Historical reference and clean-room boundary

The black-box reference for this lesson is the historical LUMON MDR / Severance Refiner, available at the archived public artifact. I studied only its observable interaction: a field of numbers, a mysterious selection task, immediate selection feedback, scoring, and a quota-shaped progression loop.

Clean-room boundary: Pattern Quota does not copy source code, text, names, artwork, layout, palette, sound, or other expressive material from the reference. The new work uses original Parallax Registry branding, an ocean-and-coral visual system, mirrored-code rules, fresh copy, a seeded generator, and an independently designed state model. The historical title and source link live here for provenance; they do not appear inside the clean demo.

The exact build prompt

Using the public number-refinement game only as a black-box behavioral reference, create an original retro-corporate pattern-selection game with generated number tiles, hidden target flags, selection feedback, accuracy scoring, batch animations, a five-round quota and restart. Invent entirely new branding, language, palette and assets.

What was rebuilt

The shared mechanic is intentionally narrow: inspect a fixed batch, choose tiles that appear to satisfy a hidden classification, submit, study the result, and work toward a finite quota. Pattern Quota turns that loop into a transparent learning system. Its briefing defines an original “echo code” rule, each 24-tile batch contains exactly six generated targets, and review separates correct selections, incorrect selections, and missed targets before calculating accuracy.

Under the surface, a seeded generator makes every batch reproducible. Passing a batch advances one of five quota segments; failing creates a retry without incrementing progress. The completion panel can restart the full run. A semantic window.PatternQuota API, strict state capsules, and a visible self-test make the behavior inspectable without coupling automation to screen coordinates or CSS selectors.

Why the safety and branding boundary matters

Clean-room work is not a cosmetic rename. It distinguishes an abstract behavior worth learning from the protected expression that made the original memorable. By documenting provenance while replacing identity, copy, visuals, assets, rules, and code, the lesson can teach deterministic UI architecture and state validation without presenting someone else’s creative work as reusable material.

The same boundary improves engineering quality. With no external assets, services, or hidden dependencies, the demo is auditable as one file. Its restrictive Content Security Policy disables network connections, its renderer uses safe DOM APIs and textContent, and imported JSON is treated as untrusted data that must prove every score and transition before it can replace live state.

Acceptance checks

  • Each batch contains a deterministic number of selectable tiles.
  • Submit is disabled until at least one tile is selected.
  • Accuracy is computed from correct, incorrect and missed targets.
  • Successful rounds advance the quota and failed rounds do not.
  • Completing five rounds shows a restartable completion state.