Screenshots
About System Design Arcade
NEW: THE AI ARCADE. Eight games on the system design round that didn't exist three years ago: RAG, LLM serving, embeddings, guardrails, and GPU fleets. Free, no subscription, and it plays offline.
The system design interview is the one where the whiteboard goes blank. You've read the books. You've watched the videos. But when the interviewer says "design Instagram," or now "design a retrieval system for our support bot," you freeze, because reading about consistent hashing at 2am is not the same as knowing it cold.
System Design Arcade is the fix. It turns every building block into 60-second arcade levels you can actually finish. On the subway. In line for coffee. While waiting for CI. On a plane with no Wi-Fi.
THE AI ARCADE
AI engineering is its own interview round now, and most prep hasn't caught up. The AI Arcade drills the systems behind it:
- Design a RAG System: chunk boundaries, hybrid retrieval, reranking, and the permission bug that ends careers
- LLM Serving: batching, KV caches, streaming, and GPU memory as the binding constraint
- Feed Ranking: retrieval, ranking, and the feedback loop that decides what a model can ever learn
- Embeddings: what "similar" means, how many dimensions you need, and migrating models without silent breakage
- Feature Stores: one definition, two readers, and point-in-time correctness between training and serving
- ML Observability: catching a model that is wrong while every dashboard stays green
- AI Guardrails: where the checks sit, prompt injection from untrusted text, and what a model is allowed to reach
- GPU Infrastructure: memory, interconnect, scheduling, and the bill
FIVE TRACKS, ONE CURRICULUM
Concept Arcade: master one building block at a time. Caches, databases, queues, replication, sharding, load balancing.
Problem Arcade: design complete systems end to end. URL Shortener, Rate Limiter, News Feed, Chat, Ride Sharing, and a new Key-Value Store. The classics, played out like boss fights.
Tool Arcade: hands-on drills with the infrastructure you'll use on the job. Redis. Kafka. Docker. Kubernetes. And Vector Vault, the vector store behind semantic search.
AI Arcade: the eight games above.
Story Mode: narrative capstones that walk an onsite round end to end: a Twitter-style home timeline, a chat system, Dropbox, typeahead search, a rate limiter, and a URL shortener.
CHALLENGE TYPES THAT DON'T FEEL LIKE FLASHCARDS
Multiple choice, ordering puzzles, fill-in-the-blank, API Builder, Architecture Builder, and Scenario-Why drills. Each level mixes formats so your brain has to actually engage. The Scenario-Why drills train the reasoning behind the answer, not just the answer.
REVIEW THAT SPACES ITSELF OUT
Miss a challenge and it comes back in Power-Up Review. Clear it and it returns on a widening schedule, so what you learn stays learned.
DAILY CHALLENGE
A fresh challenge every day, the same one for you and your study buddy. Streak counter. Optional reminder notifications, opt in from Settings.
BUILT FOR YOUR EYES
Retro arcade aesthetic by default. Pixel fonts. Neon glow. Prefer something softer? Switch to the playful sticker-card vibe, or open the Theme Lab and tune background, glow, gradient, and font until it feels right. Read Aloud narrates every explanation, so you can learn while you walk.
YOUR PROGRESS STAYS YOURS
Start playing instantly: Level 1 of every game needs no account. A free sign-in with Apple, Google, GitHub, or email unlocks every level and backs up your progress across your iPhone and iPad. Progress saves on your device. We use product analytics to see which levels need work, never the content of your answers.
FREE TO PLAY
No ads. No paywalls. No subscription. Every level is free with a free account. New games ship every sprint.
Start with Level 1.
AVALW Coverage
Not yet mentioned in AVALW articles.