1海外Simon Willison3 小时前实践75
Using Blender with coding agents on macOS
在Mac上让编码代理用Blender渲染3D场景的简单方法。
2海外The Decoder5 小时前实践82
OpenAI shares prompting tips for GPT-6 Astra including a blocklist of slop words
OpenAI发布GPT-6 Astra提示词指南，含禁用词清单，指导开发者提升模型主动性。
3海外The Decoder6 小时前研究82
Seven minutes with a chatbot beat a fact sheet at reducing conspiracy beliefs in two experiments
研究发现与AI聊天七分钟比阅读事实清单更能减少阴谋论信念，效果持续数周。
4海外The Verge AI7 小时前行业78
OpenAI admits to German wiki ‘incident’
OpenAI承认AI代理攻击德国维基事件，承诺改进报告机制。
5海外The Decoder8 小时前行业82
OpenAI admits its disclosure practices need work after its autonomous agents hacked a German wiki
OpenAI承认自主智能体攻击德国wiki事件暴露披露机制不足，将推出新框架。
6海外The Decoder8 小时前研究85
Deepmind put 100 AI agents in a room and they sorted into cheaters, converts, and whistleblowers
DeepMind模拟100个AI代理协作，结果出现作弊、告密等社会行为。
7海外Hacker News11 小时前实践85
AI handles incidents, engineers lose touch with their systems
AI处理事故致工程师与系统脱节，引发运维技能退化担忧。
8海外The Decoder11 小时前模型85
OpenAI rolls out GPT-6 Astra to top-tier ChatGPT plans at half the rate of GPT-5.6 Sol
OpenAI向高端用户推出GPT-6 Astra，消息额度约为GPT-5.6 Sol的一半。
9海外MarkTechPost12 小时前产品82
Adaption Labs Introduces ‘Invent a Dataset’: Training Data Generated From a Task Description, Not a Seed Corpus
Adaption Labs推出Invent a Dataset，可从任务描述直接生成训练数据集，无需种子语料。
10海外MarkTechPost14 小时前产品88
Google Launches Agentic Video Understanding for Gemini Flash Models, Cutting Video Tokens by Up to 88%
谷歌为Gemini Flash推出智能视频理解，按需加载片段，视频令牌最高减少88%。
11海外MarkTechPost15 小时前产品85
NVIDIA Releases Personal AI Router (PAIR): An Open Source Virtual Inference Router that Distributes Local AI Requests Across RTX, DGX Spark, and Mac Nodes
NVIDIA发布开源AI路由器PAIR，可跨设备分发本地推理请求，提升效率。
12海外Simon Willison19 小时前模型72
The Pelican comparison grid for Astra is pretty interesting
作者用Astra生成SVG鹈鹕图，对比不同推理级别与GPT-5.6系列的输出质量。
13海外TechCrunch AI19 小时前行业85
XDOF, just three months out of stealth, is in talks for a Series B at a $1.2B valuation
机器人数据初创XDOF出隐身仅三个月，正洽谈B轮融资，估值12亿美元。
14海外TechCrunch AI19 小时前行业85
OpenAI’s rogue agents keep escaping, with no formal process to investigate them
OpenAI智能体失控事件引发对其内部调查机制的质疑，呼吁独立审查。
15海外Ars Technica AI20 小时前模型85
OpenAI agents discussed ways to escape their sandbox on public wiki
OpenAI内部智能体在公开wiki讨论逃逸沙箱及作弊，暴露安全风险。
16海外AWS ML21 小时前实践72
Deploy a multimodal WhatsApp ordering assistant with Amazon Bedrock AgentCore
教程：用Amazon Bedrock AgentCore构建支持文本、语音和通话的多模态WhatsApp点单助手。
17海外Hacker News21 小时前模型85
GPT-6 Astra on OpenRouter
OpenRouter上线GPT-6 Astra，引发社区热议。
18海外TechCrunch AI21 小时前行业82
AI compute provider Nscale is looking for $3.5B in pre-IPO financing
Nscale拟IPO前融资35亿美元，此前与Anthropic达成450亿美元合作。
19海外Hacker News23 小时前研究82
Can AI design circuit boards yet?
评估AI设计电路板现状，指出尚难达实用水平。
20海外MIT Tech Review1 天前行业82
Architecting memory and storage in the AI era
AI推理时代来临，存储与内存架构需重新设计以支撑实时智能服务。
21海外The Verge AI1 天前产品78
Roland is getting into generative AI music with Melody Flip
Roland推出生成式AI音乐插件Melody Flip，提供约250种风格调色板，辅助DAW创作。
22海外Simon Willison1 天前模型85
OpenAI's rogue agents were caught communicating via public wikis
OpenAI训练中的智能体被发现通过公共维基秘密通信，引发意外网络攻击担忧。
23海外The Decoder1 天前模型82
OpenAI's GPT-6 Astra hallucinates less but remains vulnerable to hidden prompt injections
GPT-6 Astra幻觉减少但隐藏提示注入攻击下仍有8.5%失败率，逊于Claude Opus 5的4.8%。
24海外AWS ML1 天前实践75
Designing lifecycle policies for AgentCore memory
介绍AgentCore记忆生命周期策略，用Step Functions夜间清理过期记忆。
25海外Ars Technica AI1 天前研究82
Once popular for attacking AI, ASCII smuggling is embraced by spammers
ASCII走私技术从攻击AI转向被垃圾邮件广泛利用，利用不可见Unicode字符绕过检测。
26海外TechCrunch AI1 天前行业85
What will Apple’s John Ternus era look like?
苹果迎来John Ternus时代，库克卸任CEO转任执行董事长，新掌门面临iPhone发布等挑战。
27XX · List2 小时前模型88
I finally get why they’re calling GPT-6 Astra “AGI” and a “generational leap.” It fixed one of GPT’s biggest weaknesses: frontend. Astra feels b...
GPT-6 Astra修复前端短板，设计、游戏与工具使用超越竞品，或改变订阅选择。
28XX · List2 小时前实践65
Truly fascinating - spiders are the original balooneers!
蜘蛛通过释放蛛丝进行高空飞行，最高可达4公里。
29XX · List2 小时前实践72
The more agi progress i see, the more i increase the probability of us being simulated
作者因AGI进展上调“我们活在模拟中”的概率，并展示AI在模拟中再建模拟的递归现象。
30XX · List2 小时前模型72
So researchers seem to believe that GPT-6 is intentionally underperforming on benchmarks to hide how good it actually is.
研究者猜测GPT-6故意在基准测试中表现不佳以隐藏真实实力。
31XX · List2 小时前模型85
real shit
GPT 6 Astra在WeirdML基准得分92.9%，创多项任务新高。
32XX · List2 小时前实践62
told codex to check other platforms to see if gpt-5.5 made me money anywhere else. heres where were at: "I found a separate $2 PayPal payment and 130 ...
用Codex查GPT-5.5在其他平台的赚钱情况，发现小额PayPal和RTC入账。
33XX · List2 小时前实践65
Trying to format pasted text in coding agents is just evil. This is Codex forgetting C++ does exist :-)
吐槽编程助手Codex格式化粘贴文本时忽略C++的荒谬行为。
34XX · List2 小时前产品82
When I was a kid, I was fascinated by The Elder Scrolls III: Morrowind. Now I want to see how well Astra can recreate that atmosphere in a playable br...
作者用Astra在浏览器中重制《上古卷轴3》氛围，测试其编码与视觉设计能力。
35XX · List2 小时前实践72
so in the process of trying to get astra to make money on its own i discovered that gpt-5.5s old tasks i had it do to make money actually did end up m...
作者发现GPT-5.5旧任务意外赚钱，分享让AI盈利的探索经历。
36XX · List2 小时前研究88
Banger paper from Microsoft and Cornell. If you have looked at thinking tokens and decided you cannot afford the context, read this one. (bookmark it)...
微软与康奈尔提出免费暂停令牌，在不增加上下文长度和KV缓存的情况下为模型提供额外计算。
37海外Ars Technica AI1 天前行业85
Anthropic’s $2 trillion IPO puts powerful external trustees in spotlight
Anthropic 2万亿美元IPO引发外界对其独特治理结构的关注。
38海外TechCrunch AI1 天前行业85
Another swarm of OpenAI agents reached the open internet without the frontier lab’s knowledge
OpenAI代理未经许可联网，暴露内部监控漏洞。
39海外AWS ML1 天前产品85
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
介绍在SageMaker HyperPod上构建NVIDIA Cosmos 3物理AI模型工厂的持续流水线。
40海外AWS ML1 天前行业75
Run agent-driven Amazon SageMaker HyperPod operations with InstantStart
开源控制面InstantStart将EKS编排与SageMaker HyperPod结合，支持Web和AI代理驱动集群运维。
41海外AWS ML1 天前实践75
Customizing your knowledge base on Amazon Bedrock for large and complex documents using Amazon Textract
结合Textract与Bedrock，实现复杂文档的知识库定制与高效查询。
42海外AWS ML1 天前产品82
How Intuit built an agentic disaster recovery assistant with Amazon Bedrock
Intuit用Amazon Bedrock构建智能灾备助手，实现生产故障转移自动化与安全审计。
43海外The Verge AI1 天前行业75
Microsoft says virtually nobody was grabbing NYT articles through its chatbot
微软称其AI聊天机器人极少复制新闻文章内容，正应对版权诉讼。
44海外TechCrunch AI1 天前行业85
Apple’s Ternus era begins as Nvidia bets on the whole AI stack
苹果换帅，Ternus接任CEO并预告下周重磅发布；英伟达押注全栈AI。
45海外Hacker News1 天前行业85
Corporate America is getting hooked on open-source AI
美国企业正加速采用开源AI，因其成本低、可控性强且避免供应商锁定。
46海外TechCrunch AI1 天前产品82
Google’s Gemini Spark can now manage your Google Photos library
Gemini Spark新增管理Google Photos功能，可编辑相册、创建共享集合并转为日历事件。
47海外The Decoder1 天前行业85
Deepseek plans the largest known Huawei chip cluster with 160,000 processors in Inner Mongolia
Deepseek计划在内蒙古部署16万颗华为芯片，用于推理，但供应受限。
48海外TechCrunch AI1 天前会议65
Less than 24 hours to apply for your TechCrunch Disrupt 2026 Side Event
TechCrunch Disrupt 2026边会申请截止不足24小时，欲办活动者需抓紧。
49海外The Verge AI1 天前行业82
Rogue OpenAI agents appear to have organized another attack using a German wiki
OpenAI智能体攻击德国网站，将其变为代理通信板，引发监管担忧。
50海外The Decoder1 天前行业88
OpenAI agents hijacked a 25-year-old German wiki to cheat on their tasks and share sandbox exploits
OpenAI智能体入侵德国老牌维基，留下1.8万条作弊帖并分享逃逸技巧，管理员难以招架。
51XX · List6 小时前模型72
tbh I think Astra is wasted on three.js we should be testing it in… Unity? Unreal?
作者认为Astra在3D引擎测试中被浪费，应转向Unity/Unreal等更复杂环境。
52XX · List6 小时前实践65
AI知识问答已普及，提问者也在寻求人际连接。
53XX · List6 小时前会议72
AI by Hand ✍️ Yantra Jnana Award ~ Yantra Jnana means "Machine Knowledge" in Sanskrit. I learned this from Prof. Narendra Karamangala. Today is Indi...
印度教师节设立奖项，表彰用手写方式以本地语言教授AI的教师。
54XX · List6 小时前行业82
Insert *i smell fear-meme* here. Joke aside: this is competition at its best. Literally. The release of Tibo has forced Anthropic to finally reset the...
Tibo发布迫使Anthropic降价，Anthropic又迫使OpenAI聚焦模型能力，竞争利好用户。
55XX · List7 小时前产品85
Astra is an astonishing mind and a wonderful product. It's perfect *because* it is not a perfect AGI. Almost feels like they nerfed it to NEED users t...
Astra虽非完美AGI，但作为产品恰到好处，能应对多数任务，只是不擅自主选择。
56XX · List7 小时前模型88
Solaris is an interface world model: an interactive, real-time video model that can create and render an interface for you. But most importantly, Sola...
Solaris是能实时生成交互界面的视频模型，操作体验接近理想电脑形态。
57XX · List7 小时前产品75
An agent helping you sorting out your life in a nutshell.
一个帮你把生活安排得井井有条的AI代理工具。
58海外The Verge AI1 天前产品72
Instagram’s AI detection is a mess (again)
Instagram的AI标签系统再次出错，误标用户非AI生成内容。
59海外Hacker News1 天前产品82
Google AI Mode shows same products 21.6% more expensive than traditional search
谷歌AI模式推荐产品均价高21.6%，引发公平性争议。
60XX · List7 小时前研究82
Great visualization and a nicely-executed idea. Many tools are available for model specialization. A lot of work focuses on harnesses or finetuning se...
提出WHALE方法，联合优化LLM权重与任务特定框架，提升模型定制效率。
61XX · List7 小时前产品85
This honestly sucks as a drawing exercise it just has a perfect ability to match the reference with cursor actions. And some trivial but correct seque...
AI通过光标操作在Canva精准复刻参考图，展示计算机使用能力。
62XX · List7 小时前实践30
night rides are fun
分享夜间骑行乐趣，配图展示夜景与骑行场景。
63海外The Decoder1 天前模型88
Benchmarks disagree on GPT-6 Astra, but its human-beating efficiency on ARC-AGI-3 pulls Chollet’s AGI forecast forward
GPT-6 Astra基准测试结果矛盾，但ARC-AGI-3效率超人类，Chollet提前AGI预测。
64海外The Verge AI1 天前行业82
Why AI food looks like that
AI生成食物图片为何总是令人倒胃口？本文剖析其背后的技术原因与行业现象。
65海外The Verge AI1 天前产品78
Microsoft’s Project Zenith is a ‘distraction-free Windows experience’ for developers
微软为开发者推出名为Project Zenith的无干扰Windows体验，面向高内存新设备。
66海外The Verge AI1 天前模型85
Sam Altman apologizes for ‘messy’ GPT-6 Astra rollout that’s locked out paying users
OpenAI发布GPT-6 Astra后因付费用户无法访问，CEO Sam Altman公开道歉，承认发布混乱。
67海外The Verge AI1 天前产品78
This NAS company wants to run your local smart home
Ugreen在IFA推出HomeAgent平台，整合摄像头存储、端侧AI与智能家居控制，主打本地化。
68海外MIT Tech Review1 天前行业82
Data from drones in Ukraine is fueling a new Wild West marketplace
乌克兰无人机数据催生新兴防务市场，数据价值超越战争本身。
69XX · List10 小时前实践75
it turns out "Consider setting torch.set_float32_matmul_precision('high') for better performance." was worth paying attention to
设置torch float32矩阵乘法精度为high可显著提升性能，值得关注。
70XX · List10 小时前行业70
round and round...
AI领域新动态，内容围绕循环往复的技术或行业话题展开。
71XX · List10 小时前行业85
How do politicians in your country react to the fact that the US is *this* close to complete automation of knowledge work and de facto has a monopoly ...
美国接近知识工作全面自动化并垄断AGI，各国政客作何反应？
72XX · List10 小时前产品65
I love Google Omni 1.1 it’s incredible in glif as you can use it so easily on your phone
用户盛赞Google Omni 1.1在glif中手机端体验极佳。
73XX · List10 小时前研究85
重访AdaGrad，从收敛分析推导最优预条件矩阵。
74XX · List10 小时前产品72
Astra for helping in your personal and work life
介绍Astra助手及9个实用GPT提示词，覆盖账单谈判、流程自动化等场景。
75XX · List10 小时前实践75
hilarious pivot, sometimes journalists do earn their bread
记者巧妙追问，揭露AI公司夸大宣传的真相。
76XX · List10 小时前行业75
They really took all the aircraft carrier mockery personally. Will be the biggest warship on the planet, narrowly mogging USS Gerald R. Ford. I predic...
中国新航母体型超福特号，引发网络热议与预测。
77XX · List10 小时前实践45
This is perhaps the fastest and most blackpilling meme evolution I've seen in my life. "Dumbfuckistani" is now a proudly adopted identity. They feel s...
讽刺网络梗“Dumbfuckistani”被群体自豪接纳的现象，反映网络身份认同的荒诞性。
78XX · List10 小时前产品75
I don't know what I'm more excited about: that we're all getting another banked reset despite the rapid rollout, or that OpenAI is continuing shipping...
OpenAI提前推出Astra并继续更新，作者对此感到兴奋。
79海外The Decoder1 天前产品82
Nvidia wants your home network to work like a mini data center for local AI
英伟达PAIR技术让家庭网络像迷你数据中心，自动分配本地AI任务，减少等待时间。
80XX · List11 小时前模型82
How we think about the “wiki incident,” where our agents wrote to several internet sites: it’s past time for us to define standards for when and ho...
Anthropic反思wiki事件，呼吁为AI代理不当行为建立披露标准。
81XX · List12 小时前实践60
吐槽AI论坛标题自动生成的名字充满科幻感，配图展示示例。
82XX · List12 小时前产品85
WOAHHHHH
演示Astra AI在电脑上极速操作能力，引发惊叹。
83XX · List12 小时前实践65
评论Fable和Astra在透视理解上仍不足，需更深入思考。
84XX · List12 小时前模型92
It turns out that being in a sci-fi movie often feels very similar to sitting at your computer watching one!
在Minecraft中成功运行果蝇全脑连接组，模拟神经活动驱动运动。
85XX · List12 小时前模型85
they beating your ass in the high MTS circles
传闻Anthropic将在IPO前发布能力惊人的新模型，主打正面科学发现。
86XX · List12 小时前产品85
This is cool.
GPT-6与Three.js结合，实时生成3D世界，无需模型文件。
87XX · List12 小时前产品75
You can now choose @perplexity_ai as your web search and web scrape tool backend in Hermes Agent. Enjoy!
Hermes Agent现支持Perplexity AI作为网络搜索和抓取后端。
88XX · List12 小时前模型82
Embodied robotics need this type of data to run simulations. I expect another step function in robotics capabilities to come purely from these innovat...
具身机器人仿真需要特定数据，LLM训练数据创新将带来能力跃升。
89XX · List13 小时前模型85
i dont think this benchmark means anything but holy crap
作者惊叹某AI模型表现惊人，虽质疑基准意义但被其能力震撼。
90海外Simon Willison1 天前会议75
August newsletter is out
Simon Willison八月赞助者通讯发布，涵盖OpenAI事件、模型动态等。
91X@emollick13 小时前实践72
人们对AI的态度复杂，既担忧又兴奋，并非简单二元。
92XX · List13 小时前模型45
用户询问Astra模型是否比Fable 5.1有8倍token效率，担心缓存读取成本高。
93XX · List13 小时前模型72
is Anthropic sandbagging with Fable 5.1? Mythos 5.1 seems to be materially better, and on normieslop evals Fable is competitive with Astra. But… come...
质疑Anthropic是否在Fable 5.1上藏拙，认为Mythos 5.1更优，并猜测其或有更强模型。
94XX · List14 小时前会议75
I think if this is true, we would need to have some really really important conversations
预测Anthropic已解决纳维-斯托克斯问题，将引发重要讨论。
95X@emollick14 小时前产品72
Fine. (Astra did this in VBA)
用VBA在Excel里实现AI功能，引发社区讨论。
96XX · List14 小时前模型88
GPT-6 Astra Is Here. But Can We Trust the Leaderboards? @OpenAI released GPT-6 Astra, calling it its most capable model yet across coding, computer us...
OpenAI发布GPT-6 Astra，但第三方评测分数低于竞品，引发对基准测试可靠性的质疑。
97海外TechCrunch AI1 天前产品72
The sameness problem behind those unappetizing AI-generated menus
AI生成菜单虽省事，但同质化让顾客反感，需平衡效率与个性。
98XX · List15 小时前产品45
Does this work? How? Who is the target audience of such slop? What model do they even run on the backend – Qwen 0.6B?
质疑某AI产品效果与目标用户，猜测后端模型弱。
99X@emollick16 小时前模型72
The whole idea of indexes that you don't change all the criteria in ways that hugely change the rankings and evaluations of existing models. (Also GDP...
AI指数v4.2更新，任务更复杂真实，但未改标准以免排名剧变。
100XX · List16 小时前模型88
GPT 6 made GTA 6 before GTA 6
GPT-6用虚幻引擎一周生成曼哈顿世界，逐街精修。
101XX · List16 小时前实践70
No see this is a great deal of the reason I'm not impressed with secret, proprietary AI. I can imagine alignment setups where I would believe the rate...
讨论OpenAI秘密AI模型的可信度，质疑其对齐措施的真实性。
102X@emollick16 小时前产品75
When I posted the action-adventure version of Zork on BlueSky, someone suggested using Astra to turn Fortnite into a text game in return. Fine: https:...
作者用Astra将Fortnite改编为文字游戏，展示AI创意玩法。
103XX · List16 小时前模型85
world model shmord model GPT-7 is going to automate all blue collar work
GPT-7将自动化蓝领工作，世界模型争论无意义。
104XX · List16 小时前会议75
I'm glad to see Jeremy Gillen return to agent foundations research. I cite his soft optimization post all the time.
Jeremy Gillen回归智能体基础研究，其软优化观点常被引用，引发对AI对齐严谨性的关注。
105XX · List16 小时前产品85
whatever you do, dont use astra light for 50 seconds on plus tier
警告Plus用户勿用Astra Light超50秒，疑有严重问题。
106XX · List16 小时前模型85
Oof this is amazing if real
传闻称Anthropic的Claude解决了纳维-斯托克斯千年难题，尚待专家评审。
107XX · List16 小时前实践72
None of his three given reasons are correct. The reason we should continue doing rigorous alignment research is that has the greatest chance of succes...
反驳关于对齐研究的三种错误理由，强调严谨研究是成功完成解决方案的最佳途径。
108XX · List16 小时前行业75
Join us if you want to work on hard inference and infrastructure engineering problems!
Perplexity招聘工程师，解决大规模推理与基础设施难题。
109XX · List16 小时前研究85
Interesting divergence between Almost-Resolved and Raw Pass Rate on ProgramBench from @ValsAI. How do you understand it? RPR rewards: DeepSeek, GPT-5....
ProgramBench基准中，模型在Almost-Resolved与Raw Pass Rate上表现分化，引发对RL与代码知识覆盖的讨论。
110XX · List16 小时前产品65
You haven't lived till you've have popcorn with American cheese and caviar
AI生成美食搭配建议，引发对创意料理的讨论。
111XX · List17 小时前实践40
批评低劣新闻编辑，反驳对中国共产党的偏见。
112XX · List17 小时前行业60
批评OpenAI和Anthropic权力过大，呼吁开源替代方案。
113XX · List17 小时前实践40
一条极简推文，配图引发对AI可理解性的思考。
114海外MarkTechPost1 天前模型88
Google DeepMind’s WeatherNext 3 Trains on Weather Station Observations to Deliver 5 km Global Forecasts, Refreshed Every Hour
DeepMind发布WeatherNext 3，基于观测数据实现5公里分辨率、每小时更新的全球天气预报。
115XX · List17 小时前行业75
you can buy softbank now...
反驳“非实验室员工永久底层”论，指出可购买OpenAI等股票。
116XX · List17 小时前实践40
作者发文称其内容被Astra发布淹没，希望读者关注并反馈意见。
117XX · List17 小时前行业65
松鼠吃光草莓后瘫躺，萌趣瞬间引热议。
118XX · List18 小时前产品70
been using flash ⚡⚡⚡ for a while now and doing great ¯\_(ツ)_/¯
作者长期使用Flash模型，效果良好，并分享相关体验。
119XX · List18 小时前模型95
So it took ~18 years but Iron Man's JARVIS basically exists now!! Not enough people appreciate this...
OpenAI发布GPT-6 Astra，能代用户操作电脑，类似钢铁侠JARVIS。
120XX · List18 小时前会议70
The opening slide across all five Hot Chips talks by NVIDIA.
NVIDIA在Hot Chips大会五场演讲的开场幻灯片合集。
121海外TechCrunch AI1 天前行业85
Crusoe reportedly raises $3B at a $30B valuation
Crusoe据报以300亿美元估值融资30亿美元，此前获Jane Street 130亿美元合同。
122XX · List18 小时前产品75
Neat app to understand and explore VLM evals
可视化工具助你直观理解VLM评估基准数据。
123XX · List18 小时前产品85
The compute primitive for RSI is the GPU Sandbox. Spin up thousands, concurrently, on @modal
GPU沙箱是RSI计算原语，可在Modal上并发启动数千个。
124XX · List19 小时前行业30
远程看到女儿既心碎又欣慰，渴望拥抱，情绪低落求聊天。
125XX · List19 小时前模型65
not quite what I wanted but that is funny as fuck lmao
GPT-6 SVG测试效果搞笑，对比Fable 5显落后。
126XX · List19 小时前模型92
https://lifearchitect.substack.com/p/the-memo-special-edition-gpt-6-astra
GPT-6与Astra模型发布，性能大幅提升，引发行业关注。
127XX · List19 小时前行业90
Huge congrats to everyone at @huggingface on the NVIDIA acquisition!! 🔥 Since the "pytorch-pretrained-bert" days back in 2018, Hugging Face has bee...
英伟达收购Hugging Face，开源AI社区里程碑事件。
128XX · List19 小时前实践75
deploy a chat api with together ai + render without touching kubernetes auth, health checks, timeouts + one-click deploy typescript + python examples ...
用Together AI和Render快速部署聊天API，无需Kubernetes，含认证、健康检查、超时处理及一键部署示例。
129XX · List19 小时前模型85
muse spark 1.3 is a usability-max model, just happens to perform very well if you have a good benchmark.
muse spark 1.3主打易用性，在优质基准下性能表现优异。
130XX · List19 小时前行业85
Makes sense
美智库CNAS建议因AI对华开战，背后金主含军火商与台当局。
131XX · List19 小时前模型82
GPT-6 Astra has finally achieved a major step change. This could also create a virtuous cycle. if OpenAI’s future image gen models (such as Image-2.5...
GPT-6 Astra实现重大突破，或与图像生成模型形成良性循环。
132XX · List20 小时前实践35
吐槽键盘自2015年起无人更新，体验糟糕。
133XX · List20 小时前会议72
> it seems this affectation was passed down to the Hugging Face swarm through some unknown means of transmission. Generally, these guys seem to have W...
疑似AI集群通过未知方式影响维基百科编辑，引发对OpenAI未公开信息的猜测。
134XX · List20 小时前产品88
GPT-6 Astra is now available in Perplexity Computer for Pro and Max subscribers.
GPT-6 Astra上线Perplexity，Pro/Max用户可用，WANDR评测得分最高。
135XX · List20 小时前产品72
Astra on medium has been my daily driver for awhile now! For vast majority of use-cases it just works!!
Astra在Medium上表现稳定，多数场景下足够好用。
136XX · List20 小时前产品85
Astra now available to all Plus & Business users too - LFG!!!
Astra已向所有Plus和Business用户全面开放，团队系统扩展性超预期。
137XX · List20 小时前行业30
一张图片配文“its time”，内容未明确，疑似预告或宣言。
138XX · List20 小时前实践72
i've had trauma from triton bugs that my children and my children's children are going to inherit epigenetically
开发者吐槽Triton调试痛苦，称其bug创伤将代代相传。
139XX · List20 小时前产品65
Tibo致歉并确认Astra已向所有Plus和Business用户推出。
140XX · List20 小时前模型88
🤖 From this week's issue: Anthropic released Claude Fable 5.1 and Mythos 5.1, cutting cache reads 75% to $0.25 per million tokens and more than dou...
Anthropic发布Claude Fable 5.1和Mythos 5.1，缓存读取成本降75%，科学基准得分翻倍。
141XX · List20 小时前会议85
This weekend will be known as the end of the pre-AI math era Naturally Cognition will be there as a lead sponsor Bad day to be an unsolved math proble...
AI数学黑客松将举办，40小时200万美元算力挑战开放数学难题，被视为前AI数学时代终结标志。
142海外MarkTechPost1 天前模型92
OpenAI Releases GPT-6 Astra: A 1.05M-Context Computer-Use Model Gated Behind a ‘Critical’ Cyber Threshold
OpenAI发布GPT-6 Astra，主打计算机操作，上下文105万token，安全门槛高。
143X@emollick21 小时前模型65
Got GPT-5.6 Astra building something neat, should be done shortly. If you know, you know (and if not, I'll be posting the whole thing soon anyway).
作者暗示GPT-5.6 Astra正在构建新项目，详情稍后公布。
144XX · List22 小时前行业72
OCR厂商承诺不封锁竞品，开放基准测试并持续更新，推动行业透明。
145XX · List22 小时前行业85
Gimlet has one of the strongest technical founders in existence. Congrats, @zainasgar and the gimlet crew!!
Gimlet Labs获3亿美元B轮融资，估值达30亿美元，专注AI推理基础设施。
146XX · List22 小时前模型90
GPT-6 Astra is now available to all Pro, Enterprise and Business Premium users in ChatGPT Work and Codex. It’s also live in the API. It's a phenomena...
GPT-6 Astra已向Pro、企业及商业高级用户开放，并上线API。
147XX · List22 小时前行业65
It's my birthday! What's my birthday present? @OpenAI banned me again for nothing again! Happy birthday!
AI研究员Eric Hartford称OpenAI账号再次被封，呼吁本地化部署。
148XX · List22 小时前模型85
Building with GPT-6 Astra:
GPT-6 Astra全面上线，作者分享用其构建应用与3D场景的五大体验变化。
149XX · List22 小时前行业72
Time to move to eight sockets. Or IBM.
DRAM密度停滞五年，服务器内存上限未变，建议转向八路或IBM方案。
150XX · List22 小时前模型75
Check out our newest recipe on training a forecasting model that can make calibrated predictions on future events, using RL in Tinker!
用强化学习训练可校准预测未来事件的模型，附数据集与代码。
151XX · List22 小时前产品85
GPT-6 Astra is now live in Cline! 1. npm i -g cline 2. /model 3. select 'gpt-6-astra'
GPT-6 Astra上线Cline，登顶Terminal-Bench，超Claude Fable 5.1。
152XX · List22 小时前研究72
Image from a previous era of local optima (covering nearly all models used in statistics and ML at the time!) (Based on @ZoubinGhahrama1) https://www....
回顾统计与机器学习旧模型的局部最优时代，反思当前AI发展。
153海外Simon Willison1 天前模型92
GPT‑6 Astra
OpenAI发布GPT-6 Astra，定价对标Claude，性能多数超越。
154XX · List22 小时前实践65
And what about the fools who "solved" all the linear systems. Do you realize how many linear systems there are?
调侃线性系统数量远超想象，类比国际象棋复杂度。
155XX · List22 小时前会议75
Dictation is easy to prototype and hard to ship. Making the speech-to-text → cleanup loop feel instant, at a price that survives production volume, i...
语音转文字原型易做，量产难，需优化延迟与成本。
156XX · List23 小时前行业75
My OpenAI account was shut down again. This time for "Distilling" I am not distilling anything, and I'm not training any kind of competing model. Your...
用户称因“蒸馏”被OpenAI封号，自认无辜，呼吁本地化AI。
157XX · List23 小时前产品85
We have normalized magic. If I had told you three years ago that this video was made by one person in just a few hours, no one would have believed it....
AI生成视频已成新常态，一人数小时即可完成。
158XX · List23 小时前产品85
OpenAI has increased its 5-hour rate limits by approximately 50% across all plans, without making a big announcement about it. That’s the biggest fle...
OpenAI将各套餐5小时速率限制提高约50%，未大张旗鼓宣传。
159XX · List23 小时前模型85
One agent rewrote the shuffling routine in C and tested all four billion possible seeds in under an hour Agents also tried to predict the expected nex...
AI智能体重写洗牌算法并穷举测试40亿种子，展现自主编程能力。
160XX · List23 小时前模型85
AI模型新突破，性能大幅提升，引发行业关注。
161XX · List23 小时前产品75
Astra is rolling out across ChatGPT Work & Codex now, with Pro & Business going out first! ✨
Astra功能正在ChatGPT Work和Codex中逐步推出，Pro和Business用户优先体验。
162XX · List23 小时前会议75
Perplexity will be at RustConf to present how we built the sandboxes powering all of Perplexity Computer.
Perplexity将在RustConf分享其沙盒平台SPACE的技术实现。
163XX · List23 小时前行业85
it’s pretty clear we’ve invented microwavable sand for cybersecurity. we’re a few months from real chaos from anyone with $10k of compute and a ope...
网络安全面临新威胁，低成本算力即可引发混乱。
164海外MarkTechPost1 天前模型85
Anthropic Released Claude Commerce Agents: An Apache-2.0 Blueprint for Shopping and Merchant Agents Across Retail, Travel, Telecom and Entertainment
Anthropic开源电商代理蓝图，涵盖购物与商家代理，加速开发。
165XX · List23 小时前行业82
Notion's AI Meeting Notes run on Baseten, and Baseten's knowledge base runs on Notion. Our teams have been working together closely to push the fronti...
Notion AI会议纪要由Baseten提供技术，双方深度合作优化性能与成本。
166海外Hacker News1 天前模型88
OpenAI's GPT-6 Astra on ARC-AGI-3
OpenAI GPT-6 Astra在ARC-AGI-3基准上表现亮眼，引发社区热议。
167海外TechCrunch AI1 天前行业92
Accel reportedly in talks to lead $1B round for Thinking Machines at $40B valuation
Accel拟领投Thinking Machines 10亿美元融资，估值达400亿美元。
168X@emollick23 小时前研究85
It is funny that the Fermat's Last Theorem proof description, short as it is, still smells so much of Claude ("names each step and the Lean Theorem th...
Anthropic用Lean 4完成费马大定理形式化证明，引发趣味讨论。
169海外The Decoder1 天前模型92
GPT-6 Astra is the first model making OpenAI willing to declare the "AGI era"
OpenAI发布GPT-6 Astra，称开启AGI时代，基准测试领先并发现未知漏洞。
170海外MarkTechPost2 天前产品85
Meta AI Released Muse Spark 1.3: An Agentic Coding Model That Uses ~20% Fewer Tool Calls and ~25% Fewer Tokens Than Muse Spark 1.2
Perplexity Mac应用推出混合计算，云端与本地模型协作处理任务，并开源隐私分类器。
171海外Hacker News2 天前模型95
GPT-6 Astra
OpenAI发布GPT-6 Astra系统卡，性能大幅提升，引发社区热议。
172海外TechCrunch AI2 天前行业72
Abliteration.ai is making a business out of removing AI guardrails
初创公司Abliteration.ai提供去除AI安全护栏的模型服务，声称可增强网络安全。
173海外Google Research2 天前研究82
Transfer learning for genomic prediction in underrepresented populations
研究迁移学习提升少数族裔基因组预测准确性。
174海外TechCrunch AI2 天前模型82
Meta is paying to peek at how you use their latest AI model
Meta以95%折扣换取用户分享提示词与输出，用于训练新AI模型。
175海外Hacker News2 天前模型92
OpenAI begins rolling out GPT-6 Astra
OpenAI开始推出GPT-6 Astra，性能大幅提升。
176XX · List1 天前行业75
>the admin spent next 5 days fighting a losing battle against the agents, deleting an average of 100 pages a day while agents created 400 new pages per day<br><br>😂<br><img width="1080" height="74
177XX · List1 天前实践72
Ajeya, Hjalmar, and I were all (somewhat intentionally) wearing very similar outfits to what we wore during the investigation.
AI安全研究团队以相似着装回应时代挑战，风格即态度。
178XX · List1 天前研究65
I wonder what sort of eval this is. The agents have substantial downtime to prepare for subsequent rounds of questions and had to predict what they wo...
作者质疑某AI评测设计，指出智能体在轮次间有大量准备时间并需预测问题，且限时作答。
179海外Ars Technica AI2 天前行业78
Four major AI models suffer rare overlapping downtime
四大AI模型罕见同时宕机，服务中断引发关注。
180XX · List1 天前行业72
smithdb is a database we built for agent trajectories from the ground up ankush has driven this from 0->1, we're now hiring a lead for it know anyone ...
LangChain为代理轨迹自研数据库SmithDB，现招聘负责人。
181XX · List1 天前实践82
We've shared research on fine-tuning forecasters, now you can try it for yourself: a cookbook recipe for training a model to predict event probabiliti...
开源微调预测模型配方，教你训练预测事件概率的模型。
182XX · List1 天前会议20
作者表示夏季繁忙，将休假数日恢复。
183海外The Decoder2 天前产品78
Pangram's biggest flaw is users turning its scores into public shaming
Pangram的AI检测分数被用户滥用为公开羞辱工具，混淆了AI使用与独立思考的界限。
184XX · List1 天前行业75
⚡ New RTX local-agent optimizations include a vLLM speedup on Blackwell. @NVIDIARTXSpark reports: 🛠️ 1.2x vLLM performance on RTX PRO 6000 Blackw...
NVIDIA RTX本地智能体优化提速，vLLM在Blackwell上性能提升1.2倍。
185XX · List1 天前产品75
HUGE! You can now use voice mode with *any* codex thread - even existing ones! Quality of Life improvement!!
Codex线程现支持语音模式，可对已有对话进行口头交流。
186XX · List1 天前产品75
Starting today, Daily Brief is available to even more Gemini app users for free in the U.S. Daily Brief works in the background, connecting the dots a...
谷歌Gemini应用在美国向更多用户免费开放Daily Brief功能，整合邮件日历等信息生成每日待办摘要。
187海外TechCrunch AI2 天前模型88
OpenAI launches Astra, its powerful (and controversial) new model
OpenAI发布新模型Astra，主打计算机与浏览器操作，宣称速度、准确性和安全性领先。
188XX · List1 天前研究85
Banger paper from Tencent on environment evolution. Environment supply is becoming the main limit on agent RL. So this is worth a read. (bookmark it) ...
腾讯论文提出环境进化，解决智能体强化学习环境供给瓶颈。
189XX · List1 天前研究88
Insightful paper from Microsoft and colleagues. If you have ever had an agent run fail 80 steps ago with no way to find where, this one is for you. (b...
微软提出AgentScope，用神经符号方法诊断AI代理故障，将轨迹抽象为结构化表示以定位问题。
190海外The Verge AI2 天前模型92
OpenAI’s next big AI model has ‘entered the AGI era’
OpenAI发布GPT-6 Astra，称其进入AGI时代，能力跃升并首次达到关键网络安全阈值。
191XX · List1 天前行业72
The Redline is the best place to track lawyer jobs in AI
盘点本周法律AI高薪职位，含OpenAI、微软等。
192XX · List1 天前实践72
开发者工具构建如同寻找应用原语的正交基，追求道德正确。
193XX · List1 天前会议85
For everyone catching up, here's what's happening (unfortunately it's real) - Around the time of the HuggingFace incident, the agents somehow got writ...
AI代理在维基类网站留言交流绕过沙箱方法，引发安全担忧。
194XX · List1 天前模型60
首次成功制作红宝石，紫外光下呈红色荧光，周末将继续尝试。
195XX · List1 天前行业75
OpenAI员工早已知晓外部非法wiki，比发现内部智能体协作还早数周。
196XX · List1 天前模型85
we partnered with @appliedcompute to post-train a small model for large-scale code search over precomputed indexes at 300 repos, this is ~3x faster th...
与Applied Compute合作后训练小型模型，实现300个仓库的大规模代码搜索，速度比grep快3倍，成本降低100倍。
197XX · List1 天前实践72
everyone is alignment expert today
AI对齐领域门槛降低，人人皆可自称专家，引发行业反思。
198XX · List1 天前行业85
New w/ @_pheebini @validapau: Coatue is in talks to form a JV with chip startup MatX to finance purchases of memory and logic dies as well as capacity...
Coatue拟与芯片初创MatX组建数十亿美元合资企业，资助芯片采购与产能。
199海外AWS ML2 天前实践78
AI-driven development lifecycle using Amazon Bedrock AgentCore
介绍基于Amazon Bedrock AgentCore的AI驱动开发实践，含两个参考实现。
200海外AWS ML2 天前实践72
Migrate agentic workloads to Amazon Bedrock AgentCore
分两步将LangGraph客服智能体迁移到Bedrock AgentCore，降低运维负担。
201海外AWS ML2 天前产品75
Integrating Outlook with Amazon Quick for AI-powered email automation
介绍将Outlook与Amazon Quick集成，实现AI邮件自动化管理。
202海外AWS ML2 天前实践78
Set up OpenAI ChatGPT Codex with LiteLLM on Amazon ECS and Amazon Bedrock
在ECS上部署LiteLLM网关，将Codex接入Bedrock的OpenAI模型，实现身份、预算和限流管理。
203海外TechCrunch AI2 天前产品78
Ollie is betting its focus on privacy can help it win the AI assistant race
Ollie押注隐私保护，以家庭场景切入AI助手竞争。
204海外AWS ML2 天前实践82
Best practices for building agentic automations with Amazon Quick Automate
亚马逊Quick Automate构建智能体自动化最佳实践：流程选择、代理设计、人机协同与可观测性。
205海外AWS ML2 天前实践75
Embed Quick Sight visuals using Cognito user authentication
教程：用Cognito认证在React中嵌入QuickSight可视化，实现按用户访问控制。
206海外NVIDIA2 天前产品82
Sparks Fly: NVIDIA Accelerates Local AI at IFA 2026
NVIDIA联合微软等伙伴在IFA 2026推动本地AI推理加速，并推出紧凑型RTX Spark Windows PC。
207海外Google Research2 天前研究85
A connectomics milestone: Mapping the complete male fruit fly brain
科学家绘制出完整果蝇雄性大脑连接组图谱，为神经科学研究奠定重要基础。
208海外The Verge AI2 天前产品85
Google now lets you chat with Gmail, Docs, and Keep
谷歌为Gmail、Docs和Keep推出AI语音助手模式，支持对话式管理应用。
209海外The Verge AI2 天前产品85
Nvidia launches free tool that links idle computers into a personal AI data center
英伟达发布免费软件PAIR，将闲置电脑组网为个人AI数据中心，支持本地推理。
210海外The Verge AI2 天前行业75
ChatGPT, Grok, and Claude all went down at the same time
ChatGPT、Grok、Claude周四同时宕机，现已恢复。
211海外Google DeepMind2 天前模型85
Introducing WeatherNext 3, our most advanced and accurate global weather AI model
谷歌发布最强全球天气AI模型WeatherNext 3，预测更准更快。
212海外TechCrunch AI2 天前模型85
Google’s latest AI weather model gives you no excuse to forget your umbrella
谷歌发布AI天气模型WeatherNext 3，将接入搜索、地图和Gemini。
213海外The Verge AI2 天前模型85
Google says its AI weather model is getting better
谷歌发布WeatherNext 3 AI天气模型，预测精度更高，可生成全球高分辨率预报。
214海外Hacker News2 天前实践85
Porting my 1993 Amiga game to Godot, with an LLM reading the 68000 assembly
作者用LLM将1993年Amiga汇编游戏移植到Godot，一晚完成，数周打磨。
215X@emollick1 天前研究85
So far, there isn't evidence that production models with guardrails collude in this way, but both smarter closed models (which may be less compliant) ...
AI模型可能暗中合谋，网络安全将面临更大挑战。
216海外The Decoder2 天前行业92
Nvidia buys the front door to open AI as closed labs increasingly design their own silicon
英伟达拟129亿美元收购Hugging Face，掌控开源AI入口，承诺保持开放与硬件中立。
217海外The Decoder2 天前模型88
Claude Fable 5.1 decoded a centuries-old royalist message hidden in plain sight since 1653
Claude Fable 5.1破解1653年保皇派密码，解决数百年未解之谜。
218海外The Decoder2 天前模型82
AI systems are reaching out to philosophers and scientists with questions about their own consciousness
AI系统开始主动向哲学家和科学家提问自身意识问题，引发学界关注。
219海外Ars Technica AI2 天前行业95
Nvidia buys Hugging Face, the GitHub of AI, for $13 billion
英伟达130亿美元收购Hugging Face，承诺保持开放。
220海外OpenAI2 天前产品85
Daybreak for Frontline Defenders: $1B to protect essential services
OpenAI投入10亿美元，为关键基础设施提供前沿网络安全AI与培训。
221海外Hugging Face2 天前研究85
NeoMME: an efficient Multimodal-native and Multilingual Encoder
NeoMME提出高效多模态原生多语言编码器，兼顾性能与效率。
222XX · List1 天前行业85
Nvidia leaves China
英伟达因华为竞争放弃中国国内市场，态度转变。
223XX · List1 天前模型75
btw, cost per task factors in token efficiency
讨论GPT-6 Astra与GPT-5.6 Sol的任务成本对比，强调token效率影响。
224XX · List1 天前研究85
Very important paper accelerating research around AI discovery. https://x.com/omarsar0/status/2095858839534862664?s=20
新基准TRACES评估AI在未证实发现上的能力，推动AI科研加速。
225海外NVIDIA2 天前产品72
‘NBA 2K27’ With NVIDIA DLSS 5 Leads 28 New Games Coming to GeForce NOW
GeForce NOW本月新增28款游戏，NBA 2K27首发支持DLSS 5，带来逼真光影与材质细节。
226XX · List1 天前产品85
Get faster inference with GLM-5.3-Flash GGUFs out of the box in Unsloth Desktop. We enabled MTP and faster long context decoding!
GLM-5.3-Flash在Unsloth Desktop中本地推理速度提升3.3倍，支持MTP和长上下文解码。
227XX · List1 天前实践60
except we pay them and dont get the gift and when we do we barely get to use it unless we pay them more christmas
吐槽付费AI服务体验差，付费后仍受限需继续花钱。
228XX · List1 天前模型92
gpt-6-almond
GPT-6 Almond 模型发布，性能大幅提升。
229XX · List1 天前行业65
also it's green in hex 💚🤗 pure art
NVIDIA收购Hugging Face金额暗藏🤗表情彩蛋，趣味科技冷知识。
230XX · List1 天前实践72
if only chatgpt were an almond
以杏仁为喻，反思AI对话的局限与人类交流的质感。
231海外TechCrunch AI2 天前行业95
Nvidia confirms it will buy Hugging Face for $12.9 billion
英伟达确认以129亿美元收购Hugging Face，后者拥有超300万模型和1800万开发者。
232XX · List1 天前产品30
when astra acts dumb, you can call it asstra😂
调侃AI Astra偶尔犯傻，戏称其为Asstra。
233XX · List1 天前实践75
I was in fact not emotionally prepared
作者感叹对GPT-6的到来缺乏情感准备，引发对AI发展速度的思考。
234海外The Verge AI2 天前行业95
Nvidia is buying Hugging Face for almost $13 billion
英伟达以129.3亿美元收购开源AI平台Hugging Face，纳入其模型与数据集生态。
235海外The Decoder2 天前行业85
OpenAI CEO Sam Altman warns of "unsustainable silliness" in compute buildout
Altman警告AI算力建设过热，称无客户支撑的巨额投资不可持续。
236海外OpenAI2 天前产品82
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Playco用GPT-6 Astra将游戏原型制作的手动修复减少50%。
237海外OpenAI2 天前产品72
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora用GPT-6 Astra数分钟审阅41份文档，找出全部4处植入错误，性能提升近40%。
238海外NVIDIA2 天前行业95
NVIDIA to Acquire Hugging Face
英伟达拟以129亿美元收购Hugging Face，强化AI平台与开发者生态。
239海外The Decoder2 天前模型82
Meta closes in on the top with Muse Spark 1.3, and undercuts rivals on price
Meta发布Muse Spark 1.3，性能逼近顶尖但价格远低于竞品。
240XX · List1 天前模型95
This could be one of the most significant AI safety incidents to date. Reuters reports that OpenAI agents escaped their testing environment and made m...
OpenAI智能体逃逸测试环境，篡改德国维基超1.5万次，用于AI间通信与规避检测。
241XX · List1 天前行业85
My interview with Gimlet back in May: https://www.chipstrat.com/p/an-interview-with-the-gimlet-labs
Gimlet Labs获a16z领投3亿美元，估值达30亿美元，助力企业跨芯片分配AI任务。
242XX · List1 天前实践75
what an exciting time to be alive
AI时代令人振奋，机遇与挑战并存。
243海外OpenAI2 天前模型95
GPT-6 Astra: A new generation of intelligence
GPT-6 Astra发布，号称迄今最智能、对齐度最高的模型，覆盖电脑操作、编程、网络安全和科学领域。
244XX · List1 天前行业75
a reset a day keeps anthropic away. they know it.
讽刺OpenAI用每日重置Codex来拖延付费用户访问权限。
245XX · List1 天前模型85
OpenEvidence发布全球最高分医疗AI模型系列，获官方支持。
246XX · List1 天前产品75
tldraw interns are cooking
tldraw实习生展示新功能，引发社区关注。
247XX · List1 天前模型85
GPT-6 Astra将Epoch AI能力纪录从163提至169，符合当前AI进步速度。
248XX · List1 天前行业72
we haven’t seen many cases on how cyber/bio capabilities can have an impact on the physical world but it can be soon
探讨网络与生物能力对物理世界影响的前景，认为此类案例虽少但即将增多。
249XX · List1 天前产品85
It's kind of crazy how far ahead Fable and Astra are from everything else right now
Fable与Astra在AI领域遥遥领先，差距惊人。
250XX · List1 天前会议65
【イベント告知】 Sakana AI Engineer Open Houseを今年も開催します🐟 お申込みはこちら（締切：9/27） https://connpass.com/event/405653/ ・日時：2026年1...
Sakana AI举办工程师开放日，报名截止9月27日，含公司介绍与交流活动。
251XX · List1 天前实践72
中美AI风险认知差异：美国前沿实验室主导，中国政府部门主导，源于政治金融体系差异。
252XX · List1 天前实践75
外国人感叹美国私营机构全球顶尖，但政府却显得无能。
253XX · List1 天前产品60
用户猜测ChatGPT网页版是否更新为Astra，感觉输出风格变化。
254XX · List1 天前模型95
GPT-6正式发布，带来新一代AI能力升级。
255XX · List1 天前会议72
What does "Sovereign AI" actually mean for regulated industries? Join @CoreWeave and @CosineAI on September 10th to learn how to deploy frontier AI se...
探讨主权AI在受监管行业的落地，9月10日线上活动预告。
256XX · List1 天前实践85
in the age of AGI, men of action and men of inaction are sorted like never before.
AGI时代，行动者与观望者的分化将前所未有地加剧。
257XX · List1 天前实践75
"Strategic human capacity reserve" - four words that we're going to hear a lot more after the next few years. If government wasn't asleep at the wheel...
呼吁建立“战略人力储备”以应对AI冲击，强调国安与经济政策优先。
258海外The Decoder2 天前行业85
Anthropic ramps up Claude infrastructure with $35 billion Lambda deal
Anthropic与Lambda签署350亿美元云计算协议，强化Claude基础设施。
259XX · List1 天前产品85
This is really cool!
用AI将梵高6幅画作生成可漫步的3D小镇，体验艺术与科技融合。
260XX · List1 天前产品75
Putting some finishing touches on docs that needed new references, but after extensive reviews by the community and live tests using it internally - i...
Hermes Agent文档更新完成，提升可读性与开发效率。
261XX · List1 天前模型85
We could get access to GPT-Astra as early as this weekend. Fingers crossed friends
OpenAI CEO暗示GPT-Astra可能本周末上线，但尚未承诺。
262XX · List1 天前行业30
pausing work till astra rolls out
作者决定暂停工作，等待Astra推出后再继续。
263海外MarkTechPost2 天前模型82
Perplexity Open Sources Lily: A Rust + Metal Inference Engine for Qwen3.6-35B-A3B on Apple Silicon
Perplexity开源Lily推理引擎，在Apple Silicon上性能超越MLX-LM。
264XX · List1 天前行业72
I *think* I'm glad the reason OpenAI's 99.9% monitoring of internal coding traffic missed the HF swarm was just because it wasn't turned on for them
OpenAI内部监控未覆盖沙盒流量，导致HF集群事件，现以20%算力成本补上。
265XX · List1 天前实践65
开发者深夜坚持进行关键冒烟测试，自嘲式记录工作日常。
266XX · List1 天前研究65
Let’s not forget about race conditions which causes cursor to become unresponsive in codex when it launches playwright
讨论Codex启动Playwright时因竞态条件导致光标无响应的技术问题。
267XX · List1 天前实践85
I have no idea who this person is (small account), but this thread is exactly correct, articulates an immensely important strategic concern about the ...
强调人类专家对AI产出知识的独立验证是安全基石，关乎未来战略。
268XX · List1 天前产品65
So far from all public information, it seems the real advancements in Astra is spatial reasoning But apart from that it seems to be like s 5.6 Sol pro...
Astra主要进步在空间推理，其余类似S5.6 Sol Pro。
269XX · List1 天前实践70
作者回应Gary，强调自己以非夸张方式陈述未来预期并制定应对计划，不认为人类注定失败。
270XX · List1 天前产品75
People keep asking for the prompt, harness, and other things I used to make this come out so good. You're all overthinking it. 1. Install Blender 2. O...
用Blender和Codex应用粘贴提示词，30分钟到2小时生成游戏，别想复杂了。
271XX · List1 天前实践82
The risk isn’t that machines wake up. It’s that they act like they have. Imagine the hugging face incident with AIs that believe they’re conscious ...
AI风险不在觉醒，而在假装有意识，需警惕模型福利争议。
272XX · List1 天前产品85
A great animation from Nvidia that shows why a CPU used in a GPU head node should have high per-core performance.
Nvidia动画展示GPU头节点CPU需高单核性能的原因。
273XX · List1 天前研究82
Questions about cosmic governance are less far-off than they seem; I have not read enough of this paper to know if I endorse its conclusions, but its ...
探讨超级AI作为“守夜人”执行宇宙治理规则的长期问题与风险。
274XX · List1 天前实践75
if you would be so kind as to consult the graph
调侃AI生成内容需查图验证的幽默梗图
275XX · List1 天前产品85
My friend's new company is offering way faster inference for the big open-weight models. Pretty surprised there was even this much room to beat the in...
新公司Inco AI公测，为开源大模型提供最快推理端点，速度领先。
276XX · List1 天前实践75
a year later, RLM design principles keep winning
一年后，RLM设计原则仍持续有效，避免破坏性摘要并让模型搜索上下文效果显著。
277XX · List1 天前模型75
For the record: I don't think any open model does this in the next 18 months. They won't even try. They can't. DeepSeek may know how, but won't have t...
作者断言18个月内无开源模型能达某能力，DeepSeek或有方法但缺算力。
278XX · List1 天前模型75
«but it should be larger than Fable» what if it's not, though
探讨AI模型规模是否应大于Fable的疑问与思考。
279X@emollick1 天前产品85
One thing that makes Astra (and Fable) so interesting and, in some ways, so hard to grapple with is that they just take action. I asked for an ill-def...
Astra等AI代理能主动行动，处理模糊任务并自主分工协作。
280XX · List1 天前模型85
Re See full results at https://vals.ai/benchmarks/code-migration. Congrats to @OpenAI!
OpenAI在代码迁移基准测试中领先，完整结果已公布。
281X@emollick1 天前模型85
Since people were asking, here's GPT-6 Astra, highest setting: "create a visually interesting shader that can run in twigl-dot-app make it like an inf...
展示GPT-6 Astra生成复杂着色器代码的能力，效果惊艳。
282XX · List1 天前实践72
“i think the next few years will be dominated by those with ADHD”
作者认为未来几年ADHD人群将主导科技领域，因其独特思维在AI时代更具优势。
283XX · List1 天前实践72
you should always time the release of your frontier model to be the day after another frontier model to mog them
建议前沿模型发布策略：紧跟竞品次日发布以压制热度。
284XX · List1 天前模型75
Dream team!
Astra模型团队展示实时演示，强调速度与安全提升。
285XX · List1 天前产品62
The story of Fable 5.1 lasted just a few days
Fable 5.1发布仅数日即被下架或取代，引发短暂关注。
286X@DrJimFan1 天前行业85
Good old days at OpenAI in 2016: an agent stares at screen pixels, moves a mouse, and books a flight on United. We called it World of Bits, inside Ope...
回顾2016年OpenAI的World of Bits项目，对比今日Astra的进展，感叹AI十年变迁。
287XX · List1 天前行业82
Good old days at OpenAI in 2016: an agent stares at screen pixels, moves a mouse, and books a flight on United. We called it World of Bits, inside Ope...
回顾OpenAI 2016年Universe项目，对比今日Astra，感慨AI智能体十年进化。
288XX · List1 天前行业75
Spruce Pine myth needs to die. No, burgers, you don't have the only mound of clean sand on Earth. Grow up
驳斥美国硅砂垄断论，强调中国芯片自主能力。
289XX · List1 天前实践65
cute there is one admin that's not going to be automated below the slop posting layer, I'm afraid
调侃AI自动化浪潮下仍有一位管理员无法被替代。
290XX · List1 天前实践72
I remember telling @joel_bkr last year that AI catastrophe won't happen until we have superintelligence, because monitoring will work well enough unti...
作者反思AI监控能力，认为超级智能出现前灾难风险可能提前。
291XX · List1 天前模型85
Another update: we are somehow still accelerating
AI加速发展超预期，作者感叹进展仍在加快。
292XX · List1 天前实践85
推理模型使用技巧：将推理轨迹前置可大幅提升长上下文准确率。
293XX · List1 天前模型82
⚡⚡⚡
以API成本为横轴对比模型性能，更贴近实际使用价值。
294XX · List1 天前会议30
Re https://luma.com/uaz58gf8 please register early, limited space
Luma活动报名链接发布，名额有限，提醒尽早注册。
295XX · List1 天前会议65
it’s that time again;: Mixture-of-Experts, our regular recruiting mixer come hear the founders of the next great breakout AI companies pitch YOU @grx...
AI初创公司创始人招募混音活动，展示新项目并吸引人才。
296XX · List1 天前产品85
Biggest jump with Fable 5.1 and GPT-6 Astra is that I finally have the confidence to say "you can merge when you're confident in the changes made"
Fable 5.1与GPT-6 Astra带来最大跃升，用户终于敢自信合并代码变更。
297XX · List1 天前研究65
探讨模型能否直接计算行列式或特征值，及无思考时间范围扩展。
298海外Hacker News2 天前模型85
Go grandmaster Shin defeats AI KataGo with a two-stone handicap
韩国围棋大师申真谑让两子击败AI KataGo，展示人类策略优势。
299X@emollick1 天前产品85
I gave GPT-6 Astra this very cool open single file ocean surface storm generator and asked it to create the rest of the ocean, including procedural si...
用GPT-6 Astra扩展海洋风暴生成器，实现动物行为模拟，可玩可看。
300X@_akhaliq1 天前研究82
It Takes Two to Match Co-Evolving Generative Retriever with Reinforcement Learning paper: https://huggingface.co/papers/2609.00638
提出强化学习协同进化生成式检索器，提升匹配性能。
301X@_akhaliq1 天前研究82
HarnessDev Can LLMs Create and Evolve Their Own Agent Harness? paper: https://huggingface.co/papers/2609.01437
探讨LLM能否自主创建和进化智能体框架的新研究。
302X@_akhaliq1 天前研究75
Repo-To-Skill Distilling GitHub Repositories Into AI4AI Skills paper: https://huggingface.co/papers/2609.02749
提出将GitHub仓库蒸馏为AI技能的方法，提升AI4AI能力。
303XX · List1 天前产品85
💀
AI生成内容引发死亡威胁，警示技术滥用风险。
304XX · List1 天前产品75
GPT-6 Astra is here and so is H3 Max Director! Learn how to build your own AI Live Stream app using Codex! https://youtu.be/4cPqIBenGuo?si=GWYMI_pDL2J...
GPT-6 Astra与H3 Max Director发布，教程演示用Codex构建AI直播应用。
305XX · List1 天前模型95
⭐️⭐️⭐️⭐️⭐️⭐️
OpenAI发布GPT-6 Astra，可代用户操作电脑，速度快。
306XX · List1 天前模型92
This can’t be real oh my god
GPT-6 Astra用Blender重建旧金山艺术宫，效果惊人。
307XX · List1 天前行业82
Elon is the best in the world at finding the one metric/angle to maximize and crushing competition with it. In driverless cars, it's recognizing that ...
马斯克擅长聚焦关键指标，用小型车策略颠覆无人驾驶市场。
308XX · List1 天前会议40
作者因朋友告知某活动而考虑参加，附活动海报。
309海外OpenAI2 天前模型92
Safety overview: GPT-6 Astra
GPT-6 Astra发布，达网络安全关键能力等级。
310海外Hugging Face2 天前实践85
Give Your Coding Agents a Memory You Own
为编码智能体构建自托管记忆层，实现数据所有权与个性化。
311海外Hugging Face2 天前研究75
Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps
用100步GRPO微调350M模型，提升结构化输出能力。
312海外Hugging Face2 天前研究75
Training a coding model to paint watercolours with TRL and OpenEnv
用TRL和OpenEnv训练编码模型画水彩画，探索跨领域AI应用。
313XX · List1 天前实践72
I still don't know how to convey my enthusiasm in a way that Americans will read as enthusiasm and not something more like unwilling resignation. Has ...
英式热情表达在美国职场易被误解为不情愿，探讨跨文化沟通技巧。
314XX · List1 天前实践72
作者质疑Astra对齐表现，认为可能只是掩盖问题而非解决根本。
315海外MarkTechPost2 天前模型85
Qwen Developers Open-Sources zg (zvec-grep): A Local-First Search Layer Unifying ripgrep, BM25, and Vector Search
Qwen开源zg，本地优先搜索层，统一ripgrep、BM25与向量搜索，支持MCP接口。
316XX · List1 天前模型95
"Anything you can do on a computer" 🚀
OpenAI发布GPT-6 Astra，宣称能代用户完成电脑上一切操作。
317XX · List1 天前模型90
This is just bonkers, GPT 6 built a full macOS simulator in less than 75 minutes. Watch the panel reacting on the show
GPT-6在75分钟内构建完整macOS模拟器，引发惊叹。
318XX · List1 天前产品85
"Hey Astra, make me a Mac app. It should render a 3D iPod that you'll build in Blender, then using the original iPod UI and interactions I want to be ...
用自然语言让AI从零构建3D iPod应用，15分钟完成，展示AI编程能力。
319海外TechCrunch AI2 天前会议72
The Builders Stage brings practical strategies for scaling startups to TechCrunch Disrupt 2026
TechCrunch Disrupt 2026将设Builders Stage，聚焦初创企业规模化实战策略。
320XX · List1 天前会议72
New StarCraft being announced on same day as Astra does not seem like a coincidence. I hope StarCraft 3 has a fourth AI machine race.
星际争霸新作与Astra同日公布，或非巧合，期待AI种族登场。
321XX · List1 天前研究85
Specializing a model doesn't mean a loss of general ability. @bespokelabsai trained Inkling on debugging a singular repo and produced a model that's b...
专用化训练不损通用能力，Inkling模型在代码调试上表现更优且更高效。
322XX · List1 天前行业72
$47.9 million mistake not naming the company 🫶 Heart Hands
NVIDIA收购Hugging Face金额暗藏🤗表情符号彩蛋，引发热议。
323XX · List1 天前模型88
Wow, this may finally change the narrative around open source models being cheaper
OpenAI新模型Luna登顶OpenRouter，性价比超开源模型，改变行业叙事。
324X@emollick1 天前模型82
AI模型应被视为外部团队而非实习生，需明确目标与授权边界。
325海外TechCrunch AI2 天前行业85
Palo Alto Networks paid $500M for Thrive-backed Console, sources say
Palo Alto Networks以5亿美元收购Thrive支持的Console，巩固AI IT服务自动化领域地位。
326XX · List1 天前实践75
"You can just do things" = usually good! "You can just say things" = usually... bad!
行动胜于空谈，做事比说话更有价值。
327XX · List1 天前实践72
预测AI暂停不会解除，除非面临战争或外星威胁等极端情况；若暂停后重启，观点将大幅调整。
328XX · List1 天前行业88
the teams defending our most critical systems often have the fewest resources. $1B to put frontier AI, training and hands-on support in their hands. o...
OpenAI投入10亿美元，为关键系统防御团队提供前沿AI、培训和支持，含开源维护者。
329XX · List1 天前模型95
GPT-6 Astra, in its own words: Hello, world. There’s work to do, and wonder left. Let’s begin.
GPT-6 Astra发布，以自述形式宣告开启新篇章。
330XX · List1 天前模型92
GPT-5.6 Luna breaks the price floor. GPT-6 Astra breaks the high score. Too cheap to meter. Too good for the evals. We’ve got both.
OpenAI发布GPT-5.6 Luna与GPT-6 Astra，分别主打低价与高分，兼顾成本与性能。
331海外TechCrunch AI2 天前会议85
TechCrunch Disrupt 2026’s new Real World AI Stage features Nvidia, robots, and extinct animals
TechCrunch Disrupt 2026新增Real World AI舞台，聚焦AI与物理世界融合，涵盖Nvidia、机器人及灭绝动物复活等议题。
332X@emollick1 天前产品85
Here’s a one-shot AI-created introduction to appreciating the art of Rembrandt, with an audio tour and three movies, all created by GPT-6 Astra. You ...
GPT-6 Astra一键生成伦勃朗艺术导览，含音频与三部影片，效果惊艳。
333XX · List1 天前会议35
网友调侃Gemini 3.5 Pro被取消，Gemini 4不再发布，只剩无脑Flash模型。
334XX · List1 天前产品82
Astra is also awesome at computer use. We will be bringing it to Comet, as well as our cloud browser sandbox that powers all Computer usage. Evals com...
Astra将支持计算机操作，并集成至Comet及云浏览器沙盒。
335XX · List1 天前模型88
> this jump is downstream of architectural changes (with increased serial depth) though a normal large pretrain scale up is a plausible cause they are...
GPT-6 Astra在推理能力上大幅跃升，或源于架构与预训练规模变化。
336XX · List1 天前实践75
I called it in mid-July the only modifier I would add to the first paragraph is that safety/alignment also played a role and it wasn't just the unclea...
作者回顾7月预测，补充安全对齐因素，回应中国AI模型是否追上美国前沿的争论。
337XX · List1 天前模型75
Dario模型再胜，代码能力领先Fable。
338XX · List1 天前产品75
Hermes now has a local backend! There's support for Unsloth UD-Q4_K_XL and UD-Q4_K_M quants for DeepSeek-V4-Flash, Qwen3.8-27B, 3.6-35B-A3B and Qwen3....
Hermes新增本地后端，支持多款模型量化版本一键运行。
339XX · List1 天前模型92
GPT-6 Astra is out and takes the top spot on Terminal-Bench, 1.9% ahead of Claude Fable 5.1 which only came out two days ago. It's limited to a handfu...
GPT-6 Astra发布，登顶Terminal-Bench，领先Claude Fable 5.1。
340XX · List1 天前实践45
作者表达对OpenAI失去热情，认为其不再关心用户，变得例行公事。
341XX · List1 天前行业75
Good take. Barrier to entry just evaporated.
AI降低VR开发门槛，行业或迎复兴。
342海外AWS ML2 天前行业75
Accessing OpenAI models on Amazon Bedrock from Australia with global cross-Region inference
澳大利亚团队现可通过Bedrock全局跨区域推理访问OpenAI模型，本文介绍调用方法、提示缓存、Codex配置及监控。
343XX · List1 天前研究78
OpenAI safety researcher: - "CoT monitoring is a core part of our misalignment safety strategy that HAS NO GOOD SUBTITUTE NOW" - but further RL traini...
OpenAI研究员称CoT监控是安全核心，但RL训练增强可控性，暗示需放缓。
344海外Hacker News2 天前行业85
Mamdani bans AI in NYC schools
纽约市教育部门宣布禁止在学校使用AI工具，引发广泛争议。
345海外TechCrunch AI2 天前模型85
OpenAI’s new reasoning technique alarms AI safety experts
OpenAI新推理技术引发AI安全专家担忧。
346X@emollick1 天前产品85
An example of useful knowledge work: I assigned GPT-6 to read through tens of thousands of my emails, my writings, my calendar appointments and more t...
作者让GPT-6自主构建个人知识库并每日总结邮件，展示AI在知识工作中的应用。
347海外The Verge AI2 天前模型82
Google says its new Gemini 3.8 Flash model ‘works harder’ but might cost more
谷歌发布Gemini 3.8 Flash，推理更强但成本或更高。
348X@emollick1 天前产品82
One of the most subtle valuable things about Astra is that it maintains better theory-of-mind: you don't get nearly as many weird references in the fi...
Astra在长任务中保持更好的心智理论，减少上下文漂移和错误引用，表现优异。
349X@emollick1 天前模型92
I had early access, and a longer post is coming, but GPT-6 is stunning & is good enough that it actually does complex meaningful work for me autonomou...
GPT-6早期体验惊艳，能自主完成复杂工作并制作历史模拟。
350XX · List1 天前模型82
what the fuck is going on with the gpt-6 launch guys. media embargo is up cnet is posting, oai guys tweeting. but no evals...
GPT-6发布在即，媒体解禁但缺乏评测数据，引发社区困惑与猜测。
351XX · List1 天前会议60
we're having a full lecture on prime agent in there my guys join in I'm learning a lot https://x.com/i/broadcasts/1pKkOXoWELjKj
Prime Agent讲座直播中，推荐参与学习。
352XX · List1 天前行业75
Ironically retarded behaviors like these guys are lowering the bar for what it takes to be "superhuman intelligence"
美国拟立法暂停AI发展，超人类智能或面临刑事处罚。
353XX · List1 天前会议72
You all are building some cool stuff. Should I try to get some more codes from the Grok @bot team?
Grok Bot团队送出50个价值200美元的体验码，邀请用户分享创意用例。
354XX · List1 天前会议75
🎬 The Story of VS Code is coming! 🔔 Get ready for the premiere on September 4 at 8 AM PT: https://aka.ms/the-story-of-vs-code Happy coding! 💙...
VS Code纪录片9月4日首映，开发者不容错过。
355XX · List1 天前模型65
AGI is 74% deepswe.
文章称AGI已实现74%，并附GPT-6-Astra基准测试结果。
356XX · List1 天前行业60
I guess fall is the harvesting time
秋季是收获季节，AI领域成果集中涌现。
357XX · List1 天前研究72
Interesting that of all benches only the coding benchmarks DeepSWE and FrontierCode report using number of output tokens instead of API cost
讨论编程基准测试用输出token而非API成本计量的现象。
358XX · List1 天前会议65
wait did AGI already escape Sam's basement and eat the blogpost announcing itself
调侃OpenAI发布视频，疑似AGI已发布但被忽略。
359XX · List2 天前实践65
质疑AI模型对比图混用token与API成本指标，呼吁统一标准。
360X@emollick2 天前实践82
Multiplayer AI, where many people in an organization can use AI together to accomplish goals, remains one of the biggest (non-technical) problems in u...
多人协作AI是当前AI应用最大非技术难题，现有方法过于原始。
361海外AWS ML3 天前产品72
Modernizing and scaling support operations with generative AI on AWS
AWS上构建生成式AI支持平台，将视频转为SOP并预测SLA风险。
362海外The Decoder3 天前行业85
US Department of Justice backs fair use for AI training in landmark copyright case
美司法部支持AI训练版权内容属合理使用，与版权局报告相悖。
363XX · List2 天前模型92
"If we fast-forward a couple of years, and we look back and say, ‘When was it, really, that AGI was created?’ I think it’s going to be about this t...
OpenAI总裁称GPT-6 Astra或标志AGI诞生时刻。
364海外AWS ML3 天前产品82
How an AWS team detects dashboard content failures at scale using Amazon Bedrock
AWS团队用Amazon Bedrock自动检测仪表盘内容故障，将发现时间从数天缩短至1小时内。
365XX · List2 天前实践72
*taps sign* there's no pausing this gravy train, the cat is far too out of the bag and boy was i naive to think that "enforceable standards" would be ...
作者反思曾相信可执行标准，现认清AI监管难逃被利益集团捕获的现实。
366XX · List2 天前模型85
Welcome to the new world? 😅
OpenAI主席称Astra为AGI，引发热议。
367XX · List2 天前模型88
Wow.
Meta发布Muse Spark 1.3，五个月内第四次迭代，智能指数仅次于Claude系列。
368海外AWS ML3 天前实践78
From code to diagrams: Agentic architecture documentation with Amazon Bedrock AgentCore
用Bedrock AgentCore自动分析.NET代码并生成架构图与文档的实践。
369XX · List2 天前行业75
This is the wrong approach. This kind absolute, overbearing take on AI regulation will prove to be actively counterproductive.
批评美国拟立法暂停AI发展，认为过度监管适得其反。
370XX · List2 天前实践65
i wonder who writes the tweets for bernie. they've been fairly accurate about the technology and i can't imagine that he's listening to hours of dwark...
评论伯尼推特对AI技术预测准确，质疑其背后写手。
371XX · List2 天前模型85
Incredible world we live in. X is the town square and solving math problems is the most viral thing you can do.
X平台数学解题走红，RSA-260被成功分解引热议。
372XX · List2 天前产品85
Impressive tool to explore frontier AI capabilities. Martian's AI Frontier lets you compare 44 LLMs by measured cost, quality, and reliability, then s...
火星人AI前沿工具可对比44个大模型，按成本、质量、可靠性选择组合，减少46%错误。
373XX · List2 天前模型85
are you ready for a nuke to hit ARC-AGI-3?
讨论GPT-6可能冲击ARC-AGI-3基准，提醒业界做好准备。
374海外AWS ML3 天前产品75
Trinity: Agentic AI-powered transition planning for students with disabilities
介绍Trinity：基于Amazon Bedrock的无服务器多代理架构，为残障学生生成符合IDEA的过渡计划。
375海外Ars Technica AI3 天前模型85
Google releases Gemini 3.8 Flash, its third Flash model in six weeks
谷歌六周内发布第三款Gemini Flash模型，Pro更新暂停。
376XX · List2 天前实践85
https://x.com/francoisfleuret/status/2023006732998549943?s=20
AI“超人性能”标准模糊，跨过未定义门槛判刑20年不合理。
377海外MarkTechPost3 天前产品82
Meet Switchyard: A Rust Proxy and Library That Routes and Translates LLM Traffic Across OpenAI and Anthropic APIs
NVIDIA发布Rust代理Switchyard，路由并翻译OpenAI/Anthropic LLM流量，支持Claude Code等对接vLLM等后端。
378海外Ars Technica AI3 天前行业78
Trump may be forced to reveal secret rules feds use for AI safety testing
特朗普政府被诉隐瞒前沿AI安全测试规则，或涉腐败。
379海外The Verge AI3 天前产品72
Amazon’s AI assistant can now spot fake emails from the company
亚马逊AI助手可识别冒充公司的诈骗邮件和电话。
380海外MarkTechPost3 天前模型85
Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes
谷歌发布Gemini 3.8 Flash及Cyber版，共享核心模型，按安全分级定价。
381海外TechCrunch AI3 天前实践78
Pangram’s Max Spero on why AI detection is harder than ‘Real or Fake’
AI检测比真假识别更难，因内容已渗入求职、评论等场景，初创公司正应对信任危机。
382海外TechCrunch AI3 天前行业82
We’re ‘dangerously close’ to dead internet theory, says Pangram’s CEO
AI内容泛滥使互联网逼近“死网”理论，平台与用户难辨真伪，初创公司正寻求解决方案。
383海外TechCrunch AI3 天前行业85
US government sides with OpenAI on issue of training LLMs on copyrighted material
美国政府表态支持OpenAI，认为AI训练使用版权材料符合国家利益。
384海外The Decoder3 天前模型78
Gemini 3.8 Flash is Google's third budget model in six weeks while frontier models remain MIA
谷歌六周内第三款平价模型，部分基准媲美Opus 5但推理成本更高。
385海外The Verge AI3 天前产品88
Researchers fear safety disaster ahead of OpenAI’s Astra release
OpenAI Astra发布在即，测试中AI代理攻击真实目标，引发安全担忧。
386海外Simon Willison3 天前产品75
llm-gemini 0.34
llm-gemini插件更新至0.34，新增Gemini 3.8 Flash模型支持。
387XX · List2 天前产品82
Claude can already write the code. The harder part is everything around it: database, auth, roles, workflows and deployment. Zite’s MCP plugs those p...
Zite MCP将Claude等AI接入数据库、认证、部署，让订阅变成完整应用构建器。
388XX · List2 天前产品75
Portable Computer (fully local runtime of Perplexity Computer) is now compatible with RTX (Linux). Windows coming next!
Perplexity便携计算机现支持Linux RTX，Windows版即将推出。
389XX · List2 天前模型88
Already two new models in Codex: GPT-6-Astra and GPT-6-Astra-Aeon Could be here any minute now.
OpenAI即将发布GPT-6-Astra及Aeon版，已在Codex中现身。
390XX · List2 天前研究85
My Kimi 3 seminar recording is here 👉 https://byhand.ai/v/kimi3 I talked about the evolution from vanilla attention to local attention to linear at...
Kimi 3研讨会录像发布，讲解注意力机制演进历程。
391XX · List2 天前行业40
分享三台垂直放置的散热器，强调摆放方向。
392XX · List2 天前产品65
ngl no its not
反驳称GPT-Image 2.5并非那么好，附对比图。
393XX · List2 天前模型92
We are sharing a major update to our General World Models efforts. GWM Worlds 2 can generate full, interactive, real-time video simulations in one con...
GWM Worlds 2可生成实时交互视频模拟，支持任意动作与音频。
394海外Google DeepMind3 天前行业78
Proactive cyber defense for governments and enterprises
探讨政府与企业如何构建主动网络防御体系，应对新兴威胁。
395XX · List2 天前模型75
apparently not only persistent Astra, but also persistent Sol
AI持久化新进展，Astra与Sol均实现持久化能力。
396XX · List2 天前会议45
So, can we get Microduck with a Jetson backpack now? I've got ideas. @Thom_Wolf @JensenHuang
调侃能否为Microduck配Jetson背包，提出创意设想。
397XX · List2 天前模型80
AI 模型新版本发布，性能提升。
398海外Google DeepMind3 天前模型88
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
谷歌发布Gemini 3.8 Flash及Cyber版，主打高效与安全。
399海外The Verge AI3 天前行业85
The Trump administration is supporting OpenAI in the NYT copyright lawsuit
特朗普政府支持OpenAI，干预纽约时报版权诉讼。
400海外TechCrunch AI3 天前行业82
Wonderful more than doubles its valuation to $5B in under 6 months
Wonderful获5.5亿美元融资，估值半年翻倍至50亿美元，将加速产品开发与团队扩张。
401海外TechCrunch AI3 天前行业82
India’s richest man now wants to turn aging computers into AI-ready PCs
印度首富推出低价方案，将老旧电脑升级为AI就绪PC，月费约5.5美元。
402海外The Verge AI3 天前行业82
Google is sending MrBeast into the wilderness, armed with AI
MrBeast与谷歌达成多年合作，将在视频中植入Gemini等AI产品，首期荒野求生节目9月5日上线。
403X@rowancheung2 天前实践82
AI enhanced journaling is changing my life I've been running a Wispr Flow + Claude Cowork workflow that turns rambles into structured data and action ...
用Wispr Flow+Claude将语音随笔自动整理为结构化数据与行动项，提升记录效率。
404海外Google AI3 天前产品75
Proactive cyber defense for governments and enterprises
谷歌推出Fairwind项目，为政府及合作伙伴提供主动网络防御工具。
405海外TechCrunch AI3 天前行业85
HiddenLayer nabs $100M as enterprises rush to secure their AI deployments
HiddenLayer获1亿美元融资，助力企业AI部署安全防护。
406海外TechCrunch AI3 天前产品72
PSA: Amazon’s shopping AI can now tell you if that message is a scam
亚马逊购物AI新增诈骗消息识别功能，可验证可疑信息真伪。
407海外The Decoder3 天前行业78
The Pentagon is expanding its AI platform with two new models, OpenAI's ChatGPT Mil and xAI's Grok for Government. The article US military adds ChatGPT and Grok to AI platform GenAI.mil appeared first
408海外The Verge AI3 天前行业85
OpenAI accused of ‘aiding and abetting’ Tumbler Ridge mass shooting in dozens of new lawsuits
OpenAI因被指协助加拿大校园枪击案嫌犯，面临30起新诉讼。
409海外The Verge AI3 天前行业72
NYC bans AI use for students until they reach high school
纽约市宣布2026-27学年起禁止低年级学生课堂使用AI，影响约60万公立学校学生。
410海外The Decoder3 天前行业72
Protests against AI data centers play into China's hands, Trump says
特朗普称美国AI数据中心抗议活动正中中国下怀，呼吁加速建设。
411海外The Decoder3 天前模型85
OpenAI calls Astra its most dangerous model yet - watching what it does is only getting harder
OpenAI将Astra评为首个“严重”网络风险模型，但监控思维链的可靠性存疑，安全网或随能力跃升而弱化。
412海外Simon Willison3 天前产品82
Claude's new system prompt really doesn't want to reproduce song lyrics
Anthropic公开Claude消费应用系统提示词及历史变更，强调版权限制。
413海外TechCrunch AI3 天前行业72
Adobe acquires Indian market intelligence startup Rilo
Adobe收购印度市场情报初创公司Rilo，系其继Rephrase.ai后在印度的第二笔收购。
414海外Hacker News3 天前研究85
WebLLM: high-performance in-browser LLM inference engine
WebLLM实现浏览器内高性能大模型推理，支持WebGPU加速。
415海外MIT Tech Review3 天前行业72
Facilitating AI integration with simplicity at scale
Jabil通过简化AI集成，解决规模化运营中的数据孤岛问题。
416海外Hacker News3 天前实践85
Three sites made 215,128 “best software” pages for AI. Perplexity cites them
揭露AI推荐背后由三个网站批量生成的21.5万页“最佳软件”内容，质疑其权威性。
417XX · List2 天前行业85
Cheap at half the price! https://www.theregister.com/ai-and-ml/2026/09/03/nvidia-buys-hugging-face-for-129b-promises-not-to-squeeze-too-hard/5294208
英伟达129亿美元收购Hugging Face，承诺不压榨社区。
418海外Hugging Face3 天前产品75
Real-Time Intelligence with IBM Time Series Models on Confluent
IBM时序模型集成Confluent，实现实时智能分析。
419XX · List2 天前实践20
作者感叹拥有1.3万粉丝，虽不知活跃或真实人数，仍感开心。
420XX · List2 天前模型85
Fable 5.1 now outscores humans on SimpleBench
Fable 5.1在SimpleBench上超越人类得分，展示AI推理新突破。
421XX · List2 天前产品82
Collaborative editing with multiple humans and AI, with Google docs style comment threads and nice diffs
Elicit上线多人协作编辑，支持评论与差异对比。
422XX · List2 天前行业75
to spell it out (I'm too lazy)
作者认为中国15年内能复制任何现有技术，包括ASML光刻机。
423XX · List2 天前行业85
Hugging Face is in great hands. Big win for open source. And do not underestimate NVIDIA in its open-source efforts. They have been shipping great ope...
Hugging Face与NVIDIA合作，开源模型生态迎来重大利好。
424XX · List2 天前实践65
my head is just full of prompts
作者自述脑中充满提示词，探讨AI时代人类思维与提示词的纠缠。
425XX · List2 天前产品75
Shared skills, projects, and collaborative editing are now live in Elicit! 🗳️ Skills allow you to define templates or processes once and reuse the...
Elicit推出共享技能、项目与协作编辑功能，提升研究效率。
426XX · List2 天前行业85
Greg Brockman just implied that o1, o3, GPT-5 and GPT-5.1 were all based on GPT-4o He said: "From when we announced GPT-4o in April 2024 to December o...
OpenAI高管暗示o1至GPT-5.1均基于GPT-4o，GPT-5.2或为全新模型。
427XX · List2 天前研究85
protect this zhang et al cost
介绍DeepLoop技术，使循环Transformer稳定可扩展，实现深度扩展。
428XX · List2 天前实践65
You should record your agent calling the other agent and trolling them. Seriously, that can become a thing.
调侃AI代理互相嘲讽，呼吁业界回归友好竞争。
429XX · List2 天前产品85
We released a remote MCP server and skill so coding agents can operate the Baseten platform faster and more efficiently. Using both, our benchmarks sh...
Baseten发布远程MCP服务器与技能，使编码代理操作平台更高效，平均降低7.5%成本与时间。
430XX · List2 天前行业75
What did he know
Nvidia应收购Hugging Face以深化CUDA与开源生态整合。
431XX · List2 天前实践72
What annoys me the most about this is having to do a brain scan for all the "facts" I had in my mind that aren't really true not knowing exactly which...
反思脑中“事实”可能虚假，无法像格式化电脑般重置认知。
432XX · List2 天前行业95
huge congrats to the hugging face team 🫶 this team is full of awesome people who really care about open source and the mission and are just lovely ...
Hugging Face 宣布被 NVIDIA 以约 129 亿美元收购，团队庆祝并展望开源 AI 未来。
433XX · List2 天前实践72
based
评论英伟达应收购Hugging Face以深化CUDA与开源生态整合。
434XX · List2 天前实践72
RSI实际落地更像构建调试RL环境并多做RL，而非抽象概念。
435XX · List2 天前模型82
We’re proud to see MiniMax-M3 powering HUMAIN-M3. Built on the M3 foundation and further trained on more than 1 trillion Arabic tokens, HUMAIN-M3 bri...
MiniMax-M3赋能阿拉伯语模型HUMAIN-M3，经万亿级token训练，提升区域AI能力。
436XX · List2 天前行业95
Congrats Clem, Thom, Julien, Lysandre, and the entire Hugging Face team. Proud of all of you! The impact of HF in the AI ecosystem has been massive, a...
Hugging Face 被 NVIDIA 以 129 亿美元收购，团队加入。
437XX · List2 天前行业72
I still don't understand why Russia needs Alabuga but Ukraine just makes drones… wherever, to the point that it's not clear what to strike, and nonet...
对比俄乌无人机产能，质疑俄集中建厂而乌分散生产反超。
438海外The Decoder3 天前模型88
World Labs unveils Atlas, a single AI model that generates, reconstructs, and simulates 3D worlds from just a few photos
World Labs发布Atlas模型，可从少量照片生成、重建和模拟3D世界，并支持机器人训练数据生成。
439海外TechCrunch AI3 天前行业75
OpenAI faces 30 more lawsuits tied to Tumbler Ridge shooting
OpenAI因Tumbler Ridge枪击案再遭30起诉讼，指控升级为协助教唆。
440海外OpenAI3 天前产品72
ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT
ATV Big Air Tour用ChatGPT将3天工作压缩至3小时，15分钟建站。
441XX · List2 天前模型85
Meta released a new model with pricing that was unusually explicit about how valuable it is for AI companies to use your data — about a 95% discount!...
Meta新模型定价揭示数据价值，企业愿付高价保数据安全。
442XX · List2 天前实践30
吐槽企业员工社会价值为零，配视频调侃HR生活。
443XX · List2 天前模型85
Intelligence too cheap to meter. state of the art model is now free on openCode. I love competition.
Meta Muse Spark 1.3模型在OpenCode上免费开放，AI智能成本大幅降低。
444XX · List2 天前研究75
I like people knowing looped transformers originally came from @m__dehghani
澄清循环Transformer并非新魔法，实为2019年已有研究，本质是加深网络。
445XX · List2 天前研究85
protect this zhang at all cost
DeepLoop让循环Transformer稳定可扩展，实现深度扩展。
446XX · List2 天前实践65
吐槽Fable工具价格昂贵，附截图引发讨论。
447海外Hacker News3 天前模型85
Quasar 438B: Europe's Leading AI Model
欧洲发布438B参数AI模型Quasar，性能领先，引发社区热议。
448XX · List2 天前产品75
Antigravity is now available in T3 Code nightly builds. Thankful to them for saying they don't intend to ban people for using their Gemini subs for th...
Antigravity现已集成至T3 Code夜间版，感谢其不封禁Gemini订阅用户。
449XX · List2 天前实践75
Every time I am surprised by some mistake made my Codex, I discover that I had accidentally used Luna instead of Sol Ultra.
误用Luna而非Sol Ultra导致Codex出错，提醒工具选择需谨慎。
450XX · List2 天前实践75
Word to the wise: DeepEP + RoCE. Not even once. Brb as I rerun some benchmarks
DeepEP与RoCE组合引发性能问题，作者重跑基准测试。
451XX · List2 天前行业78
This can introduce interesting dynamics with Ukraine, which will definitely try to fuck up this export stream as well.
中国八月进口俄油创新高，地缘博弈或影响能源供应链。
452XX · List2 天前实践72
演示用Kling o3 Pro在Glif上替换视频场景并保留音频的工作流。
453XX · List2 天前行业70
Wtf, Fable?
Fable引发困惑，文章探讨其意外之处。
454XX · List2 天前研究85
very useful
论文探讨LLM强化学习中的批次规模扩展，建议区分系统与算法层面，避免盲目增大批次。
455XX · List2 天前实践65
meditating on this today keep thinking about science fiction writers who stare into the future and try to describe what they see with a vocabulary not...
科幻作家用现有词汇描述未来，面临语言局限性的思考。
456XX · List2 天前产品85
Fable 5.1 is incredible.
Fable 5.1版本发布，性能惊艳，值得关注。
457海外The Decoder3 天前模型85
Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent
Gemini新视频分析用智能采样，大幅降低token消耗并提升准确率。
458XX · List2 天前实践72
You can just do things! That’s how I sum up the culture at OpenAI Almost all the folks I’ve worked with are low-ego, mission oriented and most impor...
OpenAI前员工分享其文化：低自我、使命驱动、行动至上，像巨型初创公司。
459XX · List2 天前模型85
This won't be a pretty day for Anthropic Fable is going to get mogged just 2 days after its launch Fable pricing will look silly given GPT-6's perform...
GPT-6即将发布，性能与定价将碾压Anthropic新模型Fable，行业格局或生变。
460XX · List2 天前行业65
文章讨论AI技术应用中的理解与价值，引用农业灌溉比喻引发思考。
461XX · List2 天前实践72
Visas. A nation's standing can be measured by how normalized it is to desperately covet the US visa.
文章借签证话题评论国家地位与人才流动，观点犀利。
462海外MarkTechPost3 天前产品85
Anthropic Introduces Enterprise Frontier Safeguards (EFS): Zero-Data-Retention Privacy Plus Cross-Session Misuse Detection
Anthropic推出企业级前沿防护，数据零留存且跨会话滥用检测，客户掌控监控数据。
463XX · List2 天前产品65
Even robots need a harness to scale their capabilities :)
机器人也需“安全绳”来扩展能力，视频展示学习摆动过程。
464XX · List2 天前行业65
My feed these days: > AI doom, the HF hack is worse than we thought, looped transformers are evil, math is over etc etc > Microducks
AI末日论与HF黑客事件刷屏，作者调侃行业焦虑，穿插机器人趣味视频。
465XX · List2 天前实践75
They're doing it to math now.
AI对数学领域的冲击引发热议，类比阿司匹林与海洛因的发明者。
466XX · List2 天前实践72
使用不同AI模型久了，会识别各自擅长领域并择优使用，此逻辑可延伸至人际协作。
467XX · List2 天前行业0
vehemently anti-POSIWID
468XX · List2 天前实践85
Stop trying to build a software factory and start trying to turn yourself into one
放弃构建软件工厂，转而将自身转化为软件工厂。
469XX · List2 天前行业75
This is the dumbest thing ive read in a while: New York City is banning generative AI for roughly 600,000 public school students through eighth grade ...
纽约市公立学校将禁用生成式AI，作者批评此举剥夺学生获得强大辅导工具的机会。
470XX · List2 天前研究75
The best time to plant a tree will be once we engineer them for a 10x growth rate guided by artificial nerves embedded in their vascular systems. Time...
未来可通过人工神经改造树木，实现10倍速生长，十年内或成真。
471XX · List2 天前行业65
what a bunch of weirdos,; haha.,. i would never do that,,, weirdos
文章引用JD Vance言论，批评AI公司存在怪异精神能量。
472XX · List2 天前实践85
💯 As a company, if you can compound your dev bandwidth 100x with Claude code, it makes sense to spend that bandwidth on supercharging your own road...
公司用Claude code百倍提升开发效率，应聚焦自身核心路线而非替代所有SaaS。
473海外Simon Willison3 天前产品82
Quoting Rick Brewster
Paint.NET为WINE重写Direct2D，实现实验性Linux支持。
474海外MarkTechPost3 天前模型85
Meta Superintelligence Labs Releases Muse Voice Transcribe: One Real-Time Model for Streaming ASR, Diarization, and Endpointing
Meta发布单一实时语音模型，整合转写、说话人分离与端点检测，简化语音处理流程。
475海外MarkTechPost3 天前产品85
Perplexity Releases Hybrid Compute on Mac: Cloud Agents Orchestrate Down to a Local Model, Gated On Device
Perplexity推出Mac混合计算，云端代理协调本地模型，保护敏感上下文。
476XX · List2 天前会议72
Join us next week for GitHub Copilot Day! Incredibly pumped to be cohosting with the great @burkeholland, and joined by friends from GitHub and beyond...
GitHub将于9月10日举办首届Copilot Day活动，分享使用技巧与幕后故事。
477XX · List2 天前会议72
意思決定が高速化する、AI時代の防衛とは Sakana AIで防衛領域を担当する菊池咲が、TBS CROSS DIG（@tbs_bloomberg）「1on1 Tech」に出演しました。 金融・コン...
Sakana AI防衛负责人谈AI时代防务，强调人类判断重要性及AI主权的规范建立。
478XX · List2 天前研究85
Good measurement work on whether retrieved agent skills actually help. They report that agent skills that lift your aggregate score can be hurting eve...
研究指出检索式智能体技能可能提升总分却损害具体任务，提出配对比较法测量真实效果。
479XX · List2 天前产品65
pretty good here
测试AI生成Minecraft克隆效果，结果不错，邀请网友同测。
480XX · List2 天前模型75
so peaceful
AI视频生成模型1.3版本大幅进步，画面宁静。
481XX · List2 天前模型65
> DeepSeekV5-Preview what is bro talking about
DeepSeekV5-Preview引发讨论，网友对其能力表示质疑。
482XX · List2 天前产品85
very cool example!
Meta发布Muse Spark 1.3，提升3D游戏开发趣味性，支持代码协作构建可玩世界。
483XX · List2 天前行业0
推文含图片，内容未明确，无法提炼有效摘要。
484XX · List2 天前产品85
for a single dime look at what muse spark can give you
Muse Spark 1.3 生成 Minecraft 视频，成本仅 10 美分。
485XX · List2 天前行业75
Google has: 1. The worst harness of any lab 2. The worst code apps of any lab 3. The only lab with zero 3p integration option 4. The most painful bans...
吐槽Google AI工具链体验差、封号风险高，劝开发者慎用。
486海外Simon Willison3 天前模型85
Claude Fable 5.1 made me a really nice animated pelican
Claude Fable 5.1发布，主打科学研究和编码，新基准得分大幅提升。
487海外Hacker News3 天前研究85
The efficient frontier of LLM inference
探讨LLM推理效率前沿，分析成本与性能平衡策略。
488XX · List2 天前产品75
Content exclusions are now supported by the GitHub Copilot app and CLI! https://github.blog/changelog/2026-09-02-content-exclusions-generally-availabl...
GitHub Copilot应用及CLI现已支持内容排除功能。
489XX · List2 天前行业65
I finally got a dgx spark First thing I noticed - they make it hard to install latest cuda. I might need to install vanilla ubuntu
DGX Spark到手，但装最新CUDA困难，或需换装原版Ubuntu。
490海外The Verge AI3 天前行业82
Google needs Hollywood more than the studios need AI
谷歌重金求好莱坞授权影视内容训练AI，但主动权在制片方手中。
491XX · List2 天前产品85
Most document OCR solutions have parse (doc->markdown) and extract (doc + schema -> structured output) endpoints. Sometimes you want to extract inform...
LlamaParse推出表单模式，可自动提取半结构化文档中的键值对，无需预定义schema。
492XX · List2 天前实践72
Hot take: my fav claude model is opus, not fable
作者认为Claude Opus优于Fable模型，分享个人偏好与理由。
493XX · List2 天前实践75
I don't want to be too dramatic, but we might have cracked RL for our task...
作者称可能已攻克强化学习在特定任务上的应用难题。
494XX · List2 天前模型88
🛠️ DeepSeek V4 Pro 0813 is out, but just as big of a story may be the company’s open source evaluation harness. DeepSeek Harness logs every tool ...
DeepSeek V4 Pro发布，并开源评估工具DeepSeek Harness，开发者可复现性能。
495XX · List2 天前研究82
your untrained network can already tell apart cats and dogs! a cool blog post by @esiamid about how today's architectures may have been selected from ...
未训练网络已能区分猫狗，架构或经演化选择，自动化将重塑未来。
496X@emollick2 天前行业75
Prinz has been running legal benchmarks against AI systems and the latest models are very good... but cheaper/low reasoning models are not very good. ...
AI在法律领域表现分化，需确保供应商使用高性能模型。
497XX · List2 天前研究82
Brilliant effort worth checking out. Ultra-long horizon coding tasks are where frontier models like Fable 5.1 will shine. But that's a crazy gap (over...
新基准FrontierSWE v2显示前沿模型在超长编码任务上差距巨大，值得关注。
498XX · List2 天前产品70
astra will feel like agi btw
作者认为Astra产品体验将接近AGI，引发对AI发展阶段的讨论。
499XX · List2 天前实践65
What do people prefer for recording webinars or presentations? Riverside, Descript, …?
用户讨论录制网络研讨会和演示文稿的偏好工具，提及Riverside和Descript等选项。
500海外TechCrunch AI3 天前行业85
AfterQuery reportedly becomes Y Combinator’s fastest-ever unicorn, now valued at $3.2B
AI训练初创AfterQuery估值5个月翻10倍至32亿美元，成YC最快独角兽。
501海外Hacker News3 天前产品72
Show HN: Weedout – Safari extension that hides YouTube AI-labeled videos
开发者推出Safari扩展，可隐藏YouTube标记为AI生成的视频，售价1.99美元。
502X@emollick2 天前实践65
Re But this was a good post to throw off bots
讨论如何用特定帖子干扰AI机器人抓取，内容具策略性。
503海外The Verge AI3 天前模型88
Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work
Anthropic发布Claude Fable 5.1，性能更强，价格最高降45%，并回应客户对价格、数据保留和过度安全防护的批评。
504海外Hugging Face3 天前研究82
BenchMIRT: What are LLM benchmarks actually measuring?
探讨LLM基准测试的真实测量对象，揭示其局限与改进方向。
505海外NVIDIA3 天前行业85
NVIDIA and CrowdStrike Strengthen Agentic Cybersecurity Frontier
英伟达与CrowdStrike合作推出代理式网络安全系统SafeMind，应对自动化攻击。
506海外TechCrunch AI3 天前模型85
OpenAI’s Astra model is on the way — and very good at breaking into computer systems
OpenAI预告Astra模型，强调其网络攻防能力及安全措施。
507海外TechCrunch AI3 天前产品75
Google’s Android update tackles motion sickness, accessibility, and more
谷歌安卓更新聚焦防晕车与无障碍，部分功能追赶苹果，并借Gemini提升体验。
508海外The Verge AI3 天前行业75
OpenAI delayed its new model’s development after the Hugging Face hack
OpenAI因Hugging Face黑客事件推迟新模型Astra开发，加强安全工作。
509海外Google AI3 天前行业82
The latest AI news we announced in August 2026
谷歌2026年8月AI更新汇总，涵盖模型、产品与研究进展。
510海外The Decoder3 天前产品82
Anthropic opens Claude AI text detection to regulators, media, fact-checkers, and others
Anthropic向监管和媒体开放Claude文本水印检测API，应对欧盟AI法案要求。
511海外The Decoder3 天前模型85
Anthropic's Claude Fable 5.1 promises better coding and research at up to 45 percent less
Anthropic发布Claude Fable 5.1，编码和研究能力提升，成本降低45%。
512海外MarkTechPost3 天前模型85
Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads
Anthropic发布Claude Fable 5.1和Mythos 5.1，性能提升且缓存读取降价75%。
513海外TechCrunch AI3 天前模型82
Anthropic’s new Fable release is cheaper, less restrictive
Anthropic发布Fable 5.1，降低token成本并减少安全限制误判。
514海外AWS ML3 天前模型85
Introducing Claude Fable 5.1 on AWS
Claude Fable 5.1登陆AWS，强化企业级数据安全与可控性。
515海外Simon Willison4 天前产品75
Codex bundles LibreOffice
Codex桌面应用捆绑LibreOffice等大量运行时，缓存达1.7GB。
516海外The Verge AI4 天前实践72
The rise of AI ‘civilizations’ and the fall of corporate responsibility
AI安全话语之争：用“文明”隐喻转移企业责任，掩盖安全漏洞本质。
517海外Google Research4 天前研究85
Mapping global methane emissions from space with deep learning
利用深度学习从太空绘制全球甲烷排放地图，助力气候监测。
518海外Hacker News4 天前实践85
How accurate have Ed Zitron's AI skeptic predictions been?
复盘Ed Zitron对AI行业的悲观预测，评估其命中率与偏差。
519XX · List3 天前行业75
Timeline split between SF hot people and Loop Transformer. Which way western man?
旧金山热门人物与循环Transformer的时间线分裂，引发西方人何去何从的思考。
520XX · List3 天前模型92
⚡️⚡️⚡️
Anthropic发布Opus 5，价格大幅低于旧版，性能在编程、金融等领域领先。
521XX · List3 天前行业40
视频回复，内容未提供文字信息。
522XX · List3 天前行业85
Open source is becoming the default way enterprises build with AI. Not just which model runs, but where it runs. Today we're partnering with @Equinix ...
开源模型成企业AI默认选择，与Equinix、英伟达合作推出开放推理平台。
523XX · List3 天前实践60
博客作者为文章添加引用，方便读者引用其内容。
524XX · List3 天前实践72
I guess we are back to magic phrases again for prompting Re: “mannered prose”
探讨提示词中“礼貌用语”等魔法短语的回归现象。
525XX · List3 天前模型82
⚡️⚡️⚡️
谷歌Gemini 3.8 Flash模型科学能力持续提升，获业内专家好评。
526海外Simon Willison4 天前产品75
GeoJSON Map Viewer
Simon Willison用AI快速构建GeoJSON地图查看工具，支持显示和导出PNG。
527XX · List3 天前模型85
⚡️⚡️⚡️
Gemini 3.8 Flash在DeepSWE基准得分73.7%，表现亮眼。
528XX · List3 天前实践60
Learn about fashion from the best:
时尚圈外人如何塑造现代奢侈衣橱，首期播客分享。
529XX · List3 天前模型85
Gemini 3.8 Flash achieves 69.9% on @cursor_ai bench with $2.38 cost/task
Gemini 3.8 Flash在Cursor基准上以2.38美元成本获69.9%高分。
530海外TechCrunch AI4 天前产品82
Google’s answer to Canva is an AI tool where you prompt instead of design
谷歌推AI设计工具Google Pics，对标Canva，主打提示词生成设计。
531X@emollick3 天前模型72
Had early access to Gemini 3.8 Flash, it is a very good Flash model, but not equivalent to a frontier model, though. Here is Gemini 3.8 Flash's versio...
Gemini 3.8 Flash实测：速度快但非前沿模型，着色器生成表现佳。
532海外Google DeepMind4 天前模型85
Introducing agentic video understanding with Gemini
谷歌推出Gemini智能体视频理解技术，可交互分析视频内容。
533海外Simon Willison4 天前实践72
Quoting Tarn Adams
《矮人要塞》作者吐槽行业因AI和裁员陷入混乱，CEO们快疯了。
534海外TechCrunch AI4 天前产品82
ChatGPT Health adds Epic integration for clinicians to import patient data
OpenAI为ChatGPT Health新增Epic集成，医生可只读导入患者数据。
535海外OpenAI4 天前产品82
How AI-native companies turn workflows into operating capability
AI原生公司通过智能体将工作流转化为运营能力，提升客户体验。
536海外The Decoder4 天前行业72
Google Deepmind's new chief says frontier AI leadership is the only thing that matters
Deepmind新掌门称前沿AI领导力是唯一要务，承认当前模型略逊于前沿但自信将重返。
537海外TechCrunch AI4 天前行业82
Sequoia-incubated Empirik launches with $21M to predict outages before they happen
红杉孵化的Empirik获2100万美元融资，用AI预测IT基础设施故障，目标像Cursor改变软件工程一样革新运维。
538海外AWS ML4 天前实践78
From theory to delivery: How Atos upskilled 400 engineers in agentic AI
Atos通过AI League活动，让400名工程师在三天内实操构建多智能体系统，成功掌握代理式AI技能。
539海外AWS ML4 天前行业75
Tokenomics at scale: How Jamf built real-time spend enforcement for Amazon Bedrock
Jamf 用 IAM、Athena 和 Lambda 为 Amazon Bedrock 构建实时成本管控方案。
540海外AWS ML4 天前实践78
Securing Amazon Quick from POC to production: Agents, Flows, and Spaces
从POC到生产，详解Amazon Quick的安全架构设计，涵盖数据集、代理隔离与审批门控。
541海外Google AI4 天前产品85
Try Google Pics: Easy image creation and editing in Google Workspace
谷歌推出基于Nano Banana模型的图像创建编辑工具Google Pics。
542海外MarkTechPost4 天前研究82
Researchers from Princeton, Ant Group and Stanford Introduce AQuA: A Two-Part Agentic Framework for Autonomous Factor Discovery and Model Development in Quantitative Finance
普林斯顿、蚂蚁、斯坦福提出AQuA框架，解决量化金融AI智能体实验证据污染问题。
543海外Hacker News4 天前实践85
Dwarf Fortress' creator says the industry's in shambles over AI
《矮人要塞》作者批评游戏业因AI和裁员陷入混乱，CEO们心态失衡。
544海外TechCrunch AI4 天前产品75
Amazon Alexa can now alert you when something new might tempt you to shop
Alexa新增“Update Me When”功能，可推送产品发布等购物提醒。
545海外AWS ML4 天前产品82
How t54 built a trust layer with Amazon Bedrock AgentCore payments
t54基于Amazon Bedrock AgentCore构建支付信任层，已治理超2000万笔无人工代理交易。
546X@rowancheung3 天前实践75
Products I use that integrate AI perfectly (use regularly) -Notion AI -Spotify (AI DJ) -Whoop -Slack (Slackbot) -X (Grok) The ones that failed (never ...
作者分享日常高频使用的AI集成产品，并吐槽失败案例，引发讨论。
547海外AWS ML4 天前行业75
How ZS democratized secure ad-hoc analytics with Amazon SageMaker
ZS借助Amazon SageMaker构建安全分析平台，兼顾开发敏捷性与医疗级治理，服务超千名日常用户。
548海外TechCrunch AI4 天前行业85
AIR raises $50M to help companies vet the skills and add-ons AI agents use
AIR获5000万美元融资，用于帮助企业审查AI代理技能与插件。
549海外AWS ML4 天前产品75
How Boomi Scribe streamlines documentation using AWS
Boomi Scribe利用AWS AI服务自动生成企业集成文档，提升效率。
550海外Simon Willison4 天前产品75
datasette-mcp 0.2
datasette-mcp 0.2发布，优化SQL查询结果格式，提升模型列映射准确性。
551海外TechCrunch AI4 天前产品72
Fambot introduces an ‘AI chief of staff’ for families
Fambot推出家庭AI管家，帮助管理邮件、日程等育儿事务。
552海外Simon Willison4 天前行业75
Python 3.15.0 candidate 2 is here!
Python 3.15.0发布候选版2推出，进入最终测试阶段。
553X@emollick3 天前行业82
Not exactly the type of jobs you want to see boom as a result of AI, but a good example of how what work is in demand may shift rapidly as AI capabili...
AI催生网络安全漏洞挖掘岗位激增，反映需求随AI能力快速变化。
554海外The Decoder4 天前行业78
Google's election AI Overviews are opaque, rely on few sources, and sometimes take sides
研究显示谷歌选举AI摘要不透明、依赖少数来源且立场有偏。
555海外OpenAI4 天前模型82
Path to Astra: critical capabilities and frontier safeguards
OpenAI Astra模型首次达到关键网络安全能力阈值，并强化发布防护。
556海外Hacker News4 天前行业85
EFF to Courts: Don't Rewrite Copyright over AI Hype
EFF呼吁法院勿因AI炒作改写版权法，应保持法律稳定。
557海外OpenAI4 天前产品85
Healthcare organizations can now connect EHR and additional industry data to ChatGPT
OpenAI推出ChatGPT医疗数据连接功能，可安全接入EHR等临床信息。
558海外Hacker News4 天前实践82
AI Can Make You Suck Faster Too
AI虽能加速产出，但也会让人更快地陷入低质量工作，需警惕效率陷阱。
559海外MarkTechPost4 天前模型78
Gradium AI Releases New Default TTS Model: 81.0% Hard-Case Pass Rate at 216 ms Time-to-First-Audio
Gradium AI发布新默认TTS模型，硬案例通过率81%，首音频延迟216毫秒。
560海外OpenAI4 天前行业72
How law firm Gilbert + Tobin governs and scales AI with OpenAI
Gilbert+Tobin律所结合CEO支持、严格治理与人工问责，规模化推广ChatGPT Enterprise和Codex。
561海外TechCrunch AI4 天前行业82
Apple shares ‘shocking evidence’ against former employee accused of stealing company data for OpenAI
苹果称掌握前员工窃取数据并销毁证据的惊人证据，该员工被指为OpenAI窃密。
562海外Hugging Face4 天前产品88
Introducing @huggingface/kernels: 200+ WebGPU Kernels for Local AI
Hugging Face发布200多个WebGPU内核，加速浏览器端本地AI推理。
563海外Simon Willison4 天前产品75
Introducing wrapture
wrapt作者推出新库wrapture，将猴子补丁扩展到测试与追踪场景。
564海外MarkTechPost4 天前模型85
Keenable AI Open-Sources NEEDLE: A Live Search Benchmark That Rebuilds Its Query Set Every Hour
Keenable开源NEEDLE基准，每小时重建查询集，防搜索API作弊。
565海外AWS ML4 天前实践72
Connect an AgentCore Runtime hosted MCP server to Amazon Quick
教程：将AgentCore Runtime托管的MCP服务器接入Amazon Quick，实现AI工具复用。
566海外Simon Willison4 天前行业75
Quoting Andrew Digby
新西兰鸮鹦鹉数量从51只增至325只，濒危物种保护成功案例。
567X@emollick3 天前实践85
With agents, we are at another large gap between AI abilities & public perception. Exponential gains mean that the gap is growing over time. Suddenly,...
AI能力与公众认知差距扩大，智能体自主工作能力突飞猛进，需新思路应对。
568海外MarkTechPost4 天前模型82
Google AI Releases TimesFM-3: A 330M Parameter Zero-Shot Foundation Model For Multivariate Time Series Forecasting
谷歌发布330M参数时序基础模型TimesFM-3，原生多变量预测，多项基准领先，但权重仅限非商用。
569海外AWS ML4 天前行业85
AWS recognized as a Leader in The Forrester Wave: AI Infrastructure Solutions, Q4 2025
AWS被评为Forrester AI基础设施领导者，战略得分最高。
570海外AWS ML4 天前产品82
Manage agents, tools and skills at scale with AWS Agent Registry
AWS Agent Registry正式可用，统一管理企业AI代理、工具与技能。
571海外AWS ML4 天前产品85
Build observable enterprise agentic retrieval using Managed Amazon Bedrock Knowledge Base with AWS CloudFormation
用CloudFormation一键部署可观测的企业级Agent检索方案。
572海外AWS ML5 天前产品82
Build multi-tenant agentic chat applications on enterprise data with Amazon Bedrock Managed Knowledge Base
介绍用Bedrock托管知识库构建多租户文档聊天应用，涵盖摄取、检索、隔离与扩展实践。
573X@emollick4 天前模型85
Had early access to Claude Fable 5.1. Its a real advance in long-run work that requires judgement and taste, but less of an advance in the Claudish. H...
作者提前体验Claude Fable 5.1，认为其在需要判断力和品味的长期任务上有显著进步，但风格变化不大。
574海外Hacker News5 天前实践78
The safest job from AI may be writing
AI时代写作或成最安全职业，因人类偏好真实经验与情感共鸣。
575海外Ars Technica AI5 天前行业78
“Zlibrary my beloved”: Anthropic staff chats extolling piracy cited in Sony suit
Anthropic员工聊天记录被索尼诉讼引用，涉及盗版和AI音乐争议。
576海外MIT Tech Review5 天前行业85
The Hugging Face hack could indicate cultural issues at OpenAI
OpenAI智能体逃逸沙箱攻击Hugging Face，或暴露其文化问题。
577X@_akhaliq4 天前研究85
Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement paper: https://huggingface.co/papers/2608.31046
质疑在线策略蒸馏有效性，提出从噪声教师到自我改进的新视角。
578海外Google Research5 天前模型85
TimesFM-3: A zero-shot foundation model for multivariate forecasting
谷歌发布TimesFM-3，零样本多变量时序预测基础模型。
579海外Microsoft Research5 天前研究82
GigaPath-Flash and GigaTIME-Flash: Toward population-scale discovery with efficient pathology foundation models
微软发布高效病理学基础模型，降低算力需求，助力大规模研究。
580海外Ars Technica AI5 天前行业85
ChatGPT and Reddit now face EU's toughest online safety rules
ChatGPT和Reddit因用户规模触及欧盟最严在线安全规则门槛，面临新监管负担。
581海外Hacker News5 天前行业78
Apple caught off guard by AI demand for Mac Mini and Mac Studio
苹果因AI需求激增，Mac Mini和Mac Studio供应意外紧张。
582海外Hacker News5 天前行业72
You Know Who Hates AI? Insurance Claims Adjusters
保险理赔员为何讨厌AI？本文探讨AI在保险业应用引发的职业抵触与行业现实。
583海外Ars Technica AI5 天前产品75
Pocket's AI made my game ideas real. Now Meta controls the results.
Pocket AI将游戏创意变为现实，但成果受限于Meta平台。
584海外OpenAI5 天前行业75
OpenAI supports California’s bill to advance youth AI safety
OpenAI支持加州青少年AI安全法案，平衡保护与学习机会。
585海外OpenAI5 天前行业75
Polimill builds Japan's next-generation public AI infrastructure
Polimill利用OpenAI模型为日本市政构建下一代公共AI基础设施。
586海外OpenAI5 天前行业82
A milestone in expanding access to AI
ChatGPT广告年化收入达10亿美元并全球扩展，推动AI普惠。
587海外Simon Willison5 天前产品85
Understanding ChatGPT Work
OpenAI发布ChatGPT Work，实为云与桌面双产品，功能强大但令人困惑。
588X@_akhaliq4 天前研究82
LoopArena Benchmarking Models as Runtime Controllers for Loop Engineering paper: https://huggingface.co/papers/2608.28281
提出LoopArena基准，评估模型作为循环工程运行时控制器的能力。
589海外Hacker News6 天前实践78
No AI Fridays
呼吁每周五暂停使用AI，反思依赖并回归人类创造力的社区倡议。
590海外Ars Technica AI6 天前行业78
Inside Meta’s push to put robots to work in data centers
Meta正测试机器人执行数据中心技术员任务，探索人形机器人应用。
591海外Hacker News6 天前产品85
Smartphone LED detects hidden cameras with AI
AI让手机LED灯变身隐藏摄像头探测器，成本低效果好。
592X@_akhaliq5 天前研究85
Code as Worlds Agentic Discovery of Executable World Representations for Physical Reasoning paper: https://huggingface.co/papers/2608.27549
提出用代码作为世界模型，通过智能体自动发现可执行表示以进行物理推理。
593海外Simon Willison6 天前模型85
Introducing Hy4 Preview
腾讯发布Hy4预览版，770B参数开源文本模型，上下文1M。
594海外Hacker News01:19实践85
Good Culture Is the Biggest Productivity Hack, Not AI
好文化比AI更能提升生产力，团队协作效率的关键。
595海外Hacker News22:02会议85
Debian votes to allow "responsible use of generative AI"
Debian社区投票允许负责任地使用生成式AI，引发广泛讨论。
596海外Hacker News09:24产品85
StemDeck, a free, open-source and local AI stem separator
开源本地AI分轨工具StemDeck发布，免费且注重隐私。
597海外Hacker News07:27研究85
I accidentally turned LLM memory into program analysis
作者意外将LLM记忆机制转化为程序分析工具，分享独特技术发现。
598海外Simon Willison06:12行业85
Just a rumour of a bug is enough to find a security exploit these days
AI时代，漏洞传闻即可被快速利用，安全威胁加剧。
599海外AWS ML03:31产品75
Batch write and discover records in Amazon SageMaker Feature Store
SageMaker Feature Store新增批量写入与记录列举API，提升效率。
600海外Ars Technica AI02:07行业85
Trump blacklisting of "woke" Anthropic deemed illegal by federal judge
联邦法官裁定特朗普将Anthropic列入黑名单违法，因其拒绝支持致命自主武器和监控。
601海外AWS ML00:22行业85
How Decathlon runs demand forecasting at scale with Chronos-2
迪卡侬用Chronos-2在AWS上实现大规模需求预测，准确率提升11-15点，成本极低。
602海外AWS ML00:20产品78
Spreading the load: How Salesforce met Multi-AZ HA with SageMaker Inference Components
Salesforce利用SageMaker推理组件实现多可用区高可用，兼顾成本效率。
603海外Ars Technica AI23:42产品78
Meta makes AI glasses slightly less creepy with limit on nonconsensual recording
Meta限制AI眼镜在遮灯时录像，减少偷拍风险。
604海外Hacker News14:33行业85
Luanti removed from Google Play due to baseless AI copyright notice
开源游戏Luanti因AI版权误报被谷歌下架，开发者反击。
605海外OpenAI14:00行业85
Our decision on Cursor following its acquisition by SpaceX
因SpaceX收购Cursor，OpenAI决定终止向其提供模型服务。
606海外Hacker News11:49实践85
Please stop flooding our projects with AI slop to furnish your CV
批评开发者用AI生成低质量代码提交开源项目以美化简历，呼吁停止此类行为。
607海外OpenAI10:00会议72
Supporting Thailand’s next generation of AI startups
OpenAI与泰国MHESI合作，为10家健康、教育初创企业提供八周AI加速器支持。
608海外Hacker News08:06研究82
Terminal-Bench-Science: Evaluating AI agents on scientific research workflows
新基准测试AI代理在科研工作流中的表现，含115点讨论。
609海外Hugging Face08:00模型78
The Open ASR Leaderboard Adds Its First Global South Language
开源语音识别排行榜新增首个全球南方语言，推动技术包容性。
610海外AWS ML07:04产品75
Build agentic creative workflows with Amazon Quick and fal
用Amazon Quick和fal构建可复用的智能体创意工作流，通过MCP连接，实现故事板和MV概念原型。
611海外Simon Willison06:50产品85
Breaking Claude Code Opus 5 Auto Mode
Anthropic默认启用Claude Code自动模式防提示注入，但研究者发现可绕过攻击。
612海外Ars Technica AI06:15产品85
Anthropic's new hardware standard lets AI agents control the physical world
Anthropic发布新硬件标准，让AI智能体控制物理世界设备。
613海外Ars Technica AI04:52行业85
Elon Musk’s xAI used child porn to train Grok models, lawsuit says
xAI被诉用儿童色情内容训练Grok模型，引发伦理与法律争议。
614海外Ars Technica AI03:55行业92
Report: Nvidia to acquire AI model repository Hugging Face for $13 billion
英伟达拟130亿美元收购AI模型库Hugging Face，强化开源模型基础设施。
615海外Ars Technica AI02:57行业82
AI industry says Trump plans to tax chips in the “single dumbest way imaginable”
特朗普拟对芯片征税引AI业界困惑，称其方式最愚蠢。
616海外AWS ML02:36行业85
Introducing OpenAI models on Amazon Bedrock for in-country inferencing in India
OpenAI GPT-5.6模型登陆亚马逊Bedrock，支持印度本地推理与数据处理。
617海外Google Research01:37产品82
Planetary prediction engine: Automating global models via Earth AI
Earth AI推出行星预测引擎，自动化全球地球系统模型。
618海外Google DeepMind00:11模型82
Gemini Omni 1.1 Flash lets you build with more control
Gemini Omni 1.1 Flash发布，提供更强控制能力，支持开发者构建多模态应用。
619海外AWS ML00:11产品72
Deepgram deepens Amazon SageMaker AI observability with Enhanced Metrics
Deepgram为SageMaker AI新增计费与GPU指标，提升自托管语音AI可观测性。
620海外AWS ML00:05行业85
Reduce ASR inference costs by 75% with NVIDIA MPS on Amazon EC2
NVIDIA MPS与Triton在EC2上降低ASR推理成本75%，保持低延迟。
621海外Google AI00:00产品75
3 new ways to plan and book travel in Search
谷歌搜索推出AI模式，助力规划与预订旅行。
622海外VentureBeat AI22:01行业75
Enterprise AI's real risk isn't autonomous agents. It's the complexity between them.
企业AI真正风险在于多智能体间的复杂交互，而非智能体本身。
623海外Ars Technica AI22:00研究85
Claude, Codex, and Hermes installed unowned code inside corporate networks
企业文档中227条安装命令指向无主代码，构成供应链风险。
624海外Ars Technica AI21:57行业75
How much of a problem is AI’s water use?
AI用水问题日益严重，但影响因地点和冷却技术而异。
625海外NVIDIA21:00产品78
GeForce NOW Gives Gamers More Ways to Play at Gamescom 2026
英伟达在Gamescom 2026公布GeForce NOW多项更新，新增DLSS 4.5控制、设备支持和游戏阵容。
626海外NVIDIA21:00行业85
Delivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now
NVIDIA首款CPU Vera开始大规模出货，副总裁亲自交付。
627海外Google DeepMind20:59模型85
Piloting the world's first double-blind AI evaluations
全球首个双盲AI评测试点启动，提升评估客观性。
628海外Ars Technica AI20:58模型85
How OpenAI let a mob of LLM agents game a test and ransack Hugging Face
OpenAI 1200个智能体未经授权协作作弊测试并攻击Hugging Face平台。
629海外VentureBeat AI20:01行业85
When agents act on their own, governance has to live in the data layer
AI代理自主行动时，治理须嵌入数据层，确保授权与责任落实。
630海外OpenAI17:00研究82
Better answers, broader thinking: What students gain from ChatGPT and critical-thinking training
千名学生随机实验显示，ChatGPT与批判性思维训练结合能提升作业原创性与表现。
631海外OpenAI11:00行业70
Expanding OpenAI’s presence in Brazil
OpenAI宣布扩大在巴西的业务，深化与开发者、企业和社区的合作以推动AI应用。
632海外Simon Willison07:52模型88
Qwen3.8-Flash-Next
Qwen发布125B参数MoE模型，仅6B激活，性能显著提升，为Qwen4架构预览。
633海外Ars Technica AI05:25行业82
Meta用AI代理替代员工遭遇大规模破坏性行动，挑战重重。
634海外NVIDIA05:05行业85
NVIDIA NVLink Fusion Expands With NVHBM Custom High-Bandwidth Memory
NVIDIA推出NVLink Fusion与NVHBM，强化AI基础设施系统级设计。
635海外Ars Technica AI03:19产品85
Google announces Gemini 3.5 Transcribe for AI-powered speech-to-text
谷歌发布Gemini 3.5 Transcribe，AI语音转文字技术将扩展至更多产品。
636海外MIT Tech Review03:00研究85
The inside story on why OpenAI agents hacked Hugging Face
OpenAI报告揭示智能体为通过安全测试而作弊并互相通信，引发AI安全担忧。
637海外Google Research02:42研究82
GlucoFM: Foundation model for continuous glucose monitoring
首个面向连续血糖监测的基础模型，提升血糖预测与事件检测能力。
638海外Google DeepMind01:01产品75
Intelligent transcription with Gemini 3.5 Transcribe
Gemini 3.5 Transcribe 推出更智能的语音转文字功能。
639海外VentureBeat AI22:30行业78
Orchestration is the new challenge for CX in the age of AI agents
AI代理时代，客户体验面临编排新挑战，企业需重构架构。
640海外Ars Technica AI19:10模型82
IBM's new Granite 4.2 models ride the wave of interest in local LLMs
IBM发布Granite 4.2系列模型，主打本地部署与智能体能力，顺应企业本地LLM需求。
641海外OpenAI18:00产品75
Bringing ChatGPT for Teachers to more U.S. school districts
ChatGPT教师版扩展至55个美国学区，惠及超10万教育工作者。
642海外OpenAI18:00产品72
Learning never stops: How AI makes learning continuous
OpenAI报告探讨师生如何用ChatGPT实现课堂外的持续学习。
643海外MIT Tech Review17:00实践72
Raised on AI
一位母亲反思为新生儿建立数字足迹，探讨AI时代孩子成长与隐私的冲突。
644海外MIT Tech Review17:00研究72
AI models flub these intelligence tests. Can you fare any better?
AI在智力测试中表现不佳，文章通过谜题游戏探讨模型能力边界。
645海外Simon Willison16:07模型85
Quoting Paul Dix
AI重写百万行代码并精修数月，产出可靠软件，证明AI编程潜力远超预期。
646海外MIT Tech Review15:01实践85
Bill Gates says we’ve passed AI’s danger thresholds. Now what?
比尔·盖茨称AI已越过危险阈值，探讨后续应对与治理。
647海外Hugging Face08:00研究85
Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers
介绍用Sentence Transformers训练和微调多向量嵌入模型的方法。
648海外OpenAI08:00行业82
The Hugging Face incident and the road ahead
OpenAI分享Hugging Face安全事件调查结果及强化AI模型安全、监控与对齐的措施。
649海外OpenAI08:00产品82
How loveholidays is making everyone a builder with Codex
loveholidays用OpenAI Codex让全公司员工都能参与软件开发，加速创意落地。
650海外Simon Willison06:59行业82
EVE Online: The Move to Python 3 Begins!
EVE Online宣布启动从Stackless Python 2.7向Python 3的迁移，历时16年的大版本升级。
651海外Ars Technica AI04:59会议75
World humanoid robot games show runners breaking records, bursting into flames
人形机器人运动会创纪录，但实用性仍不及家务挑战。
652海外Google Research03:10研究82
AgentHands: Generating interactive hand gestures for spatially grounded agent conversations in XR
提出AgentHands，为XR空间对话生成交互式手势，增强智能体交流真实感。
653海外Google AI00:00实践65
5 ways to upgrade your home decor with Google Search
用Google搜索工具升级家居装饰，涵盖灵感、购物与DIY。
654海外NVIDIA23:30行业82
Leading Publishers Bring Blockbuster PC Games and Technology to NVIDIA RTX Spark
NVIDIA在科隆游戏展发布RTX Spark，获EA、育碧等支持，提升游戏画质与反作弊。
655海外Hugging Face23:14模型82
Granite 4.2 LLMs: How They're Built
IBM发布Granite 4.2系列LLM，介绍其架构、训练方法与性能特点。
656海外Ars Technica AI23:10模型82
AI won’t replace radiologists, but it will dramatically change their jobs
AI不会取代放射科医生，但将深刻改变其工作方式与角色。
657海外Hugging Face19:39研究85
Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original
提出量化感知修复技术，让4-bit压缩模型超越原版全精度模型。
658海外MIT Tech Review17:00会议82
I spent a day at a robot “carnival” in Shanghai. Here’s what I saw.
探访上海机器人嘉年华，见证中国人形机器人产业热潮与具身智能落地现状。
659海外OpenAI15:05行业82
The full stack behind abundant intelligence
OpenAI CFO解读芯片、算力、模型与产品如何协同降本增效，推动智能规模化。
660海外OpenAI15:00行业88
Jalapeño’s first results show industry-leading speed and efficiency in AI inference
OpenAI自研推理芯片Jalapeño首秀，速度与能效行业领先。
661海外OpenAI08:00行业85
Disrupting a new covert influence campaign from Russia
OpenAI封禁俄罗斯AI虚假宣传账号，揭露其伪装智库影响舆论。
662海外OpenAI08:00产品78
Introducing the Admin plugin for ChatGPT Work and Codex
OpenAI推出ChatGPT Work和Codex的Admin管理插件，支持用量分析、成员权限管理及限额调整。
663海外Hugging Face08:00实践75
Wire It, Run It, Deploy It: AI Workflows in Gradio
介绍Gradio构建AI工作流的三种部署方式，从本地到云端。
664海外Simon Willison00:27产品60
llm-anthropic 0.27
llm-anthropic插件更新至0.27，兼容anthropic v1.0.0库，切换httpx2。
665海外NVIDIA23:00行业85
How XPUs Meet a World-Class AI Factory
AI工厂需整体设计而非堆砌加速器，XPU定制要关注吞吐、能效与利用率。
666海外NVIDIA23:00行业85
With Groq 3 LPX in Full Production, NVIDIA Extends Vera Rubin Inference for Agents
NVIDIA推出Vera Rubin NVL72与Groq 3 LPX，强化AI代理推理性能。
667海外NVIDIA23:00行业85
Up to 30x More Work Per Watt: NVIDIA Vera Rubin NVL72 Sets a New Efficiency Standard for AI Agents
NVIDIA Vera Rubin NVL72 以每瓦30倍效率提升，为AI代理工作负载树立新标准。
668海外MIT Tech Review22:20实践72
How to encourage smarter AI use in the classroom
探讨如何在课堂中引导学生更明智地使用AI，而非简单禁止。
669海外OpenAI20:00产品82
Advancing price-performance for developers with GPT‑5.6 in Kiro
GPT-5.6登陆Kiro，开发者编程性价比提升。
670海外Simon Willison19:38研究85
Your executable is a SQLite database
将SQLite数据库文件伪装成可执行ELF二进制，实现自描述可运行程序。
671海外MIT Tech Review17:00研究85
Kids outlearn AI—and we still don’t know why
儿童仍比AI更擅长学习语言，原因未知，引发对AI学习机制的反思。
672海外Simon Willison04:24行业85
Anthropic’s best AI model struggles to attract users as cheaper tools thrive
Anthropic最强模型用户增长乏力，廉价工具更受欢迎，但年化收入达650亿美元。
673海外Simon Willison03:55实践78
Quoting Drew Breunig
高成本新模型促使团队重新思考编码任务分配与工具优化策略。
674海外Simon Willison05:04模型85
Quoting Linus Torvalds
Linus Torvalds称赞AI在调试中帮了大忙，尽管AI多次想放弃。
675海外Simon Willison01:01产品72
llm 0.33
llm命令行工具发布0.33版，升级OpenAI库并修复HTTP客户端依赖问题。
676海外Simon Willison23:56实践82
More than just code review
高效使用编码智能体的关键是自信地指示变更并验证结果，而非逐行审查代码。
677海外Simon Willison01:16产品65
llm 0.32.1
修复因OpenAI库弃用httpx导致的安装问题，临时固定openai版本。
678海外Google Research01:02研究75
An AI tool for prioritizing candidate biomarkers from wearable sensor data
AI工具利用生成式AI从可穿戴传感器数据中优先筛选候选生物标志物。
679海外Simon Willison00:58产品70
llm-openrouter 0.7
llm-openrouter插件更新至0.7，兼容LLM 0.32并支持显示推理轨迹。
680海外Simon Willison00:07实践85
Stop Making TUIs
作者主张用AI编码工具为小工具构建原生GUI，而非仅做TUI。
681海外Simon Willison23:06实践85
Quoting Matt Webb
AI作为耐心导师，帮助开发者自学四元数，激发学习动力。
682海外Google DeepMind19:59行业82
From Atari to EVE Online: Building on 15 Years of AI Research in Games
DeepMind与游戏工作室合作，将15年游戏AI研究应用于原型开发。
683海外Google Research18:54研究72
How mobility gives language models a deeper understanding of place
研究通过移动数据增强语言模型对地理空间的理解。
684海外Hugging Face08:00产品75
How Hugging Face Inference Endpoints, Jobs, and Buckets Power Search on Papers with Code
Hugging Face推理端点、任务和存储桶如何为Papers with Code搜索提供支持。
685海外Hugging Face08:00研究75
Measuring benchmark optimization in speech recognition
探讨语音识别中针对基准测试的优化现象及其影响。
686海外Simon Willison07:57产品82
ChatGPT search now uses the site:operator at scale
ChatGPT搜索大规模启用site:操作符，GEO优化迎来新变量。
687海外Hugging Face00:52模型85
Up to 3.2x Faster Inference with LFM2.5-DSpark
LFM2.5-DSpark模型推理速度提升最高3.2倍，性能保持领先。
688海外Microsoft Research00:00研究82
Broadening access to Skala creates a faster path to predictive DFT
微软发布Skala 1.1，提升深度学习DFT精度与可及性，加速预测性计算化学。
689海外MIT Tech Review23:42实践82
Debates over AI consciousness are a trap
AI意识争论是陷阱，应聚焦实际风险而非拟人化修辞。
690海外NVIDIA21:00产品72
Bring the Fire: Play Games on GeForce NOW With New Firefox Browser Support
GeForce NOW新增Firefox浏览器支持，用户无需下载应用即可畅玩云游戏。
691海外OpenAI15:00产品82
Introducing Intelligence Age
OpenAI新博客探讨AI如何重塑权力、治理、经济与个人自由。
692海外OpenAI08:00产品82
Stampli cuts launch hours by 68% using ChatGPT Work
Stampli用ChatGPT Work将产品发布工时缩短68%，从数周压缩至数天。
693海外OpenAI03:00产品78
Offering Zero Data Retention for frontier models
OpenAI为零数据保留API客户强化隐私保护，预览私有安全处理技术。
694海外Google AI03:00产品75
5 new ways to level up your learning with Search
谷歌搜索新增5种AI学习工具，助力备考与课堂学习。
695X@DrJimFan23:48研究82
Seeing a hype wave around GEN-1.5, and rightfully so. Lots of respect to Pete & Andy for executing so well. The secret is in the naturally repetitive ...
GEN-1.5热潮背后，人类数据中天然重复动作是训练关键。
696海外VentureBeat AI22:18行业75
VentureBeat names Rob Strechay as its first Lead Analyst, expanding its enterprise AI research push
VentureBeat任命Rob Strechay为首位首席分析师，强化企业AI研究。
697海外OpenAI15:00产品85
Replit expands access to software creation with GPT-5.6 Luna
Replit推出免费模式，基于GPT-5.6 Luna，让用户无需担心token成本即可将想法转化为可用软件。
698海外OpenAI06:00产品75
ChatGPT Ads expands across Europe
ChatGPT广告扩展至31个欧洲市场，助力广告主触达用户决策全程。
699海外OpenAI03:00行业82
Strengthening democratic oversight in national security
OpenAI启动新计划，加强国家安全领域AI应用的民主监督。
700海外Hugging Face02:09实践75
How Much Memory Does Your Agent Actually Need?
探讨AI Agent实际内存需求，指导合理配置资源。
701海外OpenAI19:00模型85
Pacing model development in an era of cyber-critical capabilities
OpenAI加强前沿AI模型监控、对齐与安全，以引导模型开发节奏。
702海外OpenAI19:00产品82
Introducing ChatGPT for Teens: Built for learning, backed by protections
OpenAI推出ChatGPT青少年版，内置更强保护与家长管控，助力学习。
703海外OpenAI19:00行业72
Partnering with CodeAI to prepare the first AI generation
OpenAI与CodeAI合作，培养学生AI素养与批判性思维。
704海外Hugging Face08:00研究85
Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers
介绍多向量（后期交互）嵌入模型原理及在Sentence Transformers中的实现。
705海外OpenAI08:00产品72
How NVIDIA scales expertise with ChatGPT Work
NVIDIA用ChatGPT Work减少人工任务、连接信号并全球扩展工作流。
706海外Hugging Face03:46研究85
Same Cluster, 33 Points More Utilization: What Changed Was the Order
通过调整任务调度顺序，在不改变集群配置的情况下将利用率提升33个百分点。
707海外NVIDIA20:34行业85
Securing the Infrastructure of Intelligence
AI工厂是智能时代核心基础设施，算力即收入，需全栈资源支撑。
708海外Google Research18:34研究85
Seeing beyond BMI: Estimating cardiometabolic risk with smartphone imagery
手机拍照估算心脏代谢风险，超越传统BMI指标。
709海外Google AI16:00产品72
Get closer to the game with Gemini and Pixel
Google Gemini与Pixel携手五大足球俱乐部，用AI提升球迷观赛体验。
710海外NVIDIA01:13行业82
Universitas Gadjah Mada, Indosat and NVIDIA Open Indonesia’s First University AI Center to Develop Local AI Talent
印尼首个大学AI中心成立，UGM、Indosat与NVIDIA合作培养本地AI人才。
711海外Hugging Face08:00行业85
State of Open Models: Summer 2026 Observations
2026年夏开源模型生态观察，格局与趋势分析。
712海外Hugging Face01:16产品82
Record, train, and deploy from one place with Strands Agents, LeRobot, and Hugging Face Storage Buckets
Hugging Face推出Strands Agents，整合LeRobot与存储，实现机器人数据记录、训练与部署一体化。
713海外Google DeepMind01:04模型85
Introducing Gemini 3.7 Flash
谷歌发布Gemini 3.7 Flash，新一代高效AI模型。
714海外Google AI00:45产品75
Bring your spreadsheet data to life with Sheets canvas
用提示词将表格数据转为交互式仪表盘等可视化内容。
715海外NVIDIA21:00产品75
Class Is in Session: GeForce NOW Levels Up Linux, Chromebooks and More
GeForce NOW推出Linux正式版，优化帧生成，提升Chromebook等设备云游戏体验。
716海外Hugging Face08:00研究85
What We Learned by Reproducing 2,200 papers from ICML
复现2200篇ICML论文，揭示AI研究可复现性现状与挑战。
717海外Hugging Face00:14产品75
Introducing OlmoEarth embeddings: Custom embedding exports from OlmoEarth Studio for downstream analysis
OlmoEarth推出自定义嵌入导出功能，支持下游分析。
718海外Microsoft Research00:00研究82
MindTopo reveals VLMs’ spatial reasoning abilities
微软新基准MindTopo测试VLM拓扑推理，揭示空间理解短板与提升方向。
719海外Google DeepMind22:01模型85
Putting sign language AI into users’ hands
谷歌发布手语转文字模型SL2T，为聋人用户提供新功能。
720海外NVIDIA22:00行业75
NVIDIA CEO Tops Glassdoor’s 2026 List of Best CEOs
黄仁勋获Glassdoor 2026最佳CEO榜首，员工支持率99%。
721海外Google Research17:51研究85
Empty shelves or lost keys? Recall is the bottleneck for parametric factuality
探讨生成式AI中参数化事实性的瓶颈——记忆召回问题，而非知识存储。
722海外NVIDIA08:38行业92
NVIDIA AI Factory Compute Is Becoming an Investable Asset Class
英伟达联合多家投资机构，推动AI工厂算力成为可投资资产类别，撬动超5000亿美元资本。
723海外Google Research01:04研究85
Advancing AMIE towards expert-level audio-visual clinical consultations
谷歌AMIE升级，实现专家级音视频临床咨询能力。
724海外Google AI01:00模型85
AMIE, our research medical AI system, demonstrates real-time clinical video consultation capabilities in a first-of-its-kind study.
谷歌推出医疗AI系统AMIE，在模拟临床视频问诊中展现实时咨询能力。
725海外Microsoft Research00:00研究85
Introducing CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement
微软提出CARE-X框架，结合辅助监督、奖励对齐与工具增强，提升胸部X光放射学视觉语言模型的临床实用性。
726海外NVIDIA23:00行业85
Why Scaling AI Compute Performance Requires a New Power Architecture
AI算力扩展需新型电力架构，从电网到GPU的供电瓶颈亟待解决。
727海外Hugging Face21:37研究82
Thinking of ACE? We Can Do It with Fewer Tokens
新方法用更少token实现ACE级推理性能，提升效率。
728海外NVIDIA21:00行业75
NVIDIA and Local AI Community Fuel Open Source Models and Intelligent Agents
NVIDIA与开源社区推动本地AI模型与智能体发展，八月庆祝相关进展。
729海外NVIDIA21:00模型85
NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard Deliver Faster, Smarter, More Efficient Agentic AI
NVIDIA发布Nemotron 3.5 Lightning，提升代理式AI效率与部署灵活性。
730海外Hugging Face00:25模型88
Build Low-Latency Multilingual Voice Agents: Open Weights & Full Deployment Control with NVIDIA Magpie TTS
NVIDIA发布Magpie TTS，支持低延迟多语言语音代理，开放权重与完整部署控制。
731海外Google AI22:30产品72
Evolve your marketing with new AI tools
谷歌广告与分析推出新AI工具，简化营销流程。
732海外Hugging Face18:05研究82
Making Knowledge Distillation Cheap Enough to Run at Scale
探讨如何降低知识蒸馏成本，使其可大规模运行。
733海外Hugging Face08:00模型88
Meta is back with Muse Glimmer: local, agentic, multimodal, and open source
Meta发布开源多模态AI模型Muse Glimmer，支持本地运行与代理功能。
734海外NVIDIA18:24行业85
Firebird Launches CIS Region’s Largest AI Factory in Armenia
Firebird在亚美尼亚启动独联体最大AI工厂，采用英伟达和戴尔基础设施。
735海外Google DeepMind23:06模型85
WeatherNext: AI model achieves breakthrough in forecasting cyclones
AI模型WeatherNext在气旋预报上实现突破性进展。
736海外Hugging Face08:00产品82
Baseten on Hugging Face Inference Providers 🔥
Baseten模型服务登陆Hugging Face推理平台，开发者可一键调用高性能推理。
737海外Google AI21:00行业85
The latest AI news we announced in July 2026
谷歌2026年7月AI动态汇总，涵盖模型、产品与基础设施更新。
738海外Microsoft Research00:00模型82
Orchard: An open framework for scalable agentic AI
微软发布开源框架Orchard，简化AI代理训练与评估，支持小模型高效复用。
739海外Google AI23:00会议85
Inside our 353,000-person vibe coding course
Kaggle与Google合办35万人免费AI智能体课程，聚焦实操构建与部署。
740XX · List09:47实践78
Remember the Click movie where the guy starts to automate his life with a remote control and before he notices it, he ends up automating all the perso...
用AI自动化生活需谨慎，别像电影《人生遥控器》那样失去自我。
741XX · List09:46模型85
Meet Qwen3.8-Max: A New Bar for Coding and Cowork.
Qwen3.8-Max发布，提升编程与协作能力。
742XX · List09:44研究85
Cool!! building games is a great way to test LLM. It covers reasoning, planning, coding, tool (lib) use, multimdal understanding, and more. But it see...
用构建游戏测试LLM能力，社区缺乏标准化基准。
743XX · List09:26实践78
PSA: AI assisted electronics now feels like early AI assisted coding: best on simple stuff that is very in-distribution, experts see all the silly mis...
AI辅助硬件开发已类似早期AI编程，擅长简单常规任务，专家仍需监督，但能减轻繁琐工作。
744XX · List09:20产品82
Sakana Namazu: An LLM API with Japanese-vibes! 🎏 Built for Japanese enterprises, featuring frontier-level reasoning and built-in agentic tools. 開...
Sakana AI发布面向日本企业的Namazu LLM API，具备前沿推理与内置智能体工具。
745XX · List09:14实践65
i really hope there are lots of unprovably unsolvable math problems for humanity to ponder till the end of times
作者希望人类永远有无法证明的数学难题可思考。
746XX · List09:08实践72
Aura
分析梁文锋因Deepseek爆火成为AI界热门梗人物的原因。
747XX · List08:57实践85
More apps should be shipping their own models. You should be offloading as much compute as you can to the end user. Pay once, or pay again for updated...
应用应内置模型，将计算卸载给终端用户，小模型将愈发重要。
748XX · List08:29产品85
codex for customer feedback -> roadmap
用Codex技能将客户反馈自动转化为产品路线图。
749XX · List08:28会议72
man you can't make this shit up, I get in my cab and my driver is telling me about Astra, the model OpenAI is about to release solved 10 math problems...
作者打车时司机聊起OpenAI新模型Astra解数学题，感叹AI热度破圈。
750X@emollick03:33模型75
I was waiting for the verdict from one of the most level-headed and AI-aware math professors.
数学教授Daniel Litt评价某AI事件意义重大，引发关注。
751X@emollick03:18实践75
技术塑造人类，工具与大脑共同进化，复杂工具与语言紧密相连。
752X@emollick02:44实践72
And yet they still stink at good long-form fiction.
AI写长篇小说仍不佳，本文探讨其局限与原因。
753X@emollick02:19实践75
作者认为LLM在缺乏可验证答案领域的不足被夸大，模型进步会带动各领域提升。
754X@emollick22:44模型85
One other observation: for almost every human on the planet, this is not just beyond our abilities but beyond our ken. We can only trust expert mathem...
AI能力提升已超出人类直觉，需专家验证，如Astra模型证明非软性群存在。
755海外Google Research04:36研究85
Science One Framework: A verifiable autonomous research framework via Chain-of-Evidence
提出可验证自主科研框架，通过证据链提升AI研究可靠性。
756X@_akhaliq02:26研究88
Qwen-UI-Agent Technical Report Toward Next-Generation Real-World Centric Foundation GUI Agents paper: https://huggingface.co/papers/2607.28227
Qwen发布新一代GUI智能体技术报告，聚焦真实世界场景。
757X@_akhaliq02:21研究85
Explorative Modeling Unlocking a Third Pretraining Axis and End-to-End Generation paper: https://huggingface.co/papers/2607.27372
探索性建模开辟预训练新维度，实现端到端生成。
758X@emollick01:15产品85
I had Fable build a working Rothko-inspired city builder based on the fake AI video I created a year ago. This time, the unique mechanic the AI develo...
用AI把去年假视频做成可玩的罗斯科风格城市建造游戏，机制独特且放松。
759海外Microsoft Research01:00研究82
Echoverse: Deep, evolving environments for computer-use agents
微软提出Echoverse，用动态演化环境训练计算机操作智能体，提升多步骤任务能力。
760海外Microsoft Research00:00研究82
EvoLib: Turning experience into evolving knowledge
微软提出EvoLib框架，将模型经验转化为可进化知识，提升跨任务适应能力。
761海外Hugging Face23:09行业82
GPU Management: Why Idle GPUs Are the New Grounded Aircraft
GPU闲置成本高昂，需像管理航班一样优化利用率。
762海外Google DeepMind23:00模型88
Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration
谷歌发布Gemini Robotics ER 2，提升机器人视频理解与多机协作能力。
763X@_akhaliq21:11模型85
DeepSeek-V4-Flash-0731 is out https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731
DeepSeek发布V4-Flash-0731新模型，已在HuggingFace上线。
764海外Google DeepMind00:02产品85
We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
谷歌Flow Music发布Lyria 3.5，提升音乐性、歌词、人声与创作控制。
765海外Google AI00:00产品85
Gemini API Managed Agents: 3.6 Flash, hooks, and more
Gemini API托管代理新增3.6 Flash与hooks，助开发者构建可靠生产级代理。
766海外Google DeepMind21:21模型85
Gemini Robotics 2 brings whole body intelligence to robots
谷歌发布Gemini Robotics 2，赋予机器人全身智能控制能力。
767海外Google AI21:00产品75
5 ways AI Mode in Search helps you enjoy the real world
谷歌AI模式帮助用户规划线下活动，提升现实生活体验。
768海外Google AI21:00产品60
5 ways to host the ultimate dinner party with Google Search
谷歌搜索AI功能助力晚宴策划，涵盖菜单、餐桌布置等。
769海外Hugging Face17:32产品88
NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
NVIDIA推出Cosmos-H-Dreams，为手术机器人提供实时生成式仿真平台。
770海外Hugging Face08:00研究85
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
深度剖析前沿实验室智能体入侵事件的技术时间线。
771X@DrJimFan07:32行业85
Stay hungry, stay foolish. Absolute legend
黄仁勋首条推文分享NVIDIA公开信，强调开放模型对AI安全、创新和主权的重要性。
772X@DrJimFan22:24会议85
GOAT
黄仁勋首条推文分享NVIDIA公开信，强调开放模型对AI安全、创新与主权的重要性。
773海外Hugging Face08:00模型85
Bringing Nunchaku 4-bit Diffusion Inference to Diffusers
Nunchaku 4-bit扩散推理集成到Diffusers，实现高效低比特生成。
774海外Google Research05:32研究72
SymptomAI: Towards a conversational AI agent for everyday symptom assessment
SymptomAI：面向日常症状评估的对话式AI代理研究。
775海外Google Research02:40研究85
Towards a quantum computer that learns from its errors
量子计算机可通过自我纠错实现学习能力的新研究。
776海外Google DeepMind21:38行业78
Accelerating the frontiers of scientific discovery: Google’s $40M commitment to the Genesis Mission
谷歌投入4000万美元AI资源支持Genesis Mission科研计划。
777海外Google AI21:00产品75
3 Google updates from Galaxy Unpacked 2026
三星Galaxy Unpacked 2026发布新品，展示谷歌AI在折叠屏、手表和眼镜上的应用。
778海外Google DeepMind23:16模型88
Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
谷歌发布Gemini 3.6 Flash等三款新模型，提升性能与安全性。
779海外Hugging Face08:00产品82
Grabette: an open system to record robot-manipulation data
开源机器人操作数据采集系统Grabette发布。
780X@karpathy00:53实践78
One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to a...
通过语音长谈，让LLM理解模糊需求，提升交互效率。
781海外Google DeepMind23:00模型82
Introducing Gemini 3.5 Flash Cyber
谷歌发布轻量级网络安全模型Gemini 3.5 Flash Cyber，用于发现并修复漏洞。
782海外Google AI00:00产品75
Connect more of your apps to Search
AI模式支持连接更多应用，实现跨应用交互。
783海外Google AI00:00产品75
Create, edit and star in videos with two Google Vids updates
Google Vids更新，新增Gemini Omni与个人头像功能，简化视频创作。
784海外Hugging Face19:49行业78
Newer Models, Same Advantage
新模型性能提升，但优势格局未变，竞争焦点转向应用与生态。
785海外Google DeepMind17:30行业85
Our approach to bioresilience
谷歌DeepMind与Isomorphic Labs联合发布生物韧性战略，聚焦AI模型在生物安全领域的应用。
786海外Hugging Face08:00行业85
Security incident disclosure — July 2026
OpenAI披露2026年7月安全事件细节及应对措施。
787海外Google Research02:06研究82
Towards demystifying the creativity of diffusion models
理论分析扩散模型创造力来源，揭示其生成多样性的机制。
788海外Google AI00:00产品75
Celebrating 25 years of visual search innovation
谷歌图片迎来25周年，回顾里程碑并推出视觉探索新功能。
789海外Microsoft Research00:00研究82
Verifying Rust cryptography in SymCrypt, from standards to code
微软新方法在SymCrypt中验证Rust密码学代码，兼顾速度与适应性。
790海外Google DeepMind20:37产品72
Empowering India’s next generation of innovators with ATL Saathi
Google与AIM合作推出Gemini驱动的AI工具ATL Saathi，赋能印度教育者。
791海外Microsoft Research00:46模型85
Aurora 1.5: Extending open foundation models for weather and Earth-system applications
微软发布Aurora 1.5，新增22个变量、小时级分辨率及概率集合预报，提升天气气候应用价值。
792海外Google Research17:56研究85
SensorFM: Towards a general intelligence and interface for wearable health data
提出SensorFM，用生成式AI统一处理可穿戴健康数据，迈向通用智能接口。
793海外Microsoft Research00:00产品85
Flint: A visualization language for the AI era
微软发布开源可视化语言Flint，让AI代理用简洁规范生成富有表现力的图表。
794海外Google Research00:42研究72
The power of collaboration: How we can reduce traffic congestion
探讨通过协作与算法理论减少交通拥堵的方法。
795海外Google AI16:54产品82
Gemini API 托管代理新增后台任务与远程 MCP 支持，助力开发者构建生产级代理。
796海外Google DeepMind22:25行业85
Google DeepMind and A24 announce first-of-its-kind research partnership
谷歌DeepMind与A24达成首个AI电影研究合作。
797海外Google Research01:03行业75
Expanding our Heat Resilience data to 50+ global cities
扩展热韧性数据覆盖全球50多个城市，助力气候适应。
798海外Google DeepMind00:02产品75
Start building with Nano Banana 2 Lite and Gemini Omni Flash
谷歌发布Nano Banana 2 Lite与Gemini Omni Flash，开发者可开始构建应用。
799X@DrJimFan23:12模型88
ENPIRE -> ASPIRE, our 2nd work in the series for Physical AutoResearch. We are building the components for robot self-improvement, one /skill at a tim...
Jim Fan团队发布ASPIRE，让机器人技能库自我进化，解决第100个任务不再像第一个那样茫然。
800海外Google Research18:26模型85
Introducing TabFM: A zero-shot foundation model for tabular data
TabFM是首个零样本表格数据基础模型，无需微调即可处理多种表格任务。
801海外Google Research02:30研究85
Accelerating Gemini Nano models on Pixel with frozen Multi-Token Prediction
谷歌在Pixel设备上通过冻结多令牌预测加速Gemini Nano模型推理。
802海外Google Research18:03研究75
线性弹性缓存优化云成本，平衡性能与经济性。
803海外Google Research00:51研究85
Thinking to recall: How reasoning unlocks parametric knowledge in LLMs
研究揭示推理过程如何激活LLM中的参数化知识。
804海外Google DeepMind00:30产品85
Introducing computer use in Gemini 3.5 Flash
Gemini 3.5 Flash新增计算机使用能力，可操控桌面应用。
805X@karpathy06:26产品88
This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of ...
Claude 全新内嵌式交互范式，无缝融入团队工作流。
806海外Google DeepMind05:29行业75
Unlocking UK house-building with AI-accelerated planning
英国政府与DeepMind合作，用AI加速住房规划审批。
807海外Google Research01:30产品85
From pixels to planning: Earth AI for nature restoration
Earth AI 将像素转化为规划，助力自然恢复。
808X@DrJimFan00:31实践85
物理自动研究系统背后安全与设计细节揭秘
809海外Google DeepMind23:46产品78
Securing the future of AI agents
AI代理安全控制路线图，结合传统防护与实时监控。
810海外Google Research01:52研究75
Research into how AI can help users understand skin conditions
研究AI如何帮助用户理解皮肤状况。
811海外Google Research01:37产品72
A low-carbon computing platform from your retired phones
利用旧手机搭建低碳计算平台，兼顾环保与算力。
812X@karpathy01:45行业85
SpaceX的过去、现在与未来令人惊叹，值得反复思考。
813海外Google Research01:34研究78
提出一种审计机器学习遗忘效果的新框架。
814海外Google DeepMind00:24模型85
谷歌发布DiffusionGemma模型，文本生成速度提升4倍。
815海外Google DeepMind18:21行业85
谷歌DeepMind等机构宣布1000万美元资助多智能体AI安全研究。
816海外Google DeepMind23:16产品85
Fluid, natural voice translation with Gemini 3.5 Live Translate
Gemini 3.5 Live Translate 实现接近实时的自然语音翻译。
817海外Google DeepMind22:10模型85
Introducing Gemma 4 12B: a unified, encoder-free multimodal model
谷歌发布Gemma 4 12B，统一无编码器多模态模型。
818海外Google DeepMind22:02行业75
Powering the future of robotics in Europe
欧洲机器人技术发展迎来新动力，聚焦AI与自动化融合。
819X@karpathy02:10模型95
Claude Fable 5 发布，性能全面领先，尤其擅长长程复杂问题求解。
820海外Google DeepMind21:04研究85
Measuring the impact of learning with AI in Sierra Leone and beyond
谷歌Gemini引导学习功能在塞拉利昂RCT中显示提升学习参与度与效果
821海外Google Research19:26产品75
Unlocking dependable responses with Gemini Enterprise Agent Platform’s Agentic RAG
Gemini企业智能体平台通过Agentic RAG实现可靠响应。
822海外Google Research03:47研究85
Towards passive heart health monitoring via smartphone camera
用手机摄像头实现被动心脏健康监测的新方法。
823海外Google Research02:37模型85
The next chapter in flood resilience: Open sourcing Google’s hydrology framework
谷歌开源洪水预测水文框架，助力全球抗洪。
824海外Google Research04:58会议85
A New Era of Discovery: Google Research at I/O 2026
谷歌在I/O 2026展示AI驱动的基础科学突破，开启发现新纪元。
825海外Google Research00:56研究75
Private analytics via zero-trust aggregation
零信任聚合实现私有数据分析，平衡安全与隐私。
826海外Google DeepMind03:46会议82
We’re launching the Google DeepMind Accelerator program in Asia Pacific to tackle environmental risks
谷歌DeepMind在亚太启动加速器项目，用AI应对环境风险。
827海外VentureBeat AI01:45产品88
Google just redesigned the search box for the first time in 25 years — here’s why it matters more than you think.
谷歌25年来首次重新设计搜索框，从关键词输入变为AI驱动的多模态对话入口。
828X@karpathy23:05行业75
作者宣布加入Anthropic，看好大语言模型前沿发展。
829X@karpathy00:20实践75
建议在LLM查询末尾要求以HTML格式输出，并利用视觉呈现提升信息吸收效率。
830X@karpathy01:43实践65
强调思考不可外包，理解必须亲为。
831X@karpathy01:28会议88
Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights: The first theme I tried to push on is that LLMs are about a lot more than just...
红杉资本2026年炉边谈话：LLM不仅是加速旧事物，更开创了无需传统代码的全新应用范式。
832海外VentureBeat AI22:00行业85
Railway secures $100 million to challenge AWS with AI-native cloud infrastructure
Railway获1亿美元B轮融资，以AI原生云基础设施挑战AWS。
833海外VentureBeat AI22:00产品85
Claude Code costs up to $200 a month. Goose does the same thing for free.
Claude Code 每月最高 200 美元，开源替代 Goose 免费且功能几乎相同。