake meant a wrong answer. When software acts, a mistake means money moved, data leaked or a system switched off.

The agentnet multiplies capability. It also multiplies attack surface.

And the systems at stake are not abstract. Mobile money rails that millions of families depend on. Power grids. Hospitals. Ports. Elections. In much of Africa these are being digitised right now, in the same years that agents arrive. We have one chance to build them secure from the start instead of patching them after the first disaster.

Disaster looms if AI advances faster than our ethical decisions. Agents need security by design and humans in the loop.

That is why my next frontier is cybersecurity for the age of agents. Defences built for critical infrastructure. Systems where agents prove what they are allowed to do before they do it. Where a human stays in the loop for every decision that matters. Where power stays with the user, not with whoever controls the model.

Security is not the brake on the agentnet. It is what makes it possible to trust it at all.

10An invitation

For two centuries the story was: the future is made elsewhere, and Africa waits.

The internet started to rewrite that story. The agentnet can finish it. But only if we build. Not someday. Not after someone else proves it works. Now, while the foundations are still wet.

So this is an open invitation. If you are a builder, a researcher, an investor, an institution or a government that wants to build secure, human-centred agent systems, from Africa, for the world, I want to talk to you.

Don’t just follow or chase your dreams. Hunt them.

Let’s code a new era. #Be_Scientific