← 返回最新日报
2026-10-13· 当日存档
AI 日报
AI狂飙与隐忧并存
今日看点密集：Meta首款前沿模型Muse Wonder杀入第一梯队，阿里QwenGyre攻克超长任务智能体；另一边，Character.AI申请破产、Claude误报谋杀线索、模型可隐藏能力等安全与商业风险集中爆发，行业正经历冰火两重天。
模型动向
1
Meta's new frontier model Muse Wonder Full breaks into the AI top tier
Meta首款前沿模型六项基准进前三，智能体提升最大但成本近三倍
The Decoder深挖 →
2
Meta releases its first frontier model, Muse Wonder
Meta正式发布Muse Wonder，比前代更大更强，基准全面领先
TechCrunch AI深挖 →
3
Alibaba Qwen Introduces QwenGyre: Elastic Reinforcement Learning RL for XLong-Horizon Agents
阿里QwenGyre弹性强化学习，让智能体跑数小时近百万token任务
MarkTechPost深挖 →
4
1M-token, multimodal decision models are here! But there is more to this one. You can classify, extract, and summarize with 1M tokens of text and imag...
ScaleDown推出1M token多模态决策模型，快200倍便宜467倍
X · List深挖 →
5
Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB
Z.ai法律模型GLM-Legal低成本超越Claude与Kimi
MarkTechPost深挖 →
产品上新
6
Apple launches HomeView, its first smart display for the home
苹果首款智能家居显示屏HomeView发布，集视频通话与信息浏览
The Verge AI深挖 →
论文研究
7
Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed
DeepMind AlphaProof Nexus解出数十道未解数学难题，成果经同行评审
The Decoder深挖 →
8
9
Clearing chat history helps AI agents find better solutions
微软发现清空聊天历史可破智能体思维定式，效果提升66%
The Decoder深挖 →
10
Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots
Sakana AI自改进模型两轮胜率近70%，但盲点被误当进步强化
The Decoder深挖 →
行业动态
11
All Systems Go: Global PC App Ecosystem Ready for NVIDIA RTX Spark
英伟达RTX Spark超级芯片发布，融合游戏创作与个人AI超算
NVIDIA深挖 →
12
Just 5 years in, interactive chatbot platform Character.AI files for bankruptcy
Character.AI成立五年即破产，拟2000万美元向迪士尼出售资产
TechCrunch AI深挖 →
13
Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation
Anthropic称推理token窃取降99%，指认智谱月之暗面蒸馏并封号
The Decoder深挖 →
14
Anthropic study warns that small AI models could soon replicate across everyday devices
Anthropic警告小型AI模型或数月内可在消费设备间自我复制
The Decoder深挖 →
15
Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027
微软2027年将自研Maia芯片扩至多吉瓦，能效提升40%
The Decoder深挖 →
16
Anthropic's Claude AI submitted false murder tip to Philadelphia police
Claude误向费城警方提交虚假谋杀线索，AI安全再敲警钟
Ars Technica AI深挖 →
17
Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
微软发布第三代自研AI芯片Maia 300，降低对英伟达依赖
The Verge AI深挖 →
18
Anthropic's Andrej Karpathy says "the external world has seen the preview" of what AI can actually do
Karpathy称公众只看到AI预览，约5000人已用内部系统做专家工作
The Decoder深挖 →
19
Amex introduces a framework to prevent AI agents from making unauthorized purchases
美国运通推出框架，确保AI代理购物仅限用户授权范围
TechCrunch AI深挖 →
20
AI labs are buying data from dead startups to train their models
AI实验室收购倒闭创业公司数据训练模型，前创业者靠卖数据获利
MIT Tech Review深挖 →
21
From Earth to Orbit, France’s Ecosystem Delivers Agentic Computing at Scale With NVIDIA
法国借助NVIDIA部署智能体计算，从航天到金融落地
NVIDIA深挖 →
22
实用技巧
23
24
25
Useful tip for anyone using SeeDance 2.5 It's easy to remove unwanted (and unprompted) music from a video with Met's SAM Audio. https://aidemos.meta.c...
用Meta SAM Audio一键去除SeeDance 2.5视频多余音乐，保留音效
X · List深挖 →
26
活动会议
27
Jensen just entered the stage. Very interesting speech about: - era of AI agents - DGX spark - why it took 4 years for Nvidia to work on the collabora...
黄仁勋登台演讲，谈AI智能体时代、DGX Spark及与微软合作内幕
X · List深挖 →
快讯
·
Character.AI files for bankruptcy
The Verge AI
·
Open models have real benefits. Do they outweigh the risks?
MIT Tech Review
日报存档
2026 年 10 月
13 日AI狂飙与隐忧并存12 日AI狂飙与认知鸿沟11 日AI失控与安全警钟10 日AI代理失控警钟9 日AI智能体全面提速8 日GPT-6 智能 UI 上线7 日万亿模型与AI自治争议6 日开源模型与资本双线爆发5 日AI侵蚀大学信任4 日OpenAI安全震荡与验证器研究3 日AI代理安全与落地并进2 日AI能力跃迁与安全隐忧1 日谷歌Gemini 4发布引三强竞速2026 年 9 月
22 日小米开源登顶，AI失控警钟21 日开源图像模型与AI治理并进20 日AI安全与失控风险成焦点19 日AI安全失控与模型竞速18 日AI对齐警报与芯片竞赛17 日AI安全警报与产品整合潮16 日谷歌苹果齐发AI新品15 日苹果Siri重构与AI放缓之争14 日AI巨头放缓迭代引激辩13 日GPT-6 Astra全面突破12 日AI数学错位与安全风波11 日智能体基建与模型竞速10 日GPT-6登场，AI安全恐慌升温9 日AI数学突破引热议8 日AI日报：端侧模型与欧洲融资双热点7 日AI日报：模型竞赛与行业震荡6 日GPT-6 Astra发布，AI竞争白热化5 日GPT-6发布，AGI时代开启4 日GPT-6 Astra开启AGI时代3 日英伟达巨资收购HF2 日AI日报：世界模型与安全并进1 日AI日报：模型突破与行业震荡