Help kids build the things their communities actually need.
A project methodology built for rural classrooms — turns each student's own community concern into a real, shipped build by the end of the year.
From graduates who leave to graduates who build.
Caroline Kelly at Tanzania Education Corporation named the specific gap: many Tumaini students complete the program and return to their villages to do subsistence work — selling things, taking what jobs are available — instead of becoming the people building the future of their community.
The school produces educated students. It does not yet produce community builders. That identity can only be built by doing the work of building, repeatedly, while still in school, in service of the community the student belongs to.
This page describes a methodology designed for that specific gap. Every project a student ships during the program is a small rehearsal of the identity they could carry into adulthood. Do it three or four times across the school years, and the “community builder” identity has been practiced before the student ever needs to live it.
Built after a conversation with Caroline Kelly about the STEM & Computer Science Teacher Fellowship at Tumaini. The methodology is one I'm already running with kids in Brooklyn (chess, basketball) and would adapt to Tumaini's context. It pulls from the CHIMERA framework — body theory, gap physics, anchor-and-transfer — applied to a rural boarding-school context. Lives publicly so anyone working in similar settings can use it.
AI for the unseen
Most AI is being used by people who already have what they need — to generate ideas as products to sell. The use of AI that actually matters is the inverse: AI in the hands of people who have real problems but lacked the resources to solve them.
Voiceless to heard. Unseen to visible. Observation to built thing.
AI also surfaces invisible data — patterns in a student's own community that were always there but never legible. When students hold AI anduse it to see their own community's data, both halves activate at once. That is the curriculum.
I already teach kids who don't have laptops
The six- and seven-year-olds in my BeyondChess program in Brooklyn don't bring devices to class. The kids in the inclusive basketball program don't either. They have a version of the same technology gap Tumaini students have. What I've been teaching them is what this methodology trains — structural thinking, problem-solving, cross-domain relationships, emotional regulation — durable things that don't need a device in front of them. The geography differs. The pedagogical reality is closer than it looks.
From a student's observation to a community-deployed build
Identify
Start with the student's own observation or concern — a problem they actually see in their own life, family, or village. Not a topic assigned. Not a tutorial. Their thing.
Research
The student interviews classmates, teachers, family members, and people in the community around the concern. Listening becomes the first lesson. The build hasn't started yet; the student is becoming a researcher of their own world.
Connect
Together we map the concern to the people it will affect, to what they already have, to what's missing. This is when the cross-domain pattern recognition kicks in — what concept from class is actually structuring this real problem?
Brainstorm
The student leads on what their solution should do; the teacher guides on what's buildable. The student stays in the driver's seat of their own concern.
Build
The teacher drives the AI build process; the student shadows. They watch how the build unfolds, ask questions, redirect when they see the build drifting from their intent. They are not consuming AI content; they are stewarding the build of their own idea.
Ship
We deploy the build — a website, a system, a community tool — in a form their family, classmates, and village can actually use. Their name is on it. Their reasoning is on it. They wrote it.
The reward function students are actually optimizing
Most schools incentivize the wrong thing — and the students aren't wrong for noticing. Schools specify outcomes (a high test score, a finished assignment, a passing grade) and the student becomes the optimizer. The optimizer correctly delivers the specified outcome by the cheapest available path. In the 2000s that was SparkNotes. In the 2010s it was paying someone for an essay. Today it's asking ChatGPT to do the work. Learning was the thing we actually wanted, but we never put it into the reward function. So the optimizer ignores it. The student isn't broken. The student is aligned — to the reward we set, not the one we meant.
This isn't a morality problem. It's a specification problem. Scolding kids about integrity never worked because they were correctly optimizing the function we put in front of them.
The six-step methodology fixes this by rewarding the trajectory instead of the endpoint. Identify, Research, Connect, Brainstorm, Build, Ship — each is an observable process step. There's no SparkNotes for noticing a real problem in your own village. There's no shortcut to interviewing six families about water access. The shipped tool at the end isn't the rewarded target — it's the natural consequence of having actually walked the process. What the methodology rewards is the walk, not the arrival. That's why it resists the AI shortcut in a way an outcome-graded curriculum cannot.
The same logic applies to AI itself. AI doesn't repair broken reward functions; it amplifies whichever one is already running. Drop AI into an outcome-based school and you get faster cheating. Drop it into a process-rewarded curriculum and you get something closer to a master tutor for every student. AI is a multiplier pointed at whatever reward function the surrounding system already set.
Naserian and the school the herd keeps leaving behind
A concrete walk through all six steps with one student on one real concern. The six steps above are abstract without one. This is what the methodology actually does in a year.
Naserian, a Form 5 student.
Her family is Maasai pastoralist. Her younger brother and sister attend a village school nearby. This term the rains failed in March instead of October; her father moved the herd toward grazing land five months earlier than the usual September migration. Her siblings disappeared from school attendance overnight. They came back six weeks behind. Research finds around 43% of Maasai families now experience separation from livestock migration, and the timing has become unpredictable as climate patterns shift. The school calendar assumes attendance is binary; in this community, attendance follows the herd, and the herd follows the rain.
She names the concern in class: the school treats migration weeks as failure weeks. Her own siblings are marked behind for being absent during something nobody in their family chose. It's the first time she frames the gap as a problem with a structure rather than just how things are.
Over two weeks she interviews three village-school teachers, two parents who've migrated mid-term, and her own siblings. Teachers tell her they feel guilty marking absent kids "behind" when the absence wasn't a choice. Parents wish there were a way to bring the lesson plan with the herd. Siblings remember which weeks they missed and can describe what they wished they'd had.
In class, the teacher helps her map this to her Form 5 Systems Development unit (Chapter 6) — actors (school, family, child), information flows (attendance, lesson plan, migration calendar), the bottleneck (no shared knowledge between school and family once migration starts). The same shape that lives in any information system.
She designs a school-family migration bridge. Teachers log who's present each week. Families register their migration calendar in advance when they can. The system SMS-summarizes the week's lesson topics to migrating families so the child can continue with the family during migration. Works on feature phones — which every family has.
The teacher drives the build — Africa's Talking SMS API, a simple web form for attendance logging, a family-registration intake. Naserian shadows. When the build drifts (the AI suggests an English-only interface), she redirects: families read Swahili and Maa, not English. The messaging language was never the AI's call to make.
Eighteen families in the catchment subscribe for one term. Teachers report less guilt around migration weeks; they now treat the absence as a structural fact, not a learning failure. Families say they know what to teach during the migration. Naserian's own siblings come back caught up enough to keep the term moving.
Naserian presents in Swahili and English to families, village teachers, her own siblings. She explains the choices she made and the ones she got wrong. Her name is on the system. Her teachers tell the room that the tool changed how they think about the migration weeks — they grade those weeks differently now. She is, by this point, a community builder, because she has been one repeatedly for nine months. The next year's cohort inherits the system and extends it — the local teacher co-driver (now Year 2 of their own arc) carries the continuity.
Naserian is a placeholder name. The migration-attendance gap is real and well-documented in Maasai catchments. The real student, with their real concern, comes from TEC staff and the kids themselves. The methodology is the shape; the content is theirs.
Mastering one domain teaches you all the others
Underneath the methodology is a deeper principle about how people actually learn anything well. It's the principle I've been using in BeyondChess, in the inclusive basketball program I coach in Brooklyn, and in the broader framework I've been developing for the last few years. It works in three stages:
- Anchor in one dimension.Pick a domain you're willing to actually master. Chess. A guitar tune. A basketball play. The map and movement of a video game. A piece of your own community's daily life. The specific choice matters less than the depth of attention.
- Master the movement first, then the rhythm.At first your attention is on the mechanics — where the piece goes, where your hands go, where your character is on the screen. With repetition, the mechanics dissolve into rhythm. You stop thinking about the controls; you just play. Your consciousness is freed from the surface details and can attend to the pattern beneath.
- Transfer the freed consciousness into a new dimension. Once the rhythm is grooved in one domain, the same attentional muscle becomes available to apply elsewhere. The student who has truly grooved the rhythm of a chess game can recognize the same shape in algorithm design, then in problem solving in their own village, then in the structure of an argument. The depth of one domain enables the breadth across many.
For Tumaini students, the anchor domain is already in their lives. The rhythm of herding livestock through the seasons. The pattern of which weeks the market price for cattle moves. The way a family compound organizes its work across the day. The shape of a chess opening, if they play. Students arrive deeply rhythmic in their own lived domains. The methodology's six steps are the bridge that transfers that familiarity into the new domain of building.
The depth of mastery in one domain is the substrate that lets consciousness move freely across all the others.
Students don't arrive empty. The work of teaching is showing them that the same rhythm they already carry in herding, in family work, in chess, in songs, carries over into building something for their community. The methodology's job is making that transfer visible and practiced.
Cross-Domain Primitive Transfer
How the same attentional movement structures itself across four different fields of practice.
Chess
Noticing an undefended king-side pawn vulnerability.
Analyzing previous tournament games and opponent's opening preferences.
Mapping the tactical weakness to core positional principles.
Visualizing candidate move-sequences and branch alternatives.
Playing the sequence out on the board under pressure.
Reflecting on the match outcome during the post-game review.
Basketball
Noticing defensive pressure crowding the ball-handler.
Calling out court patterns and checking teammate positioning.
Mapping court geometry to spacing and cutting patterns.
Selecting a high-pick-and-roll play structure dynamically.
Executing the screen and crisp kick-out pass.
Scoring or resetting the half-court offensive set.
Music / Guitar
Noticing an emotional gap in the musical phrasing.
Listening closely to chord progressions and key changes.
Mapping chord scales to the physical fretboard shapes.
Improvising melodic riffs or variations over the rhythm.
Plucking or strumming the composition into physical vibration.
Performing the duet in perfect, unforced alignment.
Community Tech
Noticing local water failure or agricultural loss patterns.
Interviewing family members and market vendors about the problem.
Mapping local dynamics to NECTA CS systems theory.
Collaborating on a clean, buildable system solution.
Shadow-building the application powered by AI models.
Deploying the usable tool for local community use.
Shadow → Interrogate → Drive
The most important pedagogical feature of this curriculum is the arc students move through with AI itself. Three stages, mapped to Form level, each load-bearing.
Shadow
Younger students do not touch AI directly. The instructor drives the build; students watch how the work unfolds, ask questions, redirect when the build drifts from their intent. They are mirroring, not consuming. The structured-thinking lesson is preserved because nothing replaces their reasoning.
Interrogate
Mid-program students get AI access but with a different mandate. They compare AI output against their own field research and their own community context. Where does the model misread their village? Where does it produce confident nonsense? This stage teaches discernment — how to use a tool that's sometimes wrong in the very places that matter.
Drive
Senior students drive their own builds, alongside the instructor rather than under them. They mentor Form V interrogators and guide Form I–IV shadowing groups. The build capacity becomes the school's, not just the instructor's — which is the real endpoint of the program.
Two things this arc resolves. First, the apparent contradiction between “students don't touch AI” and “students learn to interrogate AI” — it's not a contradiction; it's a sequence. Second, the durability problem any single-instructor program faces. If the build capacity only lives in one fellow, the methodology leaves when the fellow does. The arc's explicit endpoint is students themselves carrying the methodology — peer-teaching it across cohorts. A garden that keeps growing, not a tower that needs the architect.
It's like driving with a teen learning to drive — at first they hold the wheel and you hold the gear shift. Eventually they hold both. Then they teach the next driver.
A teacher co-driver, on the same arc as the students
A fellowship typically runs one or two years. Students need three to four years to reach the Drive stage. If the fellow departs before students reach Drive, every mid-arc cohort collapses — the build capacity leaves with the fellow. That's the durability hole the student-arc alone cannot close.
The structural fix is that a local Tanzanian teacher co-drives from day one, on the same Shadow → Interrogate → Drive arc as the students. The methodology trains its own replacement.
- Year 1 — Local teacher shadows. Not just observing the fellow; learning the methodology by walking it with the students.
- Year 2 — Local teacher interrogates. Running parts of the methodology themselves, with the fellow shadowing them. Where do they drift? Where do they need course-correction? That's the work.
- Year 3 and beyond — Local teacher drives. They own the methodology. The next cohort runs without the fellow needing to be in the room. New co-teachers can join at Shadow, work through Interrogate, eventually Drive.
This is what makes the program Garden-shaped instead of Tower-shaped. The methodology doesn't depend on the fellow being permanent. It depends on training its own local replacement, on the same arc, while the students do the same work alongside them.
A note on what makes this hard: prior East African technology-in-classroom programs consistently found that the single hardest part of the work wasn't the technology — it was helping teachers shift from authority-dispenser to facilitator. The shadowing-to-driving arc above embeds that shift in the design itself, so teachers grow into the facilitator role alongside students rather than being expected to arrive pre-adjusted.
A related operational question worth naming: teachers are evaluated on exam outcomes. Every hour spent on project work is, on the surface, an hour bet against the metric the teacher's own standing depends on. The project component only works if it visibly protects exam performance rather than competing with it — which is why the methodology has to ride on Form 5 Chapters 5 and 6 (exam-mandated units), not run alongside them as separate enrichment time. The teacher isn't losing exam-prep hours to the project. The project IS the exam-prep work for those units.
The Trojan horse is Form 5, Chapters 5 and 6
Tumaini is NECTA-aligned and exam-focused, which is the right posture for a school where students' exam outcomes carry real weight. Anything I'd propose layers onto the 2023 TIE A-level Computer Science syllabus rather than competing with it. I also heard from TEC that the school has been historically test-based and is working to become more hands-on. The project-based methodology described here is built for exactly that shift — it doesn't fight the exam track; it gives the hands-on component a clear structure that respects the exam track.
The natural spine for this methodology is already in the syllabus. Form 5 Chapter 5 covers Website Development — HTML, CSS, JavaScript, APIs — and Form 5 Chapter 6 covers Systems Development — the full lifecycle of plan, analyze, design, implement. Both are exam-mandated. Both are the exact tools the youth-builder methodology wants students to be producing in.
Each student team builds one website or system over the year that solves a real Makuyuni problem. The exam prep is the community project.
C++ and the algorithmic units stay as taught — no displacement of the syllabus. The methodology runs as the spine of the project component, not as separate enrichment time. Form 6's “Emerging Technologies” unit, which is more openly defined, is the natural place to bring in AI literacy and the practice of interrogating AI output rather than consuming it.
The giants this work stands on
The shape of this work — students identifying real community concerns, walking them through a structured pipeline, shipping artifacts that matter to people they know — has been validated at meaningful scale by programs that came before. Naming them here because anyone reading this should be able to see what evidence the methodology rides on rather than treat it as fresh invention.
Technovation Girls
An 18-year-running, free program in over 120 countries including Tanzania, where teams of girls work with a mentor to identify a problem in their community, prototype a technology solution, test with real users, and pitch. Reached over 400,000 girls and family members. Multi-year longitudinal evidence: the model develops durable identities as problem solvers. That outcome — the community-builder identity formed through repeated shipping — is the same outcome this methodology targets. Technovation has already paid most of the proof-of-concept bill for the project-pipeline shape. If TEC isn't already running Technovation as an extracurricular, it would be the single highest- leverage program to add alongside anything I'd build.
Project SHINE — Ngorongoro
Two Maasai pastoralist boarding schools — the same demographic Tumaini serves. Student-led clubs plus a community science fair drawing 500–1,000 attendees per school. The critical framing was One Health— linking livestock, human, and environmental concerns, because that's the lived reality of pastoralist students.
Apps and Girls + the Tanzanian youth-tech ecosystem
Apps and Girls (Tanzania, since 2013) runs in-school sessions, after-school programs, youth boot camps, and teacher training — empowering girls and young women to create with technology. Alongside it, She Codes for Change, Buni Divaz, TechChix, and the Buni hub form an active, Tanzanian-led ecosystem already shipping in this space. Regionally, iCog Anyone Can Codein Ethiopia has reached over 26,000 kids. Anything I'd build at Tumaini ought to sit beside these, not above them.
The honest framing: the project-pipeline structure is well-validated and well-staffed. Where this concept piece adds something is narrower and more specific — the deliberate AI-staging arc (Shadow → Interrogate → Drive mapped to Form level), the observation pedagogy (pressure vs attention, self-observation as endpoint), and the NECTA-locked in-curriculum integration that complements the extracurricular programs above rather than competing with them. SHINE measured behavior and engagement, not migration retention; the values stance “build the community up” is grounded in engagement evidence, not migration data. The work earns confidence from what's measured, not what isn't.
Concrete project examples
These are starting points based on what I know about Monduli District and the Maasai pastoralist context. The actual project list comes from the students, not from me. But it helps to ground the methodology in concrete examples a teacher can imagine teaching.
A community water-source map
Students survey their own families and a sample of village families about where they get water, how far it is, and what time of year sources fail. They build a simple web app that maps it. Maps and time-series data become a teachable cross-domain primitive: the same shape shows up in CS class as a graph, in geography as a map, in math as a coordinate system.
Livestock market price tracker
Students collect weekly cattle and goat prices at the Makuyuni market and the nearest comparison markets, surface trends, and build a simple SMS-or-web tool families can check before they sell. Pastoralism economics becomes the engine of the algorithm lesson.
MWEDO microenterprise inventory app
MWEDO already runs a network of Maasai women's microenterprises across the catchment. Students build a small inventory or order-tracking tool with one or two of those enterprises as real clients. The build has a real user from day one.
Tourism-job application support
Makuyuni sits in the safari corridor. Many young adults seek work in the tourism economy and don't have help with applications, CVs, or skills articulation. Students build a tool that helps community members structure their experience into application-ready form.
School attendance tracker
Pastoralist family migration disrupts school attendance in unpredictable ways. Students build a simple tool that helps teachers track attendance patterns by family movement, and helps families anticipate which terms their children will be present for.
The real list is whatever students tell me they see. Tumaini staff input needed here
What the AI stack actually looks like at Tumaini
Before the realistic stack, one note on where I think the deepest leverage of AI actually lives. Humans are naturally great students. We learn quickly; some of us learn better than others, but the baseline is good. The bottleneck in education isn't the student side. It's the teacher side. Good teachers are rare because being a good teacher requires both deep expertise in a domain andawareness of how that domain relates to everything else in a student's life. That awareness is what lets a teacher meet a student where they actually are, in language that fits their reality. Most teachers have the expertise. Far fewer have the awareness, because the awareness takes years of practice across many domains to acquire.
AI changes that supply problem. It doesn't make students better. It makes the supply of teacher-level awareness practically infinite. The same tools that let one good teacher tailor instruction to one student now let a program tailor instruction to thirty students at the depth that used to require thirty one-to-one tutors. And on the teacher's side, AI helps existing teachers expand their own awareness — across domains they don't have time to master themselves, across student backgrounds they don't share, across pedagogical approaches they didn't train in.
AI doesn't replace teachers. It gives every teacher the awareness of a master teacher, and gives every student the attention of a tutor who knows them.
That is also what the bridge between AI and humans looks like in practice. AI is most valuable when it amplifies what a good teacher already does — meeting the student where they are, in language that fits their reality. Used that way, it's an alignment tool in the truest sense.
Practically, however — student-facing, always-on ChatGPT will not work in a rural boarding school with variable connectivity, intermittent electricity, and per-query costs that compound at scale. The realistic stack, in priority order:
- Teacher-facing AI first. Lesson differentiation, planning, grading support. One connection serves one teacher serves 30 students. Highest leverage of whatever bandwidth Tumaini has.
- Students learning to interrogate AI, not consume it. What AI gets wrong about their own community is a more durable lesson than what it gets right. Students compare AI's reasoning to their own observations and notice where AI doesn't actually understand local context. Taught alongside this: AI is the latest step in a chain of human choices — cognition, math, computing, networking, machine learning, language models — not a magic box. Students see themselves inside the lineage, not on the receiving end of it.
- On-device or LAN-hosted small models, eventually. Llama / Phi / Gemma class models on the existing lab server, for offline Swahili-aware tutoring. Powering Potential's solar Raspberry Pi labs in nearby Karatu District are the closest comparable working model.
One concrete infrastructure parallel worth naming: Powering Potential's solar-powered Raspberry Pi labs in Karatu district — the nearest comparable district to Makuyuni — run fully offline, with 20 student devices plus a content server, on solar power, in pastoralist boarding schools. It is the single closest working model to what an AI-augmented Tumaini lab could look like. The other infrastructure questions — which carrier line the school uses, what redundancy looks like for the AI calls, who maintains the LAN models — are downstream decisions I'd want to make with TEC's ops team who knows the on-the-ground reality I can't see from Brooklyn.
Two kinds of observation
Observation does two completely different jobs in a classroom, and they fight each other. Most education conversations conflate them; clean teaching keeps them separate.
Being observed
The camera. The test. The judgmental gaze. Eyes-on changes the performer — they either break or they perform, but they don't stay the same. The kid optimizing for the grade is under pressure-observation. It produces brittle behaviour: shortcuts, cheating, performance for the test rather than the learning. This is force applied to a person.
Observing
The teacher freely directing their own attention at a specific student. The student freely directing their own attention at a real community problem. This is not force; it's release. Where pressure burns out, attention keeps generating — it's the conserved resource in good teaching. Solutions surface naturally where attention lands.
Bad teaching points eyes at the student to make them perform. Good teaching points attention from the teacher toward the student, and from the student toward a real problem. Same act of looking. Opposite directions. Opposite outcomes.
The methodology is structurally a machine for converting pressure-observation into attention-observation. Every step replaces "the teacher is watching me perform" with "I am attending to something that matters." That's why it resists the shortcut-driven behaviour an outcome-graded system produces — the shortcut only exists when somebody is performing for eyes. When the student is genuinely attending to their own community's problem, there's nothing to cheat around.
Four stages of observation across the program
- The teacher attends to the student. Early on, the teacher does most of the seeing — reading where each student actually is, what they're gripped by, what they don't yet have language for.
- The student attends to their community. The methodology turns the student's attention outward, onto a real problem the people around them live with. The student becomes an observer of their own world.
- The student attends to their own work. During Build and Ship, they watch what the AI produces, redirect when it drifts, take ownership of decisions. The teacher's attention has been internalized.
- The student attends to themselves. Self-observation — without a camera, without a teacher standing over them — is the endpoint. A kid who can put focused attention on their own thinking and work, alone, is a kid who has graduated this curriculum. The community-builder identity ultimately requires it.
What this work is, underneath all of it, is translation. Every student has a domain where they think with clarity and power — herding, seasonal markets, family rhythms, chess, songs. They also have domains where they don't — programming, formal CS, written argumentation. The teacher and AI together translate a problem from the unfamiliar domain into the one where the student already has mastery, so they can solve it where they're free, then carry the solution back. When Naserian sees her family's migration-attendance gap is the same shape as the Systems Development unit, she hasn't learned a lesson — she's discovered something about herself she already had.
Don't blend. Divide the labor.
The question every reader is going to ask sooner or later is how AI and human teaching mix without one swallowing the other. The honest answer isn't that they mix. It's that they each do what only they can do, and the relation between them stays clean.
What only the teacher can do
- Care about a specific student
- Read a face in a room
- Sit with a kid through a hard moment
- Notice what actually matters in this child's life
- Physically be present at Demo Day
What AI lets the teacher reach
- Per-student gap reading across 30 students
- Differentiated examples in each kid's context
- Pattern surfacing across student backgrounds
- Grading and lesson-prep at one-teacher scale
- Translation between domains for each student
You don't blend a teacher and an AI into a paste. You keep both poles intact and let each do what it's good at. The relation is the thing.
The peer-teaching channel compounds this. Student-to-student teaching is one of the most effective practices in education — when the pair is the right distance apart, close enough that the explanation transmits, far enough that something is actually being transmitted. AI lets the program see those productive pairings deliberately and surface them. Students teach students at the gap distances where teaching actually transfers, and the teacher's attention is freed to land where it's most needed.
Open questions
These are the questions I'd most want your team's read on. They're what would make the next version of this page actually specific to Tumaini rather than informed-by-research.
- →What are the main issues your students and your community are facing right now — top of mind for you and your staff?
- →What solutions have you and your team already tried? What worked, what partially worked, what didn't?
- →What did the prior STEM/CS fellow leave behind — curriculum, lesson plans, projects, lessons learned?
- →Where does the existing NECTA-aligned curriculum leave the most room for project-based enrichment?
- →What are your students actually most concerned about in their own community? (Your staff's read, not mine.)
- →Would you be open to connecting me with a past TEC fellow whose lived experience would shape what I build?
The fellowship is one channel — not the only one
If the in-person STEM & CS Teacher Fellowship at Tumaini turns out to be the right fit, I would be honored to step in. If you find someone who is a better fit for the day-to-day work on the ground, that is genuinely fine with me — and I would still want to help in whatever way is useful to TEC and to the students at Tumaini. Specifically:
- ·Remote curriculum collaboration. Co-developing the methodology with whoever lands in the fellowship role, from here.
- ·Teacher training support. Helping your existing teachers and any new fellow ramp on the project-based, shadowing-pattern methodology — remotely or in short visits.
- ·Idea transfer. Whatever I learn from the parallel youth-builder program I'm developing in Brooklyn with Lamont Kirton, I'll share with your team as it lands. No charge. Just useful.
- ·This page itself. Use any of the methodology, syllabus mapping, project examples, or evidence framing for your own thinking. It's here for you to take what works.
If TEC isn't already running Technovation Girls at Tumaini, that's probably the highest-leverage program to add as the extracurricular project track — they've been doing this shape of work in Tanzania for over a decade, free, with mentor support already structured. What I'd build is genuinely different from that: the in-curriculum, NECTA-locked, deliberate-AI-staging spine that fits inside the school day and respects the exam track. The two would complement, not compete. Better to name that out loud than pretend my work covers ground that Technovation already covers better.
For context: the chess and basketball programs I coach in Brooklyn are unpaid volunteer work. The community builds I do — websites, apps, AI tools, automations for people who need them — are unpaid. CHIMERA, the framework underneath all of this, is open-source. None of this is monetized. I mention it because it matters for how I'm showing up here too. This concept piece is offered in the same posture as the rest of my teaching work — because I'm trying to figure out the right solution for each situation, and Tumaini is a situation worth figuring out. The role question is downstream of that.
A usable blueprint for similar contexts
This concept is written for Tumaini specifically because TEC and this fellowship are the actual conversation that occasioned it. But the constraints Tumaini operates under — strict national curriculum, single shared lab, limited connectivity, students with uneven device exposure, a community whose problems are underrepresented in conventional tech-ed pipelines — apply to many rural and under-resourced secondary schools globally. The methodology, the shadowing pattern, the syllabus-as-trojan-horse move, the “graduates as community builders” reframe — none of these are Tumaini-specific. They're context-portable.
If TEC ends up using parts of this, that's the highest outcome. If any other school, NGO, fellow, or educator looking at a similar landscape finds something here they can adapt, that's also the work. The page will keep evolving as I learn more, and if the right way to use it is for someone else to fork it for their own context, I'd be glad to help with that.
A small note on the form of this page: it was itself built using the six-step methodology described above. The concern (Tumaini's graduates-as-community-builders gap) came from a real conversation. The research is in the sources cited. The connecting move was mapping the concern to the Form 5 syllabus units the methodology can ride on. The build is the page you've just read. The ship is its public URL. If the methodology works on its own author writing it, that's a small piece of evidence that it might work on a student writing their own first build in Makuyuni.
The work is the gift. The role is one channel among several. Use whatever's useful here.
— Kareem Adedeji Richard Akabashorun
v1.2 · last updated 2026-05-22 · this page keeps updating as the conversation continues