-
1
GPT-5.6三档模型全量开放
分Luna、Terra、Sol三档定价，此前因监管审查限量预览两周，现获政府批准正式全量上线
OpenAI rolls out GPT-5.6 after government greenlight — and announces 'ChatGPT Work' ↗4.6 -
2
Meta推编程模型对标GPT-5.5
新代理编程模型经Meta Model API开放，号称HLE跑分达SOTA，1M上下文，定价低于同类竞品
Meta enters the crowded AI coding battle with Muse Spark 1.1 ↗4.4 -
3
ChatGPT Work上线做长任务代理
新agent功能可跨应用、跨文件长时间执行任务，将目标直接转化为可交付成果，对标Claude Cowork
ChatGPT is now a partner for your most ambitious work ↗4.4 -
4
4.2
-
5
Fidji Simo卸任OpenAI二号职务
因医疗休假延长，OpenAI应用业务负责人转为兼职顾问，正值公司筹备IPO关键节点
Fidji Simo steps down from leading OpenAI's AGI work due to illness ↗4.0 -
6
GPT-5.6继续力挺微软Copilot
新模型将继续驱动微软365全家桶办公应用，双方近期屡传关系生变传闻
OpenAI says GPT 5.6 is the 'preferred model' for Microsoft Copilot 365 amid breakup chatter ↗3.8 -
7
阿里禁用Claude Code防后门
据消息人士称，阿里巴巴以疑似存在后门安全风险为由，在内部工作场景中禁止使用Claude Code
Alibaba to ban Claude Code in workplace over alleged backdoor risks, source says ↗3.8 -
8
LingBot-Video开源具身视频模型
30B MoE仅激活3B参数，融合7万小时具身数据，RBench均分0.620超越现有开源与闭源模型
LingBot-Video: first-of-its-kind open-source embodied video generation model ↗3.8 -
9
NYT指控OpenAI版权案藏证据
新闻出版商指控OpenAI隐瞒可识别侵权新闻的工具与数据集，已提交新的制裁动议升级诉讼
New York Times says OpenAI hid evidence in ChatGPT copyright trial ↗3.6 -
10
马斯克承诺不断供Anthropic算力
Anthropic年收入约400亿美元依赖其算力托管，马斯克公开表态不会切断供应以换取信任
Elon Musk praises Mythos/Fable, promises not to 'cut off' Anthropic ↗3.6 -
11
OpenAI关闭Atlas浏览器不到一年
去年10月发布的AI代理浏览器停止运营，部分自动化浏览功能转移至桌面应用与Chrome扩展
OpenAI is shutting down Atlas, but its AI browser ambitions are still growing ↗3.5 -
12
Claude推出年度使用报告Reflect
新Reflect仪表盘可视化用户AI使用习惯，同时强化用户对Claude日常工作的依赖感
Anthropic's new Claude feature is quietly selling you on AI ↗3.4 -
13
744B大模型塞进25GB内存跑通
744B参数MoE模型GLM-5.2经量化后可在仅25GB内存的消费级硬件上运行，本地大模型社区实测
GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine ↗3.4 -
14
SWE-1.7逼近GPT-5.5/Opus水平
Cognition发布编程模型SWE-1.7，宣称智能水平接近GPT-5.5与Opus 4.8，HN获269赞
SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence ↗3.4 -
15
Codex推理token聚集疑致性能下降
开发者在GitHub issue中报告GPT-5.5 Codex推理token出现聚集现象，疑似拖累实际表现，HN获370赞
GPT-5.5 Codex reasoning-token clustering may be leading to degraded performance ↗3.4 -
16
谷歌广告将标注AI生成痕迹
My Ad Center新增标签，标明Search、Discover、YouTube广告是否经AI生成或编辑，此前仅选举广告强制披露
Google will now tell you if an ad was made with AI ↗3.2 -
17
Claude Mythos发布期漏洞数据异常
Epoch.ai数据显示，Claude Mythos Preview发布前后严重CVE数量明显上升，HN获156赞
New serious vulnerabilities spiked around release of Claude Mythos Preview ↗3.2 -
18
3.0
-
19
AI创企让自家agent操盘1亿融资
企业级AI agent厂商Lyzr用自研agent主导完成1亿美元融资流程，作为产品能力的实测案例
An AI agent startup just let its agent run its $100 million fundraise ↗2.9 -
20
AlphaEvolve面向谷歌云客户开放
DeepMind算法发现系统AlphaEvolve扩大到谷歌云企业客户，用于芯片设计、物流路由等复杂优化问题
We're rolling out AlphaEvolve widely to solve Google Cloud customers' hardest problems ↗2.8 -
21
Meta自研AI芯片9月投产
Meta采用模块化设计自研AI芯片，预计9月开始量产，为应对AI需求快速演变预留迭代空间
Meta's new AI chips will begin production in September ↗2.8 -
22
开发者吐槽Fable安全分类器过严
博客作者批评Anthropic为Claude Fable设置的内容分类器过度拦截，影响正常使用，HN获214赞
The classifiers Anthropic puts in front of Fable are too zealous ↗2.8 -
23
SkillCenter当AI技能应用商店
抓取GitHub技能仓库并提取SKILL.md，生成可搜索索引，一条命令即可为多种agent接入现成技能
SkillCenter – a new package manager for agent skills ↗2.7 -
24
语音AI创企Gradium获1亿种子轮
巴黎AI语音初创Gradium获英伟达参投1亿美元种子轮，资金将用于开设湾区办公室吸引人才
Paris-based AI voice startup Gradium raises $100M seed, backed by Nvidia ↗2.7