Five games, one goal: make AI knowledge stick. Pick your poison — no signup, no downloads, just you versus the machine.
Pick a difficulty, answer 10 questions, and see how well you really know artificial intelligence. You have 15 seconds per question.
You ran out of hearts. Review the flashcards and come back stronger!
Reading about a concept and being able to use it are different achievements. You can read a clear explanation of AI hallucination, agree with every sentence, and still be fooled by a fabricated statistic an hour later. The gap between recognising an idea and applying it under pressure is what these games are built to close.
Each one isolates a single skill that matters in practice, then forces you to exercise it against a timer or a score. Nothing here requires an account, a download, or any prior experience.
Ten questions per round across four difficulty settings, with hearts, streaks, XP, and a time bonus. Beginner questions cover AI fundamentals and prompt engineering; intermediate covers applied machine learning from gradient descent to model evaluation; expert covers transformers, RLHF, and inference optimisation. The fastest way to find out which track you should be studying.
Eighteen stops along one winding road, from beginner country through machine learning hills to the deep AI mountains. Correct answers move your car forward; mistakes are potholes that cost a heart and hold you in place. Difficulty climbs in dependency order, so where you break down identifies exactly which concept your understanding rests on but does not actually cover.
Three AI statements per round, one of them a realistic fabrication. Twelve rounds spanning acronyms, AI history, statistics, ethics, tokens, and agents. Every round ends with an explanation of why the false statement is false, which makes this the most directly useful game here: verification is the skill that separates people who use AI safely from people who repeat its mistakes.
Customers bring in prompts that produced disappointing output, and you repair them from a toolbox of fixes. Good repairs — naming the audience, adding real specifics, specifying the output format — visibly transform the AI’s response through four stages. Bad repairs like ALL CAPS and “be creative” cost strokes and explain why they add nothing. Par is nine across three jobs.
Label ten emails as spam or inbox, then watch a filter that learned only from your labels face eight live ones. Your mistakes become its rules, traced back to you explicitly. Two edge cases — a legitimate email containing the word FREE, and a polite phishing attempt — demonstrate overfitting and training-data coverage gaps more memorably than any diagram.
The games are practice, not instruction. Each links back to the module that teaches its concept properly, and every term you meet in them is defined in the AI glossary. If you prefer studying before testing yourself, the flashcards cover sixty core terms across all three tracks. The full curriculum is 34 free modules, starting at Module 01.