The work, told by the people it happened to
Not summaries of what we intended. These are accounts of what actually changed in a specific classroom, for a specific person, and what it cost to get there — including the parts that didn't go to plan.
Eighteen years in, and nobody had ever asked her what she needed
Akinyi teaches fifty-four children in a room built for thirty. She has done this, in one form or another, for eighteen years. She knows which of her students walked an hour to get there. She knows which ones eat at school because they don't eat at home. She has never had a projector, and for most of those years she has not had reliable electricity.
When we came to run a training week, she sat at the back with her arms folded and said nothing for the first two hours.
It would be easy to read that as hostility. It wasn't. Akinyi had watched programs arrive before. A government initiative with a launch event and no second year. An organization that delivered equipment nobody was trained to use, then stopped answering emails. Every one of them had asked her to be enthusiastic, and every one of them had left her to explain to her students why the new thing had stopped working.
Halfway through the morning she asked her only question, and she asked it flatly, without softening it: will this replace me?
“Everybody comes here to fix the school. Nobody comes here to ask the teacher what is actually wrong.”
— Akinyi, day oneThe answer we gave her was the honest one. AI will change teaching. Some of what she does now will be done differently in ten years. But the person who can tell the difference between a child who understands and a child who is pretending, who knows which student's home has fallen apart this month — that person is not replaceable by any system we know how to build. AI is a tool. She is the teacher.
She unfolded her arms. She wasn't convinced. She was listening.
The turn came on the third day. The session covered using digital tools to identify which students are quietly slipping behind — the precise thing that keeps her awake, because in a room of fifty-four, a struggling child can go unnoticed for a whole term. She leaned forward. She started asking questions that weren't skeptical. By Friday she was troubleshooting for the teachers on either side of her and proposing uses for the tools that hadn't occurred to us.
The first two days were badly designed. We opened with capability — what the technology can do — when what the room needed was reassurance about their own place in it. Three teachers disengaged early and we never fully recovered them. We now open every training with the replacement question, because Akinyi taught us it's the one everybody is carrying and almost nobody says out loud.
Nobody asked her what happened next. She simply did it. Akinyi now runs digital literacy sessions for teachers at three neighboring schools, in Swahili, on her own time, without being paid for it. When we ask if she needs support she tells us what materials she wants and gets off the phone.
A teacher stands in front of roughly 1,200 students across a career. Akinyi has quietly multiplied herself by four.
$25 puts one teacher through one training session. It is the cheapest thing we do and it reaches the most children.
Train a teacherThe lesson always ended when the light did
There is a particular kind of quiet that happens in a classroom at four in the afternoon when the light starts to go. The head teacher described it to us before we ever saw it. The children don't complain. They just begin to squint, and then they stop writing, and then the lesson is over whether or not it was finished.
This school had been on the waiting list for a grid connection for longer than most of its pupils have been alive. Every year someone comes to survey it. Every year nothing follows.
Most digital education programs would have classed this school as not ready. No power, no connectivity, no realistic timeline for either. The sensible thing, on paper, is to work with schools where the infrastructure already exists and reach this one later.
Later never arrives. That's the whole problem.
“They kept telling us we were next. My first pupils are grown now. We were never next.”
— The head teacher, on the grid connectionSo we stopped designing around infrastructure that isn't coming. Solar panels, charging kept in a locked room the head teacher controls, devices running software that works with no signal at all. Nothing in the setup depends on a promise from anybody.
The first afternoon the devices stayed lit past four o'clock, one boy asked whether the lesson was going to keep going. He wasn't excited about the technology. He was asking whether the ending had moved.
Our original charging setup put the panels where they got afternoon shade from a building we hadn't accounted for. Yield dropped by roughly a third and nobody wanted to tell us, because they were worried it would look like a complaint. We found out three weeks later. We now do a follow-up site visit at week two by default, and we say explicitly that reporting a problem is not ingratitude.
A dead device in a locked cupboard helps nobody, so repairs, replacement parts and a named local contact were budgeted into this installation before the first panel was mounted. That line item is unglamorous and it is the reason equipment is still running in month nine.
$500 equips one classroom with solar power, devices, and offline tools that work where the electricity was never coming.
Light a classroomShe had the grades. She had never met an engineer.
Her teachers had been telling her she was exceptional since she was nine. She had heard it so often it had stopped meaning anything — the way a word repeated too many times becomes a sound. She knew she was good at mathematics. What she did not know was what being good at mathematics was for.
Nobody in her family has worked in technology. Nobody in her village has. She had never had a conversation with a person who writes software, and so the job existed for her the way a foreign city exists on a map: a real place, definitely, somewhere, populated by other people.
The assumption people make about a student like this is that she doesn't know university exists. That assumption is both wrong and slightly insulting. She knew exactly what university was. She knew the cut-off points. What she couldn't picture was the version of her own life that came afterwards, because she had never seen one.
“I could say I wanted to do computer science. I could not tell you what a person who does computer science actually does on a Tuesday.”
— ScholarPath student, first mentoring sessionHer mentor is a Kenyan software engineer who grew up two counties over and now works on systems used across the region. Their first call was supposed to run thirty minutes. It ran ninety, and most of it was her asking questions that had apparently been stacking up for years with nowhere to go. What does the day look like. What do you do when you can't solve it. Did anyone tell you that you couldn't.
That last one is the question underneath all the others.
Our first mentor match failed. We paired on technical field alone and the two of them had nothing to say to each other after twenty minutes. It was awkward for both and the student blamed herself, which is the outcome we most needed to avoid. We now match on background and circumstance as much as expertise, and we tell every student upfront that a bad match is our error to fix, not theirs to endure.
She is still in school. There is no triumphant ending here yet, and we would rather say so than invent one. What has changed is smaller and harder to photograph: she can now describe, in specific terms, a life she intends to have. Six months ago she could not.
Everything else this foundation does builds a floor under someone. This is the part where somebody reaches down.
$1,000 supports one scholar through a year of mentorship and study. Or give a few hours a month and be the person she can ask.
Back a scholarEvery case on this page has a price on it
None of these were dramatic. Nobody wrote a large cheque or held a gala. Somebody covered a training session, or a set of solar panels, or a year of mentoring — and then a teacher, a classroom, and a student did the rest of the work themselves.
We publish the ones that go wrong too
Every case here includes what didn't work, because a page of unbroken successes tells you nothing about whether an organization is learning. When something fails badly enough to be worth writing about, it will appear on this page alongside the rest.