Learn

Architecture & pattern guides

Read the idea, build it on a Grid, prove it in a challenge. Short, practical guides — illustrated with RodGrid's own nodes — on system design, design patterns, flow modeling, and how to get the most out of the app.

All articles

How Challenge Mode works

Pick an exercise, build against a curated palette, ask for hints, submit for AI scoring.

Ten minutes and six nodes. That is what the card for the Observer Pattern Challenge promises, and those two numbers are most of the difference between a Challenge and a blank Diagram Grid: something keeps time, something counts, and at the end an AI reads what you drew. One attempt at Observer shows the whole loop.

From the card to the Grid

Challenges sorts its exercises into four tabs, Architecture, Patterns, Flow and Low-Level Design. Observer sits under Patterns, and a Learn article's Practice this takes you straight to a Challenge's card. Starting costs nothing on any plan; only the AI steps cost credits.

  • Start practice opens a Challenge Grid for that exercise and starts the clock. The card of an attempt you already have open shows Continue instead.
  • A Guided badge, as on URL Shortener, offers Start guided and Start classic. Classic is what this article describes; for the badge itself, read Guided challenges: four stages, one check each.

Most attempts begin on an empty Grid; the ones titled Diagnose begin with a system already in trouble.

Only one exercise can be active at a time, and neither its card nor the banner that counts it down can abandon it: the clock counts from the start whether the Grid is open or not, so you submit or wait it out. If time runs out the panel says Time's up, the attempt closes without a score, and the Grid stays only to look at.

Eight components, six places

The palette on a Challenge Grid holds only what this exercise offers. Observer's has eight: Publisher / Subject, Observer, Event, Handler, Actor, Singleton, Strategy and Saga. The brief asks for a subject, its observers and the relationships between them. Nothing in it asks for a Saga.

That mix is the curated palette: enough to build the design, often with components that tempt you away from it, because choosing what to leave out is part of the exercise. An extra node is not wrong just for being there. The grader is told to judge what it does in your design, and to file a defect when its role is wrong for this Challenge.

Pasting from another Grid does not widen the palette, and importing is refused. The panel keeps the budget in view as Nodes 4 / 6, and at the limit a new node is refused until you remove one. Reaching it means simplify the design.

Three nodes, all called Observer

Place three observers and they all read "Observer". The grader sees each node's type and name, which node each connection starts and ends at, and each connection's label. It does not see positions, sizes or arrowheads, so a design with default labels gives it little to credit, and the panel's own tip says clear labels score higher. On Observer, whose brief asks you to show subscribing and notifying, that means "Email Observer" and connections labelled "subscribes" and "notifies", and the evaluation then refers to your nodes by those names. Double-click a node or a connection to label it; the rest of naming is in Naming nodes: inline rename & the inspector.

What a hint can see

Stuck at minute six, Hint (2 cr) gets you two or three sentences that nudge without handing over the solution, naming only components on this palette. It knows the brief, the palette and which components you placed, by their catalogue names. It does not read your labels or your connections, so it can point at a missing part but not at a wrong connection. Each hint is counted on the attempt and the grader is told the number; no formula takes points off for it.

Two ways to submit, and no upgrade

Two buttons wait under the brief, and neither works until the Grid has a node. The one you press reads Evaluating… while the grade runs, and one you cannot afford says how many credits you are short.

  • Quick review (4 cr) is graded by one AI reviewer and stays off the leaderboard.
  • Submit ranked (8 cr) runs two reviewers, costs exactly double, and can be ranked.

Both return a scored rubric with strengths and improvements. Choose before you click, because a graded attempt is closed: a quick score cannot be promoted to a ranked one. A ranked result takes a new attempt, on a new Grid, with a new clock.

All of it is paid in credits, 2 for a hint and 4 or 8 for a review. On a paid plan they come first from a monthly challenge pool, kept apart from your AI credits, and once that pool runs out Challenge actions fall back to your AI credits; AI credits: what they buy and how they are spent shows where they come from. Reading the result, defending your design, your nickname and the leaderboard are all in After you submit: the score, the defense and the reference.

Ten minutes on Observer

Start Observer Pattern from the Patterns tab, label every node and connection before the clock gets tight, and send it for a quick review. Then read all of the feedback and make one thing it asks you to change the focus of a second attempt. That attempt opens on an empty Grid, so rebuild the design around that change before you spend credits on another review.

Practice this