Choose a cube
Top row first, left to right, then the middle and bottom rows.
Cube C11
3D view
Drag to orbit. Scroll or pinch to zoom.
This browser could not start WebGL. The complete 2D net remains available.
Unfolded 2D view
Each matrix follows the recorded installation directions.
Face evidence
Completeness and evidence type are separate: every cube is complete, while some hidden faces are reconstructed rather than directly seen.
All twelve cubes
Drag to orbit the shared, flat 4×3 installation grid. Every column is vertically aligned, every cube stays rigid in its photographed orientation, and wider spacing keeps the side faces readable. Gold cradles mark C21 and C24.
Camera buttons choose a side. Grid rotation turns the entire rigid installation around the current viewing axis.
This browser could not start the all-cubes WebGL view. The complete unfolded overview remains available.
Girn's Special View
Girn's alternate A'–L' reconstruction, rendered directly from his supplied six-face matrices in the supplied 4×3 layout. Gold cradles mark E' and H'.
Camera buttons choose a side. Grid rotation turns the entire rigid arrangement around the current viewing axis.
This browser could not start Girn's Special View.
All twelve unfolded
Every face is spread open in the same installation coordinate frame.
Text notation
Agent-ready cube states generated from the same data as both visual views.
Girn Mode
Copyable JSON in the supplied A–L frame. Letters run left-to-right through each row: A=C11 through L=C34.
Face orientation follows the three reference views. H keeps its black sticker as separate physical metadata; the covered color remains blue.
Coordinates and provenance
Data snapshot: problem-5/derived/face-transcription.json, schema version 1,
updated 2026-08-08. Cube IDs use Crc: row 1 is the top row, row 3 is the
bottom row, and columns run installation-left to right.
- U: rear-to-front, left-to-right.
- D: front-to-rear, left-to-right.
- L: up-to-down, rear-to-front from outside.
- Front: up-to-down, left-to-right.
- R: up-to-down, front-to-rear from outside.
- Rear: up-to-down, installation-right to left from outside.
Front, right, and Rear faces were independently read from the supplied photographs. The new high-resolution rear-side set confirms all 108 previously legality-forced Rear stickers without a discrepancy. Some U/L faces retain the teammate transcription after cross-view reconciliation. D faces C22–C34 come from a teammate bottom-face transcription whose 53 previously recorded sticker cells all matched existing evidence. Unknown stickers, if introduced by a future data update, appear gray with a question mark.
Across all twelve cubes, the evidence ledger covers 423 directly photographed stickers, 153 reconciled U/L stickers, 63 teammate D-face stickers, and 9 C21 photo-patch/legal D stickers. All 648 literal stickers are known and every literal six-face layout is unique. The gold-mounted C21 and C24 states become identical only after center-relative color normalization.
3D rendering uses Three.js r180, vendored locally under the MIT license for compatibility with this site's security policy.
Solve counts come from problem-5/derived/solve-lengths/compute.mjs, which
runs the min2phase two-phase solver on these exact reconstructions and replays every
maneuver back onto the state before accepting it. Counts use the half-turn metric and
are upper bounds: correct solutions, near-optimal, not proven minimal.