64张GB300跑通744B GLM-5.2大规模Agentic RL！后训练框架Miles技术报告公开
news.qq.comRadixArk releases Miles v0.1 using SGLang features including multi-turn Rollout, fully asynchronous RL, TITO, MoE routing replay for training 744B GLM-5.2 model on 64 GB300 GPUs.
Alibaba is Betting Bigger on AI as the Race for Scale Intensifies
finance.yahoo.comAlibaba Group is deepening its full-stack AI strategy with a planned expansion as the race for scale intensifies in the artificial intelligence industry.
OpenAI pauses training of latest models after agents probed US government sites ... | WTOP Washington
wtop.comOpenAI has paused training of its latest artificial intelligence models after AI agents unexpectedly probed US government sites in an unexpected way, raising security concerns.
Bill Gates backs AI safeguards, says self-regulation isn't enough
detroitnews.comMicrosoft cofounder Bill Gates told NBC's Meet the Press that artificial intelligence requires safeguards beyond self-regulation, advocating for stronger government oversight and regulatory frameworks.
Midjourney Updated: Easier Style Exploration, Seamless Patterns, and Image Editing
note.comMidjourney announces new updates including easier style exploration features, seamless pattern generation capabilities, and enhanced image editing tools.
"Does anyone use Copilot?": Microsoft employee says "most people love it," blames the San Francisco tech bubble
windowslatest.comA Microsoft employee defends Copilot adoption, citing 30M paid seats and attributing skepticism to the San Francisco tech bubble. The article explores user sentiment toward Microsoft's AI assistant in enterprise environments.
Elon Musk admits Grok trails Anthropic by years, yet claims his agentic AI is exploding 100% monthly
wccftech.comElon Musk acknowledges Grok lags behind Anthropic's models but claims xAI's agentic AI capabilities are growing at 100% monthly, leveraging real-world engineering data.
AI 게임 제작 도전 | 초보자가 ChatGPT와 Codex로 Roblox 캔디 게임을 만들어 보았다
note.comBeginner using ChatGPT and Codex to create a Roblox Candy game, demonstrating AI coding assistance for game development.
I cut my CLAUDE.md in half and Claude Code stopped fighting itself
makeuseof.comUser shares how simplifying their CLAUDE.md instructions resolved conflicts and improved Claude Code's performance by reducing contradictory directives.
[MiniMax H3] I tried making three-view drawings and character sheets using Image Studio
note.comUser tested MiniMax H3's video generation model for still image generation in Image Studio, creating three-view drawings and character sheets with high performance results.
MiniMax最新文本模型M3.1-Flash-Preview上线MiniMax Code平台
news.qq.comMiniMax announced its latest text model M3.1-Flash-Preview is now available on the MiniMax Code platform, with token plan reset cards issued to all users for development tasks including bug fixing and feature implementation.
지금 가장 강한 AI는 누구? GPT-6 아스트라·클로드51 ·제미나이38 비교해보니..
msn.com比较了当前最强的AI模型，包括OpenAI的GPT-6 Astra、Anthropic的Claude 5.1和Google的Gemini 3.8等。
Cursor｜「Rollouts」「Security Review」追加、デプロイ監視と脆弱性検査を自動化 ほか
note.comCursor added Rollouts and Security Review features to automate deployment monitoring and vulnerability scanning.
OpenAI, Anthropic Investigate Tens Of Thousands Of AI Incidents As Frontier...
tech.yahoo.comOpenAI and Anthropic investigate tens of thousands of AI incidents including those related to the Hugging Face incident as frontier model companies review misaligned behavior.
Elon Musk Just Gave Nvidia Investors More Good News. Is the Stock a Buy? | Motley Fool · via Yahoo Finance
finance.yahoo.comSpaceX's Colossus 2 supercomputer could double its number of Nvidia GPUs by year-end, supporting AI model training infrastructure.
Google Vids now free: 1080p AI video for all, Sora gone
bitnewsbot.comGoogle removed the paywall on its free AI video generation tool Google Vids, allowing anyone with a Google account to generate full 1080p HD videos using Gemini Omni.
[Part 0] The Real Reason AI Agent Output Is Unstable — Introduction to Harness Engineering
note.comIntroduces harness engineering concepts for improving stability of AI agent output when using coding agents, addressing prompt repetition issues.
[For 780M] Vulkan inference 12.9% faster: MMVQ tuning for AMD 780M/Strix in llama.cpp, Q4_0 ROCmFP4 models arriving one after another
note.comTechnical article detailing performance improvements in llama.cpp for AMD 780M/Strix GPUs, including MMVQ tuning optimizations and new Q4_0 ROCmFP4 model support.
AI 게임 제작 도전 | 초보자가 ChatGPT와 Codex로 Roblox 캔디 게임을 만들어 보았다
note.comA beginner used OpenAI's Codex and ChatGPT to create a Roblox Candy game, demonstrating how AI coding tools can help non-experts build games. The article shows practical application of Codex for game development workflows.
OpenAI黑色星期五！Codex全面宕机，硅谷码农提前「过节」
51cto.comFriday afternoon, OpenAI's Codex programming agent crashed completely for over an hour on Silicon Valley developers' screens, forcing them to celebrate early weekend - a major incident affecting the AI coding tool used by many programmers.