Every question you answer powers a robot you design. Then you teach it how to think — rule by rule, the way real AI is written — and send it to fight what other students built. The maths underneath is the actual KSSR/KSSM syllabus, and every wrong answer is tagged with the misconception behind it, so a teacher can see what a class is really getting stuck on.
Ninety seconds of maths at your level. Right answers earn Power. Speed and streaks pay more.
Spend Power on attack, armour, speed and frame. Every point visibly changes the robot.
Give it rules to follow — if this, then that. That is the AI half, and it is written by the student.
Send it in against the bosses, your class, or anyone on the board — and see whose thinking held up.
Every student writes their bot's decision-making themselves: a list of IF-this-THEN-that rules, in an order they choose, which the bot then follows on its own. That is a real AI technique — a rule-based agent — and the game renders it as actual Python beside the blocks.
Then comes the half that makes AI a discipline rather than a guess: measurement. The Lab reads their own fight history back and shows what share of decisions their rules actually made, which rules fire, which reads they won — as percentages over counts they can check by hand. The maths and the AI are the same numbers.
Its rules can only ask about what it can sense — so its senses decide what it can think about.
An ordered rule list is a policy. Move a rule up and you change what the AI believes matters most.
Read the data, find the weak rule, change it because the numbers said so.
Small samples mislead. AI can be confidently wrong. A human approves before anything changes.
Mapped to four of the five Big Ideas in AI from AI4K12 — the US national K-12 AI guidelines. The fifth, Natural Interaction, this game does not teach, and says so on screen rather than overclaiming.
Power is not points on a scoreboard — it is the budget for the robot. A student who works out that 3 armour beats 1 attack against a heavy boss has done real arithmetic reasoning, and they did it because they wanted to win.
Damage, armour, critical chance and health are all shown as the formulas that produce them, so a robot that loses is a problem to solve rather than a punishment.
Put a code on the board. Students open the page, type it, and they are in — no install, no logins, no email addresses. Start the round and the whole class answers the same question at the same moment, with a live scoreboard and a podium at the end.
Every wrong answer is tagged with the misconception behind it — not just marked wrong — so the report tells you why a class is stuck, whether they were racing a live round or practising alone.
Create a class — no login