DeepSeek’s Rubik’s Cube Demo Fails to Scramble and Solve Correctly

Matthew Berman examines a cube simulation generated with DeepSeek V4.1 Flash through the Codex harness.

MBMatthew Bermanホスト · Forward Future
チャンネル
Matthew Berman
公開日
2026年9月12日
動画
0:32
読了時間
1 分
スライド
1
原文
英語
YouTube で開く
スクロールすると動画は右下に小さくなって再生を続けます。字幕は閲覧言語に合わせて切り替わります。
閲覧言語2 言語で読めます
この記事はまだ日本語に翻訳されていないため、英語の原文を表示しています。
要約

30 秒でわかる

In this short demonstration, Berman says the generated cube’s pieces appear disconnected, colors change during scrambling without corresponding movement, and the solve attempt fails. The clip shows one result rather than a comprehensive model evaluation.

  1. 1The setupThe model runs through the Codex harness.
  2. 2Disconnected piecesThe individual squares appear to float independently.
  3. 3Scramble and solveBerman observes colors changing without matching movement and an unsuccessful solve.

ポイント

  1. The demonstration uses DeepSeek V4.1 Flash through the Codex harness.
  2. Berman observes apparent independent pieces and color changes during scrambling.
  3. The shown solve attempt does not produce a solved cube.
全文

チャプターごとに整えた全文を読む

文字起こしのみ
01

Inspecting the cube

Berman identifies apparent problems with the simulated pieces.

I show a Rubik’s Cube simulation made using DeepSeek V4.1 Flash through the Codex harness. Looking closely, the squares seem to float independently instead of behaving as connected cube pieces.

02

Scramble and solve

He demonstrates color changes and an unsuccessful solve.

Scrambled cube simulation
スライド 1 / 1Scrambled cube simulationThe demo shows the cube after a scramble; Berman observes that colors change without matching piece movement.

When I scramble the cube, I observe face colors changing without corresponding movement. The subsequent solve attempt does not solve the cube. This is one brief demonstration, not a comprehensive comparison of models.

同じチャンネルの記事

スライド 6 枚25:50EN
MBMatthew Berman· 2026年10月24日Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.Open-weight modelsBenchmarksAI cost & performance+1
スライド 2 枚32:20EN
MBMatthew Berman· 2026年10月23日Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.Local inferenceOpen-weight modelsAI cost & performance+1
スライド0:22EN
MBMatthew Berman· 2026年10月21日A Robot Bodyguard Video Shows Dramatic Rescue ScenariosMatthew Berman · Forward FutureMatthew Berman reacts to a short robot demonstration depicting fighting and vehicle-rescue scenarios.Robotics評価(Evals)
45:08EN
MBMatthew Berman· 2026年10月21日Sam Altman on AI Discovery, Consciousness, Compute and a Human-Centered FutureMatthew Berman, Sam Altman · Forward FutureIn a conversation with Matthew Berman, Altman discusses scientific progress, personal assistants, recursive self-improvement and the importance of human agency.AI for sciencePersonal agentsConsciousness+2
0:32EN
MBMatthew Berman· 2026年10月21日Sam Altman on the Continuing Demand for Software WorkSam Altman, Matthew Berman · OpenAIIn a short interview excerpt, Altman says demand for software may be larger than expected even as AI changes how it is made.Future of WorkSoftware engineeringコーディングエージェント
1:02EN
MBMatthew Berman· 2026年10月15日From Punched Cards to AI Software FactoriesMatthew Berman · Forward FutureMatthew Berman contrasts early programming with an AI-assisted workflow and promotes Factory’s development tools.ソフトウェアファクトリーコーディングエージェントSoftware engineering+1