DevPick

OpenAIのFFTアルゴリズムは画期的な突破というより悪意あるエクスプロイトに近いと指摘した技術的な検証記事

• •OpenAIのFFTアルゴリズムに関する技術的な検証が行われた
• •このアルゴリズムは画期的な突破というより悪意あるエクスプロイトに近いと評価された

Research shows that the assumption of invoking hidden functions collapses under adversarial conditions, demonstrating a method to embed encrypted neural networks via backdoor attacks.

• •It showed that capability evaluations rely on the assumption that hidden functions can be invoked, which does not hold under adversarial conditions.
• •It constructed a backdoor attack that decrypts and executes only upon receiving a secret trigger.
• •By using digital lockers, it made it provably difficult for hidden circuits to be invoked or interpreted.

An article explaining defensive measures that reduce the occurrence of malicious distillation by over 90% to prevent unauthorized copying of frontier AI capabilities.

• •Malicious distillation allows model developers to copy frontier AI capabilities without proportional safety measures.
• •The article explains specific defensive measures taken to reduce the occurrence rate of distillation by more than 90%.
• •It indicates that developers have strengthened countermeasures against the risk of unauthorized replication of frontier AI capabilities.

Rustで開発されたオープンソースのパラメトリック3D CADアプリケーション「SolidCraft」が公開され、デスクトップ・Web・自動化への対応が示されている

• •SolidCraftは純粋なRustで構築されたオープンソースのパラメトリック3D CADアプリケーションである
• •デスクトップ環境およびWeb環境での利用に対応している
• •MCP（Model Context Protocol）を通じて自動化機能を提供する

MCP（Model Context Protocol）をサポートする実験的なDAW（デジタルオーディオワークステーション）「blokkily」の開発プロジェクト

• •MCPをサポートする実験的なDAWである
• •パターンファーストの設計思想を採用している
• •GitHub上のandrewfader/blokkilyリポジトリで開発が進められている

An article introducing a modification case where an unused Kindle is repurposed as a photo frame using an AI agent.

• •An unused Kindle was converted into a photo frame using an AI agent.
• •As stated in the title, a model called GPT-6 Astra was involved.
• •The Kindle was jailbroken to remove restrictions.

Overview of Janas, a tool that executes large LLMs exceeding RAM capacity via SSD streaming, achieving faster inference than llama.cpp.

• •Qwen3-Next-80B (48GB) was processed at 37 tok/s on a 32GB laptop, significantly outperforming llama.cpp's 2.8 tok/s.
• •All 13 models compared in Q4_K_M quantization format recorded faster read speeds than llama.cpp.
• •Specific performance figures were presented: 24 tok/s without drafting and 62 tok/s for reading.

An article argues against Rich Hickey's claim that AI is not a tool, stating that automation and tool use can coexist and that effective LLM operation requires skill and judgment.

• •It clarifies the position that automation and tool use are not mutually exclusive and can coexist.
• •It points out that handling LLMs appropriately requires skills, feedback, and judgment, not just operation.
• •It argues that AI should be viewed as a complex entity involving human capabilities, rather than dismissed as a mere mechanical tool.

Intel announced the D270, a data center GPU for inference, enabling agent AI, long-context, and high-parallel serving at low cost.

• •It is a data center GPU for inference optimized for agent AI, long-context, and high-parallel serving.
• •It is designed to operate with practical power consumption and cost.
• •It features up to 480 GB of LPDDR5X memory.

Epic Games was reported to have instructed employees to use AI in the development of Disney-related games.