English 日本語

Tech News

Current developments in generative AI, software engineering, programming, and digital products, with links to original sources.

Linus Torvalds Puts AI in Perspective, DeepSeek-V4-Flash Local Optimization Advances, EU AI Transparency Rules Take Effect — AI News, August 3

AIニュースDeepSeekAI規制

OpenAI Unveils 'Presence' for Enterprise Agents, Tencent Hy3 Accelerates Openness by Going Apache, and 94.8% of Sites Never Get Cited by AI — AI News for August 2

AIニュースOpenAIオープンソースAI

Meta AI Takes on Agent 'Forgetting' with a Memory-Assist AI, While Apple's Bug Bounty Breaks Down Under AI Slop — AI News for August 2

AIニュースAIエージェントセキュリティ

AI-Discovered Vulnerabilities Are Rarely Exploited, UK Debate on Whether Not Using AI Is Negligence, and WinterMix Quantization for MLX — AI News, August 2

AIセキュリティローカルLLMAIコーディング

Claude Opus 5 Generates 3D Games From a Single Prompt, METR Calls for Independent Investigations of AI Agent Misconduct, and Realistic Limits Emerge for Qwen's Autonomous Coding — AI News, August 2

AIニュースClaudeAIエージェント

The Trap of "Combining" Inference Speedups Backfiring, Is Dividing Labor Between Strong and Cheap Models the Right Answer, and Snap and LinkedIn's AI Slop Countermeasures — AI News for August 2

AIニュースLLM推論最適化

A C99 Inference Engine That Runs Kimi K3 on a CPU and 8GB of RAM, a Local Follow-up on DeepSeek-V4-Flash, and a Model Price War Stoking Debate — AI News for August 2

ローカルLLMDeepSeekKimi K3

Gemini Spark Lands in Japan, Local Benchmarks for DeepSeek V4-Flash Arrive, and China's 'Token Diplomacy' — AI News for August 2

AIニュースGeminiDeepSeek

What Makes Claude Code 'Skills' Different from Plain Markdown? Coding Bias in Benchmarks and the Limits of Multi-GPU Inference Also Discussed — AI News, August 2

AIニュースClaude CodeLLM

OpenAI Surpasses 1 Billion Users and Cuts GPT-5.6 Prices; Booksellers Warn of Rare Book Destruction for AI Training; AI's Job Threat Is Wage Decline, Study Suggests — AI News for August 2

AIニュースOpenAIAI倫理