Grid Rush
Cognitive training with online ranking
Schulte-table-style exercises exist scattered across dozens of standalone pages, with no accounts or progression. Grid Rush brings them into a platform with its own identity: play without registering, and creating an account gets you a global leaderboard and sync. Behind it is a data layer with real security, not an open table.
Problem
Product: unify scattered cognitive exercises into a platform with measurable progression and social comparison, without making registration a barrier to entry. Technical: making sure a public, cloud-hosted leaderboard isn't an open table anyone can write whatever they want to.
Solution
- Guest first: everything playable with no account; local stats always work. The account adds, it never gatekeeps.
- A data layer with real security: Row Level Security on every table, revoked privileges on the scores table, and a single write path — a server-validated function that checks authentication, ranges, and allowed values.
- Platform architecture: a declarative game catalog; adding a game doesn't touch the others.
- Graceful degradation: with no backend configured, no session, or no connection, the app is still playable.
- Details that almost always get skipped: full i18n with per-profile persistence, light/dark/system themes, reduced motion for motion-sensitive users, guided onboarding, complete password recovery.
Architecture
The first version of the leaderboard stored a single global score — making a 3×3 board and a 6×6 board incomparable. I recognized the design flaw and carried out a full refactor (schema, RPC, view, UI) including a data migration. I documented the system's real limit in writing: without an authoritative game backend, server-side validation protects the write path but isn't anti-cheat. Acknowledging a system's own limits is part of professional judgment — the project is presented as actively in development, not as a finished product.
Gallery
