We Stopped Teaching Kids to Code

And that’s exactly why we built Reasonix MAC™.

There is a sentence I have heard in every education ministry meeting, every school board presentation, and every parent evening for the last five years:

“We need to teach our children to code.”

I understood it in 2019. I don’t accept it in 2026.

Here is the uncomfortable arithmetic. A child who enters Grade 1 today will graduate around 2038. By then, the specific skill of typing syntax into an editor — the thing “learn to code” usually means — will have been automated for over a decade. It is already mostly automated. The AI writing this generation’s boilerplate is better at Python syntax than any twelve-year-old will ever be, and it works for pennies.

So when we teach a child to code, we are training them to compete with a machine at the one thing the machine is best at.

That is not preparation. That’s a trap with a certificate at the end of it.

The thing the machine cannot do

Watch a nine-year-old actually solve a problem and you’ll see something the syntax never shows you.

We give them a delivery drone. The hospital needs blood in nine minutes. The drone knows one route — straight over the tower. Tonight, the tower is inside a storm.

Most children do the same thing first. They build the straight route anyway. The drone flies into the storm. They lose.

Then something interesting happens — and what happens next is the entire measurement.

Some children delete everything and start again, hoping. Some children stare at the route and count squares with a finger. Some children say, out loud, to nobody: “but what if it checks first?”

That sentence — what if it checks first — is the whole of conditional logic, invented from scratch by a child who has never heard the word else. It is also, structurally, the same move a surgeon makes before an incision, a lawyer makes before a filing, and an engineer makes before a launch. Test the condition. Branch on the answer.

We do not teach children if / else so they can write if / else. We teach it because it is a shape their thinking can take for the rest of their lives, in rooms where there is no keyboard.

That is what Reasonix MAC is for.

Mathematics, Algorithms, Coding — in that order, on purpose

MAC is built on a single structural claim: mathematics, algorithms and code are three languages for one thought, and children should meet all three simultaneously or they will understand none of them properly.

M — Mathematics. The structure underneath. Pattern, comparison, logic, probability.
A — Algorithms. The method on top. Decompose, abstract, sequence, debug.
C — Coding. The proof of thought. Where the idea becomes a thing that runs.

Every lesson runs all three. When a child meets loops, they meet them as multiplication (M), as “this pattern repeats” (A), and as repeat ×2 (C) — in the same fifteen minutes. Eventually they stop experiencing these as three subjects and start noticing they were always one idea wearing three costumes.

This is not a curriculum design flourish. It’s a claim about how transfer actually works. A skill learned in one context stays trapped in that context unless the learner can see its shape somewhere else. Teach loops only in Scratch, and you have taught a child Scratch.

The eight levels run from thinking in steps to variables — the program’s memory, and the last lesson of the last level isn’t about variables at all. It’s about an AI that sorts rubbish correctly nine times out of ten, and gets a shiny crisp packet wrong. The question we ask a thirteen-year-old is not “how do we fix the model?”

It is: “What examples did it see — and which ones didn’t it?”

Blame the data before you blame the machine. Every AI failure is, in the end, a question about what humans chose to show it. Which makes it a question about us. If a child finishes MAC understanding only that, we have done more for their AI literacy than a hundred hours of prompt-engineering tutorials.

The part nobody wants to hear: the puzzle is not the assessment

Here is where MAC breaks with almost every platform in this category.

A child who solves the puzzle but cannot say why it worked has learned a trick, not a concept. The trick will not survive contact with a new problem. It will not transfer. And, crucially, it will not be visible to anyone — because the dashboard will show a green tick, and everybody will feel good, and nothing will have happened.

So in MAC, the reflection is not optional and not skippable. After every lesson, the child writes — or, in Grades 1–3, says — why the solution works.

“Describe a decision you made this morning as an if / else. What was your condition?”

That answer is the strongest evidence in the entire system. Stronger than the puzzle. It is the difference between a child who can operate a concept and a child who possesses one.

The real product is a number that does not exist yet

The world measures reading. It measures mathematics. It measures, exhaustively, whether a fifteen-year-old can factor a quadratic.

Nobody measures thinking. Not systematically, not longitudinally, not in a way a school can act on in March rather than discovering in June.

MAC produces a Computational Thinking Index — seven constructs (decomposition, pattern recognition, abstraction, algorithm design, debugging, evaluation, AI literacy), scored per child, derived entirely from evidence of what they actually did.

Not from a test. From the shape of their behaviour:

How many attempts before it worked — and did they read between attempts, or just guess again?
Did they reach for a hint before trying once?
When it failed, did they trace their own steps, or delete everything and restart?
Can they explain the solution in their own words?

Two children can arrive at the same green tick by completely different roads, and MAC scores those roads differently — because the road is the learning. A child who fails four times, re-reads her own algorithm, finds her own error and fixes it has demonstrated the single strongest predictor of later success in computer science. The child who guessed correctly on the first try has demonstrated nothing at all.

We tell parents this in plain language on Sunday. We tell teachers on Monday, sorted by who needs them most. We tell the school at the end of term, with the intervention costed in teacher-hours.

Same evidence. Three audiences. One source of truth. That last part matters more than it sounds — because the moment a dashboard shows a number that cannot be traced back to something a child genuinely did, you no longer have a measurement. You have a decoration.

What we deliberately did not build

MAC has XP, coins, streaks, badges and a class league. It is, by any honest description, gamified.

It also has rails that cost us engagement metrics, on purpose:

XP is only earned through evidence. Rewatching a video earns nothing. Time on screen earns nothing.
No notifications after 21:00. Ever. Learning quality collapses after that hour and we will not trade a child’s sleep for a retention chart.
The bottom third of a league never sees its rank. Comparison is fuel for some children and poison for others, and we know which children we would be poisoning.
A soft stop at 45 minutes. The screen says: you’ve done good thinking today.
No confetti storms. Rewards are a single, quiet spring animation. We are not building a slot machine with a curriculum attached.

Every one of these decisions makes some number in some dashboard look worse. Every one of them is a decision I can defend to a ministry, a school board, and a parent — and I would rather defend those than defend a growth chart.

Arabic is not a translation

One more thing, and for us it is not a small one.

MAC ships Arabic-first, not Arabic-eventually. The stories are written in Arabic, not converted into it. The typography is rebuilt — Arabic loses its letters when you add tracking, and needs vertical air that Latin does not. The layout mirrors.

But the robot’s grid does not mirror, and neither does an equation. Flip a coordinate grid right-to-left and you have not localised anything; you have inverted the x-axis and destroyed a nine-year-old’s model of what forward means. That is a correctness bug wearing the costume of a cultural courtesy — and almost every RTL product in this region ships it.

We built this in Amman. Our children deserve the version where somebody actually thought about it.

The measure of success

We will know Reasonix MAC works not when a child ships an app.

We will know it works when a nine-year-old, at dinner, is asked why she took her coat this morning — and answers, without knowing she is doing anything remarkable:

“Because it was raining and it was cold. If it were only raining, I’d just take the umbrella.”

She has not learned to code.

She has learned to think — and the code was simply the place where we could finally see her doing it.

Reasonix MAC™ — Mathematics • Algorithms • Coding. Because programming is not a language… it’s a way of thinking.

Reasonix — Integrated Intelligence for Technology and Education Amman, Jordan reasonixmac.com