Teaching Kids AI Before They Can Code? My Honest Take on SafiMoyo

My eight-year-old nephew asked me last week how ChatGPT "knows things." I stared at him for a second, my mind racing through neural networks, weights, and training datasets, trying to figure out how to explain large language models to a kid whose biggest daily worry is whether his school bus will beat the Lagos traffic.

I ended up giving a clumsy explanation about "a very smart guessing machine." It wasn't great.

So when I saw that Safi Education officially launched "SafiMoyo" yesterday, aiming to teach foundational AI literacy to African kids aged 4 to 18, it immediately caught my attention. The team, led by Evans Edeha, is trying to treat AI like reading or basic math—teaching the core logic before kids ever write a single line of code.

As a developer who spends half his life debugging and the other half trying to build products that survive the local market, I have some thoughts on this.

The Good: Logic Before Syntax

I’ve always believed we rush kids into coding too fast. We throw them into Scratch or Python syntax when they don't even understand computational thinking yet.

What I like about SafiMoyo’s approach—at least from what the co-founder, Uchechukwu Banye-Edeha, is saying—is that they are focusing on how to think with AI, not just how to build with it.

If you can teach a kid in Akure or Owerri how data inputs lead to algorithmic decisions, you’ve given them a mental framework. They don't need to struggle with syntax errors on a sluggish 2GB RAM laptop to understand the core concept of machine learning. That logic sticks with them. When they finally do pick up Python or JS in secondary school, they’ll actually know why they are writing those functions.

The Reality Check: Data, Power, and Attention Spans

But let's talk about execution, because that’s where things usually get messy in our ecosystem.

SafiMoyo is delivering this via short-form video lessons. On paper, that sounds great. Kids love video. But as a tech founder, my mind immediately jumps to the infrastructure.

If a parent in a Gbagada apartment or a cold morning in Jos is trying to get their kid to watch these lessons, they are thinking about data costs. Video is heavy. If the platform isn't highly optimized, compressed, or capable of offline playback, "Sapa" and data budgets will limit its reach.

Also, the age bracket of 4 to 18 is massive.

A 5-year-old learns through play, touch, and repetition. An 18-year-old preparing for JAMB wants to know how to use these tools to build a portfolio or pass exams. Isaac Folarin, their CTO, mentioned they are building this as a "learning infrastructure" with age-appropriate pathways. I really hope the platform’s backend UX handles this separation well, because a one-size-fits-all dashboard will fail fast.

Can We Demystify the Black Box?

Despite my engineering skepticism, I’m rooting for this. Most AI tools and curricula are built in San Francisco or London. They assume high-speed fiber internet, iPads for every child, and constant electricity. They don’t design for the kid who only gets access to a smartphone for two hours after their parents get home from the market.

If SafiMoyo can successfully localize the context of AI—making it relatable to an African child's daily reality—they’ll be doing something incredibly valuable.

We don't just need consumers of AI; we need kids who grow up understanding that these models are just tools built by humans, tools they can eventually control, tweak, and build themselves to solve actual local problems. Let’s see how they execute.