Quick Maffs vs The Global Min

QuickMaffs is a simple mental math app with NO ADS or PAYWALLS! Perfect for students, Quant finance interviews, or anyone who wants to improve their math skills. Features include 13 game modes of various arithmetic types, each customizable with tailored difficulty levels. Modes: Addition, subtraction, multiplication, division, squaring, doubling, halving, linear equations, quadratic equations, equation systems, mean, percentages and trigonometry. (This app is designed and maintained by Andrew Martin.)

Walk into your next data-science interview having already made every mistake — in the app, not in the room. The Global Min drills you with more than 900 multiple-choice questions on what data-science, ML and AI-engineering interviews actually test in 2026 — and it meets you where you are. Every topic has three levels. Fundamentals check that you know what a p-value is, or what a control group buys you. Expert questions drop you into a real work situation — a RAG assistant whose answers degraded after new documents were ingested, an A/B test a PM wants to stop early at p=0.03, a churn model that aces offline metrics and flops in production — and ask the question an interviewer would: "How would you proceed?" WHY IT MAKES YOU BETTER The wrong answers are the product. Each question is engineered with distractors that sound right: answers an average-good candidate would give, and plausible alternative methods that don't fit the scenario. Picking between "correct" and "almost correct" is exactly the skill interviews measure — and exactly what this app trains. After every answer you get a short, plain-English explanation of why the best answer wins and where the runners-up fall short. NINE TOPICS, TUNED TO 2026 • GenAI & LLMs — fine-tuning vs prompting vs RAG, cost optimization, guardrails • RAG — retrieval failures, chunking, hybrid search, evaluation • Agents & Tooling — orchestration, tool safety, runaway loops, trajectory evals • Evals — LLM judges and their biases, golden sets, ship/no-ship decisions • Classic ML — leakage, drift, imbalanced metrics, offline-online gaps • Statistics — power, sampling, base rates, calibration traps • A/B Testing — peeking, interference, variance reduction, sample ratio mismatch • Causal Inference — DiD, instruments, confounders vs colliders, uplift • Sampling Design — control groups, labeling budgets, streaming data, real problems from 20 industries IT FINDS YOUR LEVEL — PER TOPIC You can be expert at A/B testing and rusty at causal inference, and the app treats you that way. Each topic carries its own level, and the questions you get are drawn to match it. Answer well and it moves you up. Go cold and it eases off, rebuilds, and brings you back — because being ground down is how people quit. The home screen is the whole point: nine topics, ranked weakest first, each showing where you stand and exactly what would move you up ("2 more correct at Intermediate → Expert"). KNOW HOW YOU'RE WRONG Being close is not the same as being lost. The app tracks whether your misses were near-misses — the right idea, imprecisely — or clearly off, and tells you which one you are, because they call for opposite fixes. Review every question you got wrong, see what you picked and what was right, and watch them leave the list as you put them right. TAILOR EVERY SESSION Weight the topic mix to match YOUR upcoming interview: focus a topic, skip another, or drill a single one. Sessions are 5, 10 or 15 questions — a coffee break, a commute, or a full mock round. PRIVATE BY DESIGN No ads, no tracking, no data ever sold. Your progress and stats live on your iPhone; questions stream from our server. Signing in is optional — everything works without an account. Sharing your answer stats is strictly opt-in (off by default) and serves one purpose: finding questions that are too easy or unfairly hard so we can fix them. Turn it off any time in Settings.