Stop Drawing Boxes. Start Defending Designs.

Most system design practice looks the same: open a blank doc, draw a few rectangles, name them “API,” “Cache,” and “DB,” then hope the interviewer is kind.

Real interviews are not kind. They ask why Redis here, what breaks at 10× traffic, and what you’d cut under time pressure. Reading architecture blogs helps. Drawing alone doesn’t. You need a loop that forces tradeoffs.

That’s what System Design Dojo is built for.

Draw it. Argue it. Survive the follow-up.

System Design Dojo is an interactive course on Codeground AI where every lesson is a full practice loop:

1. Clarify — functional and non-functional requirements (QPS, latency, consistency, cost)
2. Sketch — build the architecture on an in-browser canvas
3. Note the tradeoffs — APIs, data model, sync vs async, what you’d cut
4. Get AI critique — staff-engineer-style pushback on strengths, gaps, and alternatives
5. Answer the follow-up — one sharp question that mirrors a real interview

No whiteboard photo uploads. No “paste your notes into ChatGPT and hope.” The scenario, canvas, design notes, and critique live in one place.

What a lesson feels like

Take the classic opener: Design a URL Shortener.

You’re not asked to invent bit.ly from vibes. You get interview-shaped constraints:

• 100M new URLs / month
• 10k redirects / second peak
• 99.9% redirect availability
• Links that aren’t trivially guessable

On the left: the brief, hints, and a design-notes field for APIs, data model, and tradeoffs. On the right: an architecture canvas with the usual system-design vocabulary — services, databases, caches, clouds, arrows.

Then you hit Get AI critique.

The point isn’t a green checkmark for “correct topology.” There usually isn’t one. The critique rewards clear reasoning: ID generation strategy, redirect path, cache-before-DB, analytics off the hot path, and where uniqueness breaks under write load.

A strong answer sounds like an argument. A weak one sounds like component bingo.

What’s inside the course

Roughly ~10 hours of guided practice across classic and harder scenarios:

ModuleScenarios Foundations Design interview playbook Classic systems URL shortener, Pastebin, distributed rate limiter Realtime & social Chat, news feed, video streaming Platform patterns Notifications, search autocomplete Capstone Ride sharing

Each lesson has a rubric focus — caching, data ownership, bottlenecks, failure modes — so you practice what interviewers actually probe.

Why AI critique changes the practice

Self-study fails for a simple reason: your own diagram never argues with you.

In Dojo, the AI acts more like a tough interviewer than a cheerleader. It surfaces:

• Strengths — what already holds up
• Gaps — missing ownership, hot keys, failure paths
• Alternatives — other designs worth defending
• Scores — so you can see where the argument is thin
• One follow-up question — the kind that decides hire / no-hire

Example energy from the playbook lesson: A cache stampede hits your origin after a deploy — what do you change in the next 30 minutes?

That’s the muscle you want before a panel, not after.

Who it’s for

• Engineers preparing for mid/senior system design rounds
• Backend / full-stack folks who can code but freeze on the whiteboard
• Anyone tired of passive “watch someone else design Uber” videos

If you can draw boxes, you can start. If you can defend them under pressure, you’ll finish stronger.

Try it

Open the course, pick a scenario, sketch a real request path, write the tradeoffs, and take the critique:

https://codeground.ai/courses/system-design-dojo

Whiteboard energy. Browser-native canvas. An interviewer that pushes back.

Draw it. Argue it. Survive the follow-up.