Most ML knowledge fades the moment you close the tab. QuiddityML fixes that — rigorous exercises that build genuine understanding, spaced repetition that locks it in, and gamified daily practice that makes showing up addictive.
Watching a video is not learning. Copying a notebook is not understanding. QuiddityML is built differently.
Eleven exercise types that demand you actually think — spot bugs in real code, fill in implementations, trace through operations step by step, write from memory. Not passive. Not easy. That’s the point.
Spaced repetition adapts to your actual performance on every exercise. The algorithm knows what you struggled with and what you sailed through — and schedules reviews accordingly. No two users get the same queue.
Streaks, hearts, and a daily review queue give you just the right amount of pressure. Enough to build a habit. Not enough to patronize you. The content stays hard — the motivation system keeps you showing up.
Every concept follows the same loop — designed so understanding builds before it's tested, and nothing gets forgotten after.
Each concept opens with slides — clear explanations, diagrams, equations, and worked code examples. Real depth, not bullet points. You read and understand before anything is tested.
After each concept, exercises test your understanding from multiple angles — not just recognition, but application, debugging, and generation. You can't passively coast through.
Spaced repetition schedules every concept based on how well you actually performed — not just that you completed it. Struggled on an exercise? It comes back sooner. Nailed it? It waits longer. Your review queue is completely personal, built from your real performance data.
Four tiers of cognitive demand. Every concept gets tested at multiple levels, so understanding runs deep before you move on.
Targeted questions with four options. Tests concept recall and distinguishes real understanding from common misconceptions.
RecognitionSee the equation. Choose the correct implementation. Builds the bridge between the math in papers and the code you actually write.
Math ↔ CodeGiven an implementation, identify the corresponding equation or formula. Works the bridge in the other direction — equally important.
Math ↔ CodeSelect the correct components to build a module from scratch. Understand how pieces fit together before writing them yourself.
ArchitectureReal code with key parts removed. Fill them in from a token bank. No multiple choice on the code itself — you have to actually know it.
Active recallGiven shuffled code lines, put them in the right order. Forces you to understand execution flow, not just recognize syntax.
Control flowReal PyTorch code with subtle bugs. Identify the exact lines and understand why they fail — before you spend hours debugging your own project.
DebuggingGiven a code snippet, predict what error it raises — or whether it runs cleanly. Builds runtime intuition before it costs you an hour in prod.
Error intuitionChoose the correct test or assertion for a given function. Forces you to think about what code should guarantee, not just what it does.
TestingGiven a sequence of operations, predict intermediate tensor shapes, values, or errors at each step. Builds a precise mental model of how data moves through a network — the kind you can only get by thinking carefully.
Shape reasoningNo hints. No options. Write the full implementation from scratch, then compare against the reference. The highest-demand exercise in the app — and the one that turns exposure into mastery.
Free recallQuiddityML covers the modern machine learning landscape — from core foundations to cutting-edge research areas. New tracks ship regularly.
Unit 1 of every track at no cost. Unlock everything when you’re ready to go deeper.
Built for engineers who are serious about applying ML in production — not just passing a course.
Download on iOS