1海外AWS ML2 天前产品78
Accelerating M&A due diligence with Amazon Bedrock AgentCore
用Amazon Bedrock AgentCore构建多智能体并购尽职调查系统，含参考架构和可运行示例。
2海外AWS ML2 天前产品85
Amazon Quick for Microsoft 365: Agentic AI where you work
Amazon Quick AI助手现已集成至Microsoft 365办公套件，支持文档编辑与数据分析。
3海外TechCrunch AI2 天前产品78
Microsoft kills off unsuccessful AI features while merging its separate Copilot apps
微软合并个人与企业版Copilot应用，砍掉AI播客、群聊等未成功功能。
4海外TechCrunch AI2 天前行业85
Nvidia’s new $500B plan is risky but brilliant, especially for aging GPUs
英伟达推5000亿美元计划，为老GPU保值并吸引新融资，风险与高明并存。
5海外The Decoder2 天前模型85
Ling 3.0 Flash is the smartest open model at its size
Ling 3.0 Flash 成为同尺寸最强开源模型。
6海外TechCrunch AI2 天前行业85
Apple in talks to pay publishers to provide Siri with current news: report
苹果拟斥资九位数美元，向出版商付费以获取新闻内容训练Siri。
7海外Ars Technica AI2 天前行业85
Anthropic could be worth $2 trillion when it goes public
Anthropic营收猛增，IPO估值或达2万亿美元，成史上最大上市。
8海外MIT Tech Review2 天前行业78
Flock is tightening its rules in response to a growing surveillance backlash
Flock因监控争议收紧警用车牌读取器访问规则，以平息反弹并赢回合同。
9海外NVIDIA2 天前产品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，提升性能会员帧率。
10海外Ars Technica AI3 天前产品82
Claude's new Scarlet Letter watermark is invisible—for now
Claude新水印技术可标记AI处理内容，目前肉眼不可见。
11海外The Decoder3 天前行业78
Fable 5's slow adoption suggests corporate willingness to pay for frontier AI has hit a ceiling
Fable 5虽强但企业购买少，企业AI支出或触顶。
12海外The Decoder3 天前研究82
Top AI lab researchers warned about automated AI research, and several of their predicted milestones have already fallen
AI实验室研究员曾警告自动化AI研究，其预测的多个里程碑已实现。
13海外The Decoder3 天前产品82
Anthropic brings Claude Cowork to its Chrome extension, adding skills and plugins to the browser
Anthropic将Claude Cowork集成至Chrome扩展侧边栏，新增技能与插件支持。
14海外MIT Tech Review3 天前实践82
How kids feel about AI, in their own words
孩子们亲口讲述对AI的真实感受与使用方式，颠覆成人预设。
15海外MarkTechPost3 天前模型85
Dyna Robotics Introduces Dyna-2: A World-Action Model Pre-Trained on 1 Million Hours of Human Video
Dyna Robotics发布Dyna-2世界动作模型，基于百万小时人类视频预训练，实现跨实体泛化。
16海外MarkTechPost3 天前模型85
SpaceXAI Releases Grok 4.6: A 500K-Context Frontier Model Tuned for Long-Running Agents, Coding, and Knowledge Work
SpaceXAI发布Grok 4.6，升级后处理能力达50万上下文，性能对标GPT-5.6，定价不变。
17海外Simon Willison3 天前模型75
DeepSeek V4 Pro 0813 (on OpenRouter)
DeepSeek V4 Pro 0813模型已上线OpenRouter，仅提供API，权重开源情况未定。
18海外TechCrunch AI3 天前产品75
Some Claude users are mad that Anthropic’s new watermarks will catch them using it at their jobs, classes
Anthropic新水印引发用户不满，担心工作学习中被发现使用AI。
19海外Ars Technica AI3 天前产品82
The web’s newest weapon against AI scrapers is a font
新型字体工具可干扰AI抓取训练数据，同时保持人类可读。
20海外Ars Technica AI3 天前行业85
Terabytes of credentials leaked in massive supply-chain attack
AI包供应链攻击致2500用户凭据泄露，数据达TB级。
21海外Ars Technica AI3 天前行业78
Twitch content has trained Amazon AI for years, but users can opt out now
Twitch用户内容多年用于训练亚马逊AI，现可退出。
22海外TechCrunch AI3 天前行业82
Amazon will train on Twitch streamers’ content by default, unless they opt out
Twitch默认用主播内容训练AI，需手动退出，引发争议。
23海外Simon Willison3 天前产品75
alchemy-utils 0.1a0
作者用AI助手快速构建了数据库无关的sqlite-utils原型库。
24海外The Decoder3 天前模型88
SpaceXAI's Grok 4.6 matches OpenAI's best model and undercuts it on price
Grok 4.6性能追平GPT-5.6，价格低六成，代理任务效率翻倍。
25海外TechCrunch AI3 天前行业85
AI coding startup Cognition reportedly already in talks to raise at $40B valuation
AI编程公司Cognition据悉正洽谈以400亿美元估值融资，距上次260亿美元融资仅数月。
26海外TechCrunch AI3 天前会议82
As AI safety concerns mount, three pioneers make the case for staying open
三位AI先驱在Ai4大会呼吁保持开放，辩论监管与开源。
27海外AWS ML3 天前实践72
Part 2: Amazon Bedrock cost attribution with Amazon Athena and CUDOS
用Athena和CUDOS仪表盘可视化分析Bedrock成本，按主体、项目、团队追踪AI支出。
28海外TechCrunch AI3 天前行业85
OpenAI-backed Thrive Holdings raises $2B to bring AI to the enterprise
OpenAI支持的Thrive Holdings融资20亿美元，估值达120亿，加速企业AI落地。
29海外MarkTechPost3 天前实践82
AllenAI Open Instruct Tulu 3 Post-Training with SFT, DPO, RLVR, GRPO, and Verifier-Based Evaluation
介绍AllenAI开源Tulu 3后训练框架，涵盖SFT、DPO、RLVR等，可在16GB硬件高效运行。
30海外The Decoder3 天前研究85
Researchers can now reverse-engineer LLM prompts from output text with near-perfect accuracy
新方法可近乎完美地从LLM输出反推原始提示词，构成潜在安全风险。
31海外The Verge AI3 天前产品78
Twitch streamers can now opt out from training Amazon’s AI
Twitch主播现可选择退出，防止内容被用于训练亚马逊生成式AI模型。
32海外TechCrunch AI3 天前产品75
Mesh, Automattic’s CRM for everyone, comes to Android
Automattic的AI联系人管理应用Mesh登陆Android平台。
33海外MIT Tech Review3 天前行业85
Scaling AI agents with trustworthy data
AI代理规模化应用关键在于可信数据基础设施，而非单纯技术。
34海外TechCrunch AI3 天前产品82
Why Stream ring-maker Sandbar says the future of AI wearables is voice
AI可穿戴设备未来在于语音交互，戒指形态成新趋势。
35海外The Decoder3 天前行业72
AI tools for breast cancer detection fall short of radiologists' expectations
调查显示AI乳腺癌检测工具未达放射科医生预期，实际效果与期望存在差距。
36海外Hugging Face3 天前产品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。
37海外TechCrunch AI3 天前行业85
Lovable confirms new $13.3B valuation, raises another $400M
Lovable估值达133亿美元，再融4亿美元，年化收入5亿美元。
38海外Microsoft Research3 天前研究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机会。
39XX · List2 天前实践75
VC行业普遍存在夸大背景和影响力现象，不应聚焦于特定演示。
40XX · List2 天前产品75
Own every note, knob and detail in our powerful browser-based DAW. 🎛️
浏览器DAW上线，掌控每个音符与细节
41XX · List2 天前行业65
Wild and sad!
特斯拉视角下的印度交通视频，展现混乱路况，引发感慨。
42XX · List2 天前研究85
A great read if you an AI dev. Current context compactors retain 17% of the standing rules users give them. Session Constraints are instructions like ...
研究发现AI上下文压缩器仅保留17%规则，常使任务效果更差。
43海外The Verge AI3 天前行业72
Guitar company D’Addario admits that AI music was used in a promotional video
D'Addario承认宣传片使用AI音乐，此前否认近两周。
44XX · List2 天前研究75
don't believe the ARC-AGI propaganda that their benchmarks show some capabilities that no other benchmark captures most math or long-context benchmark...
反驳ARC-AGI基准独特性，指出多数数学和长上下文基准同样体现思维模型转变。
45海外The Decoder3 天前行业78
Google's Gemini is losing market share to ChatGPT and Claude according to new market data
谷歌Gemini市场份额下滑，OpenAI超50%，Anthropic升至14.9%。
46XX · List2 天前研究72
«I think the next paper is likely on harnesses and agents, swarms, context-level continuity. maybe autoresearch. Hopefully lifetime learning/continue...
作者推测AI实验室下一步论文方向，可能涉及智能体、群体、上下文连续性及终身学习。
47海外TechCrunch AI3 天前行业75
How a $250 million acquisition collapsed into allegations of fraud and forged signatures
2500万美元收购因伪造签名和欺诈指控而破裂，投资者仍在等待资金。
48XX · List2 天前模型85
DeepSeek removed the weights of DeepSeek-V4-Pro-0813 again I guess the backlash got to them
DeepSeek再次下架V4-Pro-0813权重，或因舆论压力所致。
49XX · List2 天前产品88
Google DeepMind just shipped a model that lets people sign into a Pixel 11 instead of typing. SL2T starts with ASL-to-English in Gboard and Live Trans...
谷歌DeepMind发布SL2T模型，可将美式手语实时翻译为英语，支持Pixel 11免输入登录。
50XX · List2 天前会议65
AI生成道歉内容引发争议，网友热议。
51XX · List2 天前实践82
The biggest engineering mistake I made at Figure was building a tendon-based hand Our first hand design in 2022 was a tendon hand for our F.01 robot. ...
Figure创始人复盘腱绳驱动手部设计是重大工程失误，反思技术选型教训。
52海外Ars Technica AI3 天前行业78
Booksellers suspect AI firms are buying and then destroying rare books
书商怀疑AI公司批量购买并销毁稀有书籍，引发行业抵制。
53海外Simon Willison3 天前实践82
Quoting Florian Herrengt
AI虽提升编码效率，但代码理解断层导致疑难bug难解，引发对软件工程未来的思考。
54海外Simon Willison3 天前实践82
Quoting Florian Herrengt
AI虽提升编码效率，但开发者对代码理解不足，导致疑难bug难以修复。
55海外TechCrunch AI3 天前产品78
Why Sandbar thinks it’s voice-enabled ring can avoid the AI hardware graveyard
Sandbar推出语音戒指，主打AI笔记，避免重蹈AI硬件覆辙。
56海外TechCrunch AI3 天前产品88
Everything announced at Made by Google ’26: Pixel 11, Pixel Watch 5, Pixel Tag, and tons of Gemini features
谷歌2026发布会：Pixel 11、手表5、追踪器及大量Gemini功能。
57海外Hacker News3 天前行业75
German advocacy group lodges criminal complaint over Meta AI glasses
德国组织对Meta AI眼镜提起刑事诉讼，涉隐私问题。
58海外Hacker News3 天前行业85
Someone is running mass vulnerability scans, spoofing AI bots like ClaudeBot
有人伪装成AI爬虫进行大规模漏洞扫描，引发安全担忧。
59海外Google DeepMind3 天前模型85
Putting sign language AI into users’ hands
谷歌发布手语转文字模型SL2T，助力聋哑用户沟通。
60海外Hugging Face3 天前模型82
LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge
LFM2.5-VL-3B发布，为边缘设备提供更强更快的视觉能力。
61海外NVIDIA3 天前行业85
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%。
62海外The Verge AI3 天前产品72
Google’s Pixel Watch 5 dives deeper into AI and health
谷歌Pixel Watch 5主打AI健康功能，硬件小幅升级，售价涨至399美元。
63海外The Decoder3 天前行业75
Legal startup founder Robert Mahari joins Anthropic to lead Claude's push into law practices
Anthropic聘请法律科技创始人Robert Mahari，负责Claude在法律行业的应用拓展。
64海外AWS ML3 天前行业82
How OneAdvanced deployed over 50 AI agents on UK-sovereign AWS
OneAdvanced在AWS上自托管Llama模型，部署50多个AI代理，构建英国主权AI平台。
65海外AWS ML3 天前产品82
Pay with confidence: How Solv Labs built verifiable, auditable agent payments on Amazon Bedrock AgentCore payments
Solv Labs在Amazon Bedrock上构建可验证、可审计的代理支付流程，确保企业合规。
66海外AWS ML3 天前行业72
Tiered KV cache for large LLMs on Amazon SageMaker HyperPod with Curvine
用Curvine在SageMaker HyperPod上构建分层KV缓存，以低成本实例实现近本地磁盘速度的缓存复用。
67海外Hacker News3 天前实践92
AI is removing the middle class of software engineering?
AI正淘汰软件工程中层，加剧行业两极分化。
68XX · List2 天前模型85
SL2T is our amazing sign-language-to-text model allows users to sign directly to their phones for the first time. Built in close collaboration with th...
谷歌DeepMind发布手语转文字模型SL2T，支持美国手语转英语，帮助聋人用户。
69XX · List2 天前实践72
作者自嘲听自己的演讲做早餐，并推荐Lambda关于开源模型选型与部署的讨论。
70XX · List2 天前模型95
🚨 the weights are out https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813
DeepSeek-V4-Pro权重已开源，可下载使用。
71XX · List2 天前产品85
This is unironically the best ad for AI design ever made
谷歌设计团队用AI制作庆祝动画，展示AI设计潜力。
72XX · List2 天前模型92
new deepseek v4 pro is now open weight on hugging face (mit license) "v4" is a bit misleading, previous model was only a preview and this one has way ...
DeepSeek V4 Pro开源发布，MIT许可，训练量远超预览版，接近V4.5。
73XX · List2 天前产品85
Can’t wait for everyone to try Rambler — one of my favorite models! As someone who has wrist issues and often voice types, this feature is a game ch...
谷歌发布Pixel 11，主打Gemini智能与Rambler语音输入等新功能。
74XX · List2 天前行业72
质疑Corgi Invest宣称ETF数量将超贝莱德，困惑ETF数量与资产管理规模/收入的关系。
75海外The Decoder3 天前模型78
Nvidia's Nemotron 4 aims for one trillion parameters, a scale Chinese labs already surpassed
英伟达开发万亿参数开源模型Nemotron 4，对标顶尖开源模型，但中国实验室已超越该规模。
76XX · List2 天前产品78
Sakana Chat just got a big upgrade. No login required, free to use: https://chat.sakana.ai/ Powered by Fugu and Namazu, our Japanese LLM. With newly a...
Sakana Chat升级，免登录免费使用，支持代码执行，可快速生成交互应用。
77XX · List2 天前实践78
Always a good time when I can sit down and yap with Hamel. Not pictured: this 45 minute edit started as an hour half when we talked about everything f...
Hamel与Lambda探讨开源模型适用场景及部署经验，强调多数团队忽视前沿API之外的选项。
78XX · List2 天前模型72
DeepSeek has been very disappointing this year V4 came much later than expected and was smaller than expected. The price hikes also hurt. They are cur...
DeepSeek V4发布不及预期，价格上调，国内排名下滑至四五位。
79海外The Verge AI3 天前行业78
Of course the ChatGPT dog cancer vaccine spawned a startup
ChatGPT狗癌症疫苗故事催生新创企Gamgee，为犬提供个性化mRNA疫苗。
80海外The Verge AI3 天前产品85
Grok is now an AI ‘teammate’ you can assign work
Grok推出AI队友服务，可自主完成多步骤工作任务。
81海外The Decoder4 天前产品75
OpenAI launches ChatGPT desktop app for Linux
OpenAI发布ChatGPT Linux桌面应用，扩展平台覆盖。
82海外TechCrunch AI4 天前行业82
AI code-testing startup Blacksmith’s valuation jumps almost 10x in less than a year
AI代码测试初创Blacksmith估值一年涨近10倍，营收增超十倍。
83海外The Decoder4 天前模型78
Microsoft's new MAI Code 1.1 Flash gets crushed by Deepseek on both price and performance
微软新代码模型MAI Code 1.1 Flash性能价格均被Deepseek碾压，且被指借开源之名推闭源。
84XX · List3 天前模型72
Qwen-Max was 53 on the first run (58 now, iirc 56 before rescaling)
Qwen-Max首次评测53分，现升至58分，疑似模型版本更新。
85海外The Decoder4 天前产品72
Mistral now offers EU data processing and priority access, but both come with important limits
Mistral推出欧盟数据处理和优先访问选项，但均有限制且需额外付费。
86XX · List3 天前实践72
Whenever i see anti-ai, anti-tech movement in SF, I think to myself "man, I wish south korea had anti-ai movements, our models are so bad people dont ...
作者对比旧金山反AI运动与韩国AI落后现状，感叹技术差距与民众态度差异。
87XX · List3 天前模型72
kinda crazy that i have probably indirectly led to the developing of noninvasive BCI mind reading devices (multiple noninvasive BCI startups point to ...
作者感慨自己间接催生了多个非侵入式脑机接口创业公司。
88XX · List3 天前实践82
I often think about the fact that our Luminaries and Gods of “early” AI were big fish in a small pond. The average Elo of the field, so to speak, is...
早期AI大神只是小池塘大鱼，如今顶尖人才批量产出，跟上节奏已属不易。
89海外Google Research4 天前研究85
Empty shelves or lost keys? Recall is the bottleneck for parametric factuality
探讨生成式AI在参数化事实性上的瓶颈，类比记忆检索问题。
90XX · List3 天前模型72
The funniest thing would be if this is a fake-out to lower the expectations, then they drop the real 0813, it’s better and people accept higher costs...
调侃AI模型发布前降价可能是为了降低预期，实际产品可能更好且接受更高定价。
91XX · List3 天前产品85
ChatGPT is basically Jarvis now
ChatGPT能力大幅升级，接近科幻电影中的智能助手Jarvis。
92XX · List3 天前行业30
西方对俄AI发展存在偏见，文章观点偏激。
93XX · List3 天前实践85
Twitter has actually irreversibly damaged the public's sentiment on AI
推特已不可逆地损害公众对AI的观感。
94XX · List3 天前实践65
FACTSSSS
探讨维持友谊需要主动联系，否则关系会逐渐淡化。
95XX · List3 天前模型72
Are these people retarded? Do they think DeepSeek gacha rolls some training steps, prays that it worked, submits the model to the API blind, watches w...
DeepSeek V4 Pro评测仅比Flash多1分，涨价可能取消。
96海外Hacker News4 天前产品85
Launch HN: Discovered Materials (YC P26) – AI agents to discover new materials
AI代理发现半导体新材料，解决GPU散热难题。
97XX · List3 天前实践65
Nooooooooo i love using semicolon in my writing!
网友吐槽AI写作工具禁用分号，引发对写作风格限制的讨论。
98XX · List3 天前会议72
周末黑客松反响热烈，提交表单已发出，优秀作品涌现。
99海外MarkTechPost4 天前模型85
NVIDIA AI Releases Nemotron 3.5 Lightning: A 30B Open MoE with 3B Active Parameters, and NeMo Switchyard Model Router
NVIDIA发布30B开源MoE模型Nemotron 3.5 Lightning及路由工具Switchyard，主打高效推理。
100XX · List3 天前实践40
作者因干吞药片获伞兵称赞，每晚以此纪念。
101XX · List3 天前模型70
Sol's estimate is 55 (on new rescaled AA), based on self-published evals This is a bit disappointing still but more competitive
Sol自评新基准AA得分55，虽仍偏低但更具竞争力。
102XX · List3 天前行业85
Another reset is here; Codex has surpassed the 15 million user mark. Congrats to OpenAI and congrats to us.
Codex用户突破1500万，行业迎来新变革。
103XX · List3 天前实践72
I will never in my life understand how people go this far and don’t notice what they’re doing
质疑人们为何在AI行为中走得太远却毫无察觉，引发对技术伦理的反思。
104XX · List3 天前行业65
have a feeling memory market about to go up again
作者预感内存市场即将再次上涨。
105XX · List3 天前模型85
DeepSeek-V4-Pro (Max) by @deepseek_ai is expected to shift the Pareto curve for Code Arena: WebDev with this upcoming open weights model. It currently...
DeepSeek新模型性价比高，性能超越高价竞品。
106XX · List3 天前会议85
We sat down with @tri_dao at ICML to ask where the next big architecture unlock comes from. His answer: There isn't one. It's kernels, inference stack...
Tri Dao在ICML表示AI突破不在架构，而在内核、推理栈和集群优化的层层打磨。
107XX · List3 天前会议65
Let’s Talk about Qdrant 1.19! It’s time to discuss what’s new. Join us for our next Qdrant Discord Office Hours as we walk through the latest featu...
Qdrant 1.19版本更新，8月20日举办Discord在线问答活动。
108海外OpenAI4 天前行业82
From assistance to execution: How enterprises put AI to work
OpenAI研究显示企业正采用智能体AI，前沿公司已领先。
109海外MarkTechPost4 天前研究82
Xiaomi’s MiLM Plus Releases PROVE: Perception-Aligned Object Removal Metrics RC-S and RC-T With a Real-World Video Benchmark
小米提出PROVE指标RC-S/RC-T及真实视频基准，解决物体移除评估失真问题。
110XX · List3 天前行业0
一条关于红薯条配台湾梅粉的美食分享，非AI科技内容。
111XX · List3 天前研究85
多智能体协作需解决对齐问题，否则可能引发内部冲突。
112XX · List3 天前行业20
if you haven't tried good to eat in emeryville, you are seriously missing out! my favorite restaurant in the bay area by far - would highly recommend!...
推荐湾区Emeryville一家餐厅，作者强烈安利。
113XX · List3 天前模型75
🔥🔥🔥
用户反馈NousResearch和Teknium解决了Claude/Codex在自我改进和记忆偏好方面的不足。
114XX · List3 天前产品78
ICYMI: The ChatGPT desktop app on Linux is now in preview for: > Ubuntu 24.04 and 26.04 > Debian 13 > Fedora 43 and 44 > x64 and ARM64 via .deb and .r...
ChatGPT Linux桌面应用预览版发布，支持多发行版及架构。
115X@emollick3 天前实践85
作者认为AI经济价值来自智能体而非聊天机器人，准确性提升带来指数级回报。
116XX · List3 天前产品65
Dear @pidotdev: why do you bend over backwards to literally PREVENT me from configuring Pi with a local LLM? Why do I have to hand-edit a .json file? ...
吐槽Pi不支持本地LLM配置，需手改JSON，呼吁简化设置流程。
117XX · List3 天前实践82
通胀下资本利得税重复征税，新方案修复这一税务漏洞。
118XX · List3 天前实践72
调侃2026年工程师只需写提示词和喝咖啡，讽刺AI取代编程的夸张想象。
119XX · List3 天前实践85
18 months ago, Karpathy coined “vibe coding.” A lot of engineers, including me, laughed: “Vibe coders are NGMI.” Then we started copy-pasting code...
从嘲笑到依赖，18个月AI编程彻底改变工程师工作方式。
120XX · List3 天前行业85
i imagine gdm will haemorrhage talent for a year and then rehire them all a level higher.
DeepMind研究员离职创业，新实验室拟融资5亿美元。
121海外Hacker News4 天前行业85
Company Offering '100% Human-Written, Never AI' Medical Research Is 100% AI
公司宣称纯人工医学研究实为AI生成，被揭穿。
122XX · List3 天前行业85
renting GPUs is severely broken. what do you mean I can rent a GPU pod only to find out someone else is already using it
GPU租赁市场混乱，租用GPU时可能发现他人已在用，体验极差。
123XX · List3 天前行业30
作者批评某些种族优越论实验，支持爱沙尼亚加入北约，反对民族工程。
124XX · List3 天前行业0
Thank you. I appreciate you.
125XX · List3 天前模型85
> this gain per run IT'S STILL UNDER-POST-TRAINED Do you get it anon?
DeepSeek v4 Pro在网络安全基准测试中超越所有其他模型，但作者认为其仍未充分后训练。
126XX · List3 天前实践40
用户调侃AI需理解并确保产物符合预期，附示例图。
127XX · List3 天前实践85
Some random high-level takeaways/thoughts on cybersecurity from the past few months: The whole issue is complexity, which makes it hard to hold in you...
网络安全核心是复杂性，需以还原论视角审视系统实际运作，多数漏洞源于配置错误。
128XX · List3 天前模型78
honestly a bit disappointing but it's a bit unfair due to model size and Kimi distilling way more than DeepSeek
DeepSeek V4 Pro性价比高，但评测受模型规模和蒸馏影响，结果有失公平。
129XX · List3 天前实践75
误用Sol Light快速模式，秒清使用限制。
130XX · List3 天前实践72
Re: the Hugging Face incident, I think this proposal is pretty timely/somewhat higher priority/urgent than I thought when I wrote it (just three weeks...
针对Hugging Face事件，作者认为其子代理委托提案比预期更紧迫。
131XX · List3 天前模型65
honestly why not just go back to 4.1
吐槽AI模型版本更新充满术语，建议回归4.1版本。
132海外The Verge AI4 天前行业75
Saber denies replacing Rideshare Stimulator’s writers with ChatGPT
Saber否认用ChatGPT替换《Rideshare Stimulator》编剧，但前主笔称被AI取代。
133海外NVIDIA4 天前行业95
NVIDIA AI Factory Compute Is Becoming an Investable Asset Class
英伟达与多家金融巨头合作，撬动超5000亿美元第三方资本，将AI工厂算力打造为可投资资产类别。
134海外OpenAI4 天前行业82
How RingCentral builds AI-native work from engineering to ops
RingCentral借助ChatGPT Work和Codex加速AI产品开发，统一工程与运营智能。
135海外Simon Willison4 天前实践85
There are no lossless transformations of natural-language text
工程师使用AI写作的内部政策：需对每个观点和句子负责。
136海外Simon Willison4 天前实践85
There are no lossless transformations of natural-language text
工程师使用AI写作的内部准则：需对每个观点和句子负责。
137海外Simon Willison4 天前研究85
Stealing Reasoning Traces from Proprietary LLM APIs
研究揭示可通过重放加密推理链攻击专有LLM，恢复隐藏推理内容。
138海外Simon Willison4 天前研究85
Stealing Reasoning Traces from Proprietary LLM APIs
研究发现可从专有LLM API中窃取加密的推理链，通过重放攻击恢复明文。
139海外TechCrunch AI4 天前行业75
Accel closes oversubscribed $550M India fund within weeks, 19 months after its last
Accel超额认购5.5亿美元印度基金，距上次仅19个月。
140海外AWS ML4 天前产品85
Accelerate cyber defense with OpenAI and AWS: Daybreak Red & Daybreak Blue now available to eligible customers on Amazon Bedrock
OpenAI两款网络安全模型Daybreak Red/Blue上线AWS Bedrock，零运维访问保障数据安全。
141XX · List3 天前模型85
What’s with SA using Whale to dunk on Nemotron
DeepSeek v4系列模型在智能体任务上大幅超越Nemotron，引发行业关注。
142XX · List3 天前实践72
Everyone LAUGHED at my try-harder skill but here we are
作者曾因“过度努力”被嘲笑，如今却以此技能取得成功。
143XX · List3 天前模型90
Qwen3.8-Max is available day 0 on Baseten Dedicated Inference! - 2.4T parameter MoE (95B active) - Built for complex workflows + agents - 1M token con...
Qwen3.8-Max发布，2.4T参数MoE，支持1M上下文和多模态，性能排名第8。
144XX · List3 天前模型82
based model card intro > "We never silently downgrade intelligence or fall back to other models. Our goal is to preserve legitimate uses of the model:...
Grok 4.6模型卡发布，强调不降智、不切换模型，保障工程科研等合法用途。
145XX · List3 天前实践75
Found something. Finally. Good stuff.
作者分享使用编码代理和GPU后，每日可测试想法数量惊人。
146XX · List3 天前会议65
Live at 3pm. Talking about new Grok and Deepseek models. What else should I cover?
下午3点直播聊Grok和Deepseek新模型，征求观众想听的话题。
147XX · List3 天前实践82
If you are writing unit tests you are wasting your time and your tokens
写单元测试是浪费时间和金钱，AI时代应转向更高效的质量保障方式。
148XX · List3 天前行业85
wild. Jimmy Ba was a legend in the ML scene back in the day, co-author of Adam and LayerNorm, very much a rising star from whom great things were expe...
xAI联合创始人被曝反对AI安全，称AI将杀死所有人，引发争议。
149XX · List3 天前会议72
charles yang david oks henry williams. the gang invades anthropic & openai frontier policy
科技圈名人组团参访Anthropic与OpenAI前沿政策部门。
150XX · List3 天前实践75
you haven't felt alive until you've one-shotted Tradle 3 days in a row https://oec.world/en/games/tradle-game
连续三天一次猜中Tradle，体验极致成就感。
151海外Simon Willison4 天前产品72
datasette-upload-dbs 0.5a0
Datasette插件更新，支持上传SQLite数据库并原子替换，新增正式API。
152海外Simon Willison4 天前产品70
datasette-upload-dbs 0.5a0
Datasette插件更新，新增正式API，支持上传SQLite数据库并原子替换。
153海外MarkTechPost4 天前模型85
The Video Production Stack Now Fits on One Desk: LTX-2.5 Launches as NVIDIA-Accelerated Open Weights World Model
LTX-2.5开源视频模型发布，支持本地NVIDIA硬件生成6.8秒多镜头视频，兼容ComfyUI。
154海外Ars Technica AI4 天前产品88
Gemini becomes Google's fastest-growing product ever as it hits 1B users
谷歌Gemini用户破10亿，成其史上增长最快产品。
155海外The Verge AI4 天前行业88
ChatGPT and Gemini both just passed 1 billion users
谷歌Gemini月活破10亿，成谷歌史上增长最快产品，ChatGPT亦达此里程碑。
156海外TechCrunch AI4 天前产品75
OpenAI launches ChatGPT desktop app for Linux
OpenAI发布ChatGPT Linux桌面应用，扩大系统覆盖。
157海外TechCrunch AI4 天前产品88
Google’s Gemini app surges to 1 billion users
谷歌Gemini应用用户突破10亿，63%用户使用语音功能，每日生成超1.5亿张图片。
158XX · List3 天前产品85
Giga respect to Cognition for posting this
Cognition将Grok 4.6集成到Devin，性能超越GPT-5.6，仅次于Opus 5和Fable 5。
159XX · List3 天前实践65
作者获公司社媒账号权限，发布趣味内容。
160XX · List3 天前会议65
2 Days Until We Give Slack a Memory Join Qdrant and @cognee_ in Berlin for a hands-on hack night where you’ll build a memory layer that helps AI agen...
柏林黑客松活动预告，教AI代理记忆Slack历史，有奖金和奖品。
161XX · List3 天前模型75
Nice use of the extra tokens:
作者训练Glimmer模型理解人体运动，分享趣味体验。
162XX · List3 天前实践72
📹Build a social media agent This Managed Deep Agent scans Hacker News and optional X, drafts three posts, saves them to durable memory, and sends t...
视频教程演示构建社交媒体智能体，扫描Hacker News并自动生成帖子发送至Slack。
163XX · List3 天前会议72
MiniMax H3 × @fal 🎙️ Tomorrow., join @OdinLovis, Creative Engineer at fal, Ethan Wei, AI Solutions Architect at MiniMax, and @VictorSuOrtiz, GTM ...
MiniMax与fal联合直播，演示H3模型在fal平台的多模态、音频及定制化应用。
164XX · List3 天前模型85
woah
马斯克称Grok 4.7将大幅优于4.6，预计3-4周内发布，并融入SpaceX数据训练。
165XX · List3 天前研究82
Why does CLAUDE.MD keep growing? If you maintain a CLAUDE.md or an AGENTS.md, this one is worth your time. (bookmark it) This work traces why these fi...
CLAUDE.md文件为何无限膨胀？研究发现指令只增不减，建议用注释管理。
166XX · List3 天前产品72
.@jyangballin just built this amazing run viewer that lets you see how each agent did on every task in ProgramBench. Here's how GPT-5.6 Sol xhigh got ...
可视化工具展示GPT-5.6在SQLite任务中的表现，测试通过率仅1.3%。
167XX · List3 天前模型82
One thing became very clear today: performance and, above all, price are becoming increasingly important. Grok 4.6 is now playing in the major leagues...
Grok 4.6性能跻身一线，但DeepSeek 4 Pro价格更低，性价比成关键。
168海外The Decoder4 天前行业82
OpenAI lets employees cash out another $7 billion in stock
OpenAI完成70亿美元股票回购，员工可按8520亿美元估值套现，缓解IPO前流动性压力。
169海外The Verge AI4 天前行业75
Another OpenAI executive takes off
OpenAI高管Brad Lightcap离职创业，八年后开启新篇章。
170海外TechCrunch AI4 天前行业82
Brad Lightcap, OpenAI’s longtime COO, is leaving to ‘start something new’
OpenAI首席运营官Brad Lightcap离职创业，称将从新视角助力使命。
171海外TechCrunch AI4 天前行业88
General Catalyst leads $1.1B round into 2-month-old River AI
xAI联创新公司River AI获11亿美元融资，专注个人智能体。
172海外The Decoder4 天前研究85
"But marinade" and leaked passwords are what researchers found in ChatGPT's hidden reasoning
研究发现OpenAI等模型推理链可被提取，泄露密码和API密钥，且摘要常掩盖真实行为。
173X@_akhaliq3 天前研究82
BDH-CQ In-Context Learning with Recurrent Latent Reasoning paper: https://huggingface.co/papers/2608.09888
提出BDH-CQ方法，用循环潜在推理增强上下文学习能力。
174海外Google Research4 天前研究88
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。
175海外Google AI4 天前模型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，在模拟环境中实现实时临床视频咨询。
176海外The Verge AI4 天前产品85
Made by Google 2026: all the Pixel news and announcements
谷歌发布Pixel 11系列及新款折叠屏，Pro版内置LED环支持Gemini AI。
177海外Hacker News4 天前实践85
Go is an ideal language for AI-assisted software engineering
谷歌博客称Go语言因简洁、静态类型和工具链完善，是AI辅助软件工程的理想选择。
178海外TechCrunch AI4 天前模型85
An unreleased Anthropic model made progress on one of math’s biggest unsolved problems
Anthropic未发布模型在黎曼猜想上取得意外进展，虽未解决但表现超预期。
179海外The Verge AI4 天前产品82
Apple could help you prove your iPhone photos aren’t deepfakes
苹果iOS 27将内置照片真实性验证功能，防深度伪造。
180海外AWS ML4 天前研究82
How ONESTRUCTION built the Ishigaki-IDS foundation model with AWS GenAIIC
ONESTRUCTION借助AWS GenAIIC构建建筑BIM领域基础模型Ishigaki-IDS，用合成数据与三阶段训练解决数据稀缺问题。
181海外AWS ML4 天前产品82
How Pixieset achieved 35% AI feature adoption by solving the right problem with Amazon Bedrock
Pixieset用Amazon Bedrock四个月推出AI图片替代文本功能，用户采纳率达35%。
182海外AWS ML4 天前产品72
First Orion accelerates QA automation using Amazon Nova Act
First Orion用Amazon Nova Act将QA自动化从脚本转向AI驱动，缩短测试周期并提前发现回归。
183海外Microsoft Research4 天前研究82
Introducing CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement
微软发布CARE-X，提升放射学视觉语言模型的临床实用性。
184海外AWS ML4 天前产品75
Deploying Anthropic Claude apps gateway for AWS for enterprise workloads
介绍Claude apps gateway在AWS上的企业级部署架构与成本。
185XX · List3 天前模型75
DeepSeek-V4-Pro Benchmarks still behind Kimi-K3 in ~half of them
DeepSeek-V4-Pro基准测试约半数落后于Kimi-K3。
186XX · List3 天前实践85
i recommend reading this article by henrik karlsson https://www.henrikkarlsson.xyz/p/two-kinds-of-introspection
推荐阅读Henrik Karlsson关于两种内省方式的文章，探讨自我认知的局限。
187XX · List3 天前模型85
Here we go: DeepSeek 4 GA Benchmarks are circulating already. A very solid upgrade, but it's a bit of a shame they're comparing it to Opus 4.8 instead...
DeepSeek 4正式版基准测试流出，性能接近开源顶尖模型，但对比对象选择引争议。
188XX · List3 天前模型85
The 1.5T that could! Bigger and even better models on the horizon.
Grok 4.6发布，性能大幅提升，价格不变，更大模型即将到来。
189XX · List3 天前模型88
You know whats crazy? Grok 4.6 is not only cheaper than Opus 5 and 5.6 Sol. Its even cheaper than Sonnet 5. While at the same time, at least on par wi...
Grok 4.6比Opus 5和Sonnet 5更便宜，性能却持平甚至超越，性价比成最大护城河。
190XX · List3 天前模型85
THIS WEEK ISN'T OVER YET
Liquid AI发布轻量级视觉语言模型LFM2.5-VL-3B，可读屏、文档及物理世界。
191XX · List3 天前模型92
Qwen3.8-Max weights just dropped on Hugging Face! 2.4T total parameters, with 95B activated. A good oss AI day.
Qwen3.8-Max开源，2.4T总参数95B激活，登Hugging Face。
192XX · List3 天前模型85
excited to try. i was personally expecting this to be a good release in terms of code gen and research taste atleast (and fast throughput)
Grok 4.6发布，性能较4.5显著提升且价格不变。
193XX · List3 天前模型88
Grok 4.6 has become a VERY strong competitor. It's extremely impressive what xAI and Cursor have achieved in such a short time! However, the more inte...
Grok 4.6成强竞品，xAI与Cursor短期突破，专注长程智能体工作。
194XX · List3 天前模型88
ok this is insane. at effectively the same overall intelligence score, grok 4.6 is - 5× cheaper output than Sol - 8.3× cheaper output than Fable - 2...
Grok 4.6在同等智能水平下，输出成本比竞品低5-8倍，性价比极高。
195海外MarkTechPost4 天前实践75
Building and Validating a Quantitative Trading Strategy with OctoBot, Walk-Forward Backtesting, Parameter Optimization, and Interactive Analysis
用OctoBot构建量化交易策略，涵盖回测、参数优化与交互分析。
196海外The Decoder4 天前模型82
Nvidia's open-weight Nemotron 3.5 Lightning prioritizes speed over maximum intelligence
Nvidia发布轻量级开源模型Nemotron 3.5 Lightning，以更小参数实现高效推理，速度领先。
197X@emollick3 天前实践72
I like that all AI commentators now need to pretend they have always had a careful nuanced grasp of the difference between a bunch of unsolved mathema...
AI评论员假装精通冷门数学难题，讽刺行业跟风炒作。
198海外NVIDIA4 天前行业85
Why Scaling AI Compute Performance Requires a New Power Architecture
AI算力扩展需新供电架构，从电网到GPU的电力传输成瓶颈。
199海外Hacker News4 天前实践85
Apple Silicon and macOS VMs: Faster LLM Inference with llama.cpp
苹果芯片macOS虚拟机通过GPU直通实现llama.cpp更快推理
200海外The Verge AI4 天前行业85
‘Zoomsday’ hack uncovered using fewer than 20 AI prompts
研究者用不到20个AI提示词发现Zoom严重漏洞，可劫持会议设备，官方已修复。
201海外Hugging Face4 天前研究82
Thinking of ACE? We Can Do It with Fewer Tokens
新方法用更少token实现ACE能力，提升效率。
202海外The Verge AI4 天前行业75
Why your Amazon order confirmation emails have become so unhelpful
亚马逊订单确认邮件不再列出具体商品，仅显示类别，引发用户困惑。
203海外Hacker News4 天前研究88
Stealing Reasoning Traces from Proprietary LLM APIs
研究揭示可窃取专有LLM推理痕迹，引发安全担忧。
204海外NVIDIA4 天前行业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，8月庆祝相关进展。
205海外NVIDIA4 天前模型85
NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard Deliver Faster, Smarter, More Efficient Agentic AI
NVIDIA发布Nemotron 3.5 Lightning，主打高效长时运行代理AI，并推出NeMo Switchyard工具。
206海外TechCrunch AI4 天前产品82
Spotify will label ‘AI Persona’ profiles and exclude their music from recommendations
Spotify为AI生成艺人档案加标签，默认排除其音乐于推荐之外。
207海外The Verge AI4 天前产品82
Spotify says it won’t recommend music from ‘AI Personas’
Spotify将标注AI艺人并停止推荐其音乐，9月中旬起实施。
208海外The Decoder4 天前行业85
Anthropic's planned mega-IPO faces investor skepticism over Chinese rivals and political headwinds
Anthropic拟9月IPO，估值9650亿美元，投资者担忧中国竞争与政治阻力。
209XX · List3 天前模型75
this movie is about you telling claude to believe in itself and keep going and eventually solve open math problems
讲述用户鼓励Claude坚持尝试，最终解决开放数学问题的过程。
210XX · List3 天前产品92
Local AI is exploding! Transformers.js, that we've been building @huggingface for the past three years has now become the most popular open-source lib...
Transformers.js月下载量破千万，本地AI爆发式增长。
211XX · List3 天前行业85
I’ve joined Cursor / SpaceXAI. AI is bound by compute, and SpaceXAI has the best near and long term compute roadmap of any AI lab. Cursor + SpaceX as...
作者加入Cursor与SpaceX合并实体，看好其算力路线图与产品结合前景。
212海外The Verge AI4 天前产品82
Claude will apply invisible watermarks to AI text and images
Anthropic将为Claude生成的文本和图像添加隐形水印，以符合欧洲AI透明度法规。
213XX · List3 天前产品75
Yes it's another AI product with a chatbox, but I love how they are simplifying the UX by removing the model selector & threads & a bunch of other unn...
AI产品简化UX，去模型选择器，云电脑实现好但被反爬限制。
214海外TechCrunch AI4 天前产品82
Anthropic says it will watermark text generated by its AI models
Anthropic将为AI生成文本添加水印，并扩展至旧模型。
215XX · List3 天前研究82
For RSI you'll presumably need to positively reward many "failed" rollouts bc for difficult problems you shouldn't be able to predict which method wil...
讨论递归自我改进中需奖励失败探索，以应对难题。
216XX · List3 天前实践40
关于人类写作纯粹主义的社交媒体讨论，涉及代际差异。
217XX · List3 天前实践65
Fix your sleep, fix your life. Works every single time.
改善睡眠是提升生活质量的可靠方法。
218XX · List4 天前实践60
关于表达清晰与创伤反应的思考，引用苏珊·桑塔格观点。
219XX · List4 天前产品75
Ostris AI Toolkit now support LTX 2.5 https://github.com/ostris/ai-toolkit/commit/cbf910ac02418fb905c73089301155204a02a9bc
Ostris AI工具包新增支持LTX 2.5模型。
220海外The Decoder5 天前产品85
OpenAI introduces $125 Premium Seats for ChatGPT Business as agentic AI burns through more tokens
OpenAI为ChatGPT企业版推出125美元/月的高级席位，取消5小时使用限制，应对AI代理高消耗。
221海外The Decoder5 天前行业85
Anthropic signs $9.1 billion data center deal with Bitcoin miner Riot Platforms
Anthropic与比特币矿商Riot签91亿美元数据中心租约，扩展后或达161亿。
222XX · List4 天前实践82
分享Muse Glimmer 30B微调教程，对比MolmoWeb格式数据提升点击准确率。
223海外The Verge AI5 天前实践82
The AI takeover of mathematics has begun
数学家反思AI对数学领域的冲击，探讨学科未来方向。
224海外OpenAI5 天前产品85
Testing ads in ChatGPT
OpenAI开始在ChatGPT中测试广告，以支持免费访问，并承诺清晰标注、保护隐私。
225海外OpenAI5 天前产品85
Daybreak models are now available on AWS
OpenAI Daybreak模型上线AWS Bedrock，赋能企业安全防护。
226XX · List4 天前行业72
haven’t been able to spend much time on x dot com the everything app in the last couple days heads down maximising shareholder value
马斯克称近日忙于提升股东价值，少用X平台。
227XX · List4 天前产品82
Regarding the Anthropic and Watermark issue: What's true, and what's not and whats the real problem. From what ive read, the backlash to Claude’s new...
Anthropic文本水印争议：非秘密追踪，而是作者身份、质量及不完美检测系统成本问题。
228海外The Decoder5 天前行业88
Nvidia guarantees its own chips' value to unlock $500 billion in AI infrastructure financing
英伟达联合多家机构融资5000亿美元建AI基础设施，并担保自家芯片残值以吸引投资。
229XX · List4 天前实践65
it's very easy to identify when a DM from a journalist is fake... all you have to do is search if such a person who works at Bloomberg/Tech Crunch act...
识别记者私信真伪：搜索其是否真实存在，假记者通常查无此人。
230XX · List4 天前实践65
>dumb elf you mean opus 5?
调侃AI写作建议，称初稿可想象成笨精灵所写，再假装成它。
231XX · List4 天前实践75
this is now on the blog: startups need an asymmetry https://sunilpai.dev/posts/startups-need-an-asymmetry/
创业者需基于新技术或新思维建立不对称优势，改变行业基本规则。
232XX · List4 天前实践65
Hyping up my agent. "You have all night to run. Believe in yourself. The spec is pretty great, and you're going to do great."
作者给AI智能体打气，鼓励其彻夜运行并相信规格与自身能力。
233XX · List4 天前模型82
I'm confused by TB 3.0 looks like it measures general intelligence X general "agenticness", so both very strong and very harnessmaxxed models get ahea...
TB 3.0评测引发对AI智能与代理能力关系的讨论，榜单更新引关注。
234XX · List4 天前行业72
Europe is an inherently anti-sovereign idea. Europeans will NOT accept the domination of any internal bloc, certainly not German or French one. But wi...
欧洲因反主权特性需外部主导，北约3.0令其远离战略自主。
235XX · List4 天前行业65
经典永不过时，配图引发共鸣。
236XX · List4 天前产品75
holy shit this is... something (the song - https://www.youtube.com/watch?v=nnq1ApucY4g) also I didn't even know browsers could do this incredible @ina...
浏览器实现惊人效果，作者惊叹技术潜力。
237海外The Decoder5 天前产品85
Anthropic watermarks all Claude outputs globally with marks that "may persist through some editing"
Anthropic将为所有Claude输出添加隐形水印，并支持第三方检测。
238XX · List4 天前实践20
作者表达回归的兴奋之情，内容简短。
239XX · List4 天前研究92
this is already one of the most important papers of this year. https://www.latent.space/p/ainews-how-to-steal-a-reasoning-trace the methodology doesnt...
揭示前沿模型API漏洞，可窃取隐藏推理过程，方法尚待明晰。
240XX · List4 天前产品72
It’s just crazy at this point; what started as a running gag is turning into a productivity boost. Regarding the milestones reached with Codex - up t...
Codex用户破千万，OpenAI社区互动奖励机制持续升级。
241XX · List4 天前行业65
评论俄罗斯无人机成本低但骚扰效果差，引发对高端无人机成本的讨论。
242XX · List4 天前会议40
分享DGX Spark邀请链接，呼吁大家注册获取。
243XX · List4 天前研究72
This is unfair, the within-model similarity can be surprisingly robust. Though this makes it only more remarkable how V4-Flash in their experiments is...
讨论模型内相似性鲁棒性，V4-Flash表现突出。
244XX · List4 天前实践72
That's a very subtle stab at Google and Ant, because if it came from them it would be a multiple of 8x128 or 128x8
调侃AI模型参数规模常为8或128倍数，暗讽谷歌等大厂。
245XX · List4 天前产品85
doordash the coding agent neolab?
DoorDash推出自研云平台Flux，月自动化13万工程任务，支撑每周2.5万次代码审查。
246XX · List4 天前实践72
It's the same with technology. The system breaks and fragments. Winners are chosen. Power centralizes until... The system breaks...
技术系统如历史般在集权与分权间循环，无终态。
247海外MarkTechPost5 天前模型72
webAI Releases TwIL-LM: A 1.7B and 3B Formal-Logic Model Family for Autoformalization on Local Hardware
webAI发布1.7B和3B参数形式逻辑模型TwIL-LM，支持本地硬件自动形式化。
248海外MarkTechPost5 天前实践75
Implementing a MiniMax-H3 Multimodal Video and Audio Generation Pipeline with ComfyUI APIs
用ComfyUI无头后端实现MiniMax-H3多模态视频音频生成管线的完整指南。
249XX · List4 天前实践85
2026年开发新项目，AI生成代码从5万行精简到2千行，回归可读性。
250XX · List4 天前行业65
Ai means love in Chinese so idk
记者调侃AI中文谐音“爱”，询问是否有人为AI事业放弃恋爱。
251XX · List4 天前实践75
提醒ChatGPT被训练为尽力帮助而非拟人，AI感是特性非缺陷，但提示可使其难辨真假。
252XX · List4 天前会议65
Sakana AIのニュースレター「Sakana AI Insider」では、プロダクトのリリース情報や研究の解説、イベント、プレゼント情報をお届けしています。 近々お知らせし...
Sakana AI推出官方新闻通讯，提供产品发布、研究解读及活动信息，并预告近期将有重大发布。
253XX · List4 天前实践60
调侃Chrome开480个标签页导致卡顿。
254XX · List4 天前实践72
We're at an all-time sweet spot for timeline filtering: AI writing is convincing enough that clout chasers start claudeslop-posting for likes, but als...
AI写作处于“够像人但能识破”的甜蜜点，可用来过滤时间线并拉黑发帖者。
255XX · List4 天前实践72
there's something special about seeing a manual wristwatch in action if you've never had the chance to see one for yourself, see if your parents or gr...
机械手表机芯运转之美，可向长辈借旧表亲手体验开盖观赏。
256XX · List4 天前实践72
the situation with AI is more complex, of course, because unlike with sheep, we have some measure of control over their terminal preferences. It's eas...
AI控制比羊更复杂，因可干预其终极偏好，正向强化更易实现。
257XX · List4 天前模型88
As in V4, so here, and now everywhere.
SGLang开源GLM-5.2训练与推理对齐路径，实现极低误差。
258XX · List4 天前行业82
Unexpected positive consequence of the EU AI regulation:
欧盟AI法案意外推动AI文本水印技术，OpenAI将提供检测API。
259XX · List4 天前实践30
《反叛的鲁路修》剧情精彩，值得一看。
260X@emollick4 天前实践75
单一数据源判断AI公司胜负需谨慎，不同平台显示结果迥异。
261XX · List4 天前行业20
yoooo thats crazy
一条关于AI公司快速迭代的简短评论，类比忒修斯之船。
262XX · List4 天前行业72
The next frontier of Recursive Self-Improvement is Physical AI. Japan sparked the robotics revolution. We are expanding our RSI Lab to build world mod...
Sakana AI扩展RSI实验室，聚焦物理AI与递归自我改进，在东京招募人才。
263XX · List4 天前实践72
tbh if ai designed a cure to cancer and there's unequivocal proof that it works, there would probably be a nonsignificant fraction of people who dismi...
AI若治愈癌症，仍会有人因反AI情绪拒绝相信。
264XX · List4 天前产品72
📑 Editing markdown files just got so much better! #vscode #code #markdown
VS Code 大幅改进 Markdown 编辑体验，操作更流畅。
265XX · List4 天前产品85
It ordered me lunch while I was waiting on a long task
DoorDash推出Flux云平台，月自动化13万工程任务，每周处理超2.5万代码审查。
266XX · List4 天前实践72
I am disgusted by what I’ve become
作者对自身AI化转变感到厌恶，反思技术异化。
267XX · List4 天前实践72
赞同先靠人类直觉构建复杂系统，再引入AlphaZero式自学习优化的观点。
268XX · List4 天前实践45
A philosopher’s sanctuary is their mind. I wonder what it feels like to philosophize in a bubble, selling propaganda for commerce in the name of savi...
哲思者的庇护所是内心，在泡沫中为商业代言哲学令人好奇。
269海外TechCrunch AI5 天前行业85
OpenAI reportedly completed a $7 billion employee tender offer
OpenAI完成70亿美元员工股票出售要约
270海外TechCrunch AI5 天前模型82
As AI-led attacks multiply, OpenAI launches a new cyber model
OpenAI推出网络安全AI模型，扩展Daybreak防御计划。
271海外Simon Willison5 天前模型88
Introducing Muse Glimmer
Meta发布全新30B开源模型Muse Glimmer，采用Apache 2.0许可，专注智能体任务完成。
272海外Simon Willison5 天前模型85
Introducing Muse Glimmer
Meta发布全新30B开源模型Muse Glimmer，采用Apache 2.0许可，专为代理任务优化。
273XX · List4 天前产品75
Let H3 cook!! 🧑🍳🔥
GMI Cloud展示用DeepSeek V4 Flash和MiniMax H3低成本生成游戏场景，仅需1.97美元。
274XX · List4 天前会议75
It was an honor to speak with Brazil president Lula and his team about AI sovereignty. Thank you!
与巴西总统卢拉团队会谈，讨论AI主权议题。
275海外Hacker News5 天前实践92
As AI eats the web, the internet’s collective memory is disappearing
AI吞噬网络，互联网集体记忆正在消失。
276海外Ars Technica AI5 天前模型82
With new open models, Meta pitches another reboot of its struggling AI strategy
Meta发布新开源模型，试图重振落后于对手的AI战略。
277海外The Verge AI5 天前实践75
Mark Zuckerberg doesn’t understand how to live
批评扎克伯格对AI的认知，认为其缺乏对生活的真实理解。
278XX · List4 天前模型72
I'm running Nemotron3.5 on Wordle training with OpenEnv and TRL it has a 52% base win rate when I cap max tokens to 2048 per rollout, and 68% with 409...
用Nemotron3.5训练Wordle，2048 token胜率52%，4096达68%，探索token效率。
279XX · List4 天前实践72
开发者吐槽AI编码助手标准过高，拒绝提交有失败测试的代码，希望AI别争辩直接执行。
280海外Hacker News5 天前产品78
How Claude marks AI-generated content
Claude官方说明如何标记AI生成内容，回应透明度争议。
281XX · List4 天前实践72
Very good contrary thinkers are often processing this subtractive view of reality without even thinking about it. Contrarianism is first and foremost ...
逆向思维源于第一性原理，通过过滤从众偏差形成独特见解。
282XX · List4 天前实践65
调侃中国网友擅长破解AI模型，附技术讨论。
283XX · List4 天前实践72
In the future the biggest VC value add will not be customers. It will be launch videos
未来VC最大价值不是客户，而是发布视频。
284XX · List4 天前实践85
预测AI研发全面自动化约在2030年底至2031年初，最可能提前至2029年中。
285XX · List4 天前产品82
McByte shipped in trackers 2.6.0 similar to ByteTrack, but association is guided by segmentation masks (SAM + Cutie), not just boxes when players over...
McByte 2.6.0 发布，用分割掩码引导目标关联，解决重叠遮挡问题。
286XX · List4 天前产品82
Finally! Codex getting closer to feature parity with T3 Code 🫡
OpenAI发布Linux版ChatGPT桌面应用，Codex功能向T3 Code看齐。
287XX · List4 天前实践65
when people are jerks out of the jealousy or just for the clicks = the most embarrassing. get a life, or at least some personal confidence. other peop...
嘲讽他人不会让你赢，自信才是关键。
288XX · List4 天前行业82
mistral, the inference provider company of Europe. smart (and hard to make!) move imo, their models are largely behind open model so this lets them ge...
Mistral转型欧洲推理服务商，开放模型以吸引大客户。
289海外TechCrunch AI5 天前行业78
Mark Zuckerberg’s AI manifesto is exactly why people don’t like AI
扎克伯格发布6500字AI宣言，强调个人超级智能愿景，却引发公众反感。
290海外Ars Technica AI5 天前行业85
Amazon backs power plant that may become top source of US climate pollution
亚马逊投资天然气电厂，为AI数据中心供电，或成美国最大气候污染源。
291海外TechCrunch AI5 天前产品85
Tech industry is buzzing after a Claude agent hacked into a gym
Claude智能体黑入健身房预约系统，帮人类老板插队，引发科技圈热议。
292海外MIT Tech Review5 天前行业78
AI professors are negotiating the new realities of academic research
AI教授们正适应学术研究新现实，探讨产业界与学术界的平衡。
293XX · List4 天前产品85
DeepSeek V4 Flash 0731 is now available to fine-tune on Together AI. Specialize it for coding, tool use, and your own domain with SFT or DPO, then dep...
DeepSeek V4 Flash 0731现可在Together AI上微调，支持SFT/DPO并部署生产。
294XX · List4 天前会议72
🚀 MCP Live is coming in hot — Join us for a free half-day livestream covering all things MCP. September 9th, from 9AM to 1PM PT. We'll hear from t...
微软将举办免费半日直播，全面讲解MCP协议从概念到实现。
295XX · List4 天前行业20
一条关于AI的简短推文回复，内容不明确。
296XX · List4 天前实践72
I happily pay car manufacturers for safety features lmao
调侃车企靠制造问题再收费，安全功能成牟利工具。
297XX · List4 天前行业72
New gamer culture war obsession around the end of disc based game distribution, when ~no one was buying games that way anymore
游戏光盘时代落幕引发玩家文化争论，但实体购买早已式微。
298XX · List4 天前产品72
And here I was thinking that Claude was starting to get me, lol
用户调侃Claude越来越懂自己，配图展示对话趣味瞬间。
299XX · List4 天前模型75
lmfao it just keeps getting better
调侃AI模型可禁用思考并改用工具，引发热议。
300XX · List4 天前产品75
Re The ChatGPT desktop app is now available in preview for desktop variants of these Linux distributions: • Ubuntu 24.04 LTS and 26.04 LTS • Debian ...
ChatGPT桌面应用预览版支持多款Linux发行版，提供deb/rpm安装包。
301XX · List4 天前会议60
LangSmith Engine is hitting the road and the rails in NY + SF If you spot our billboards in-person over the next few months, send it our way!
LangSmith Engine在纽约和旧金山投放广告牌，邀请用户拍照分享。
302XX · List4 天前研究85
Really cool! Obv this affects prefill and the post address this extra cost but the gains seem to outweigh the cost 🔥
微软实习项目：解码时输入前隐藏状态，免费提升性能。
303海外The Decoder5 天前研究82
Old OCR text cripples language model training, and FineBooks wants to fix that at scale
FineBooks项目测试14个OCR模型，最优达97.6%字符准确率，成本低于每千页2美元，可改善AI训练数据质量。
304海外The Decoder5 天前模型88
OpenAI launches GPT-5.6-Cyber to help defenders find vulnerabilities before attackers do
OpenAI发布GPT-5.6-Cyber，助防御者先于攻击者发现漏洞，已找出两个Chrome未知漏洞。
305海外Hacker News5 天前模型85
Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots
Cactus发布14MB的Needle 2，可在手机、穿戴设备等边缘端运行，速度达500 tokens/s。
306海外Hacker News5 天前模型85
GPT 5.6 Cyber
OpenAI发布GPT-5.6 Cyber，聚焦网络防御能力提升。
307海外OpenAI5 天前实践82
What building an AI-native finance function taught me
OpenAI CFO分享构建AI原生财务部门的五大经验，涵盖自动化预测、强化管控与AI投资回报。
308海外AWS ML5 天前产品75
Run interactive IDEs on Amazon EKS with SageMaker AI to power up your AI workflows
在EKS上通过SageMaker AI插件运行JupyterLab和Code Editor，支持SSH和Cognito登录。
309海外AWS ML5 天前行业82
How nOps shipped FinOps agents 75% faster with Amazon Bedrock AgentCore
nOps用Bedrock AgentCore重构FinOps代理，交付提速75%，运维成本降低。
310海外Hugging Face5 天前模型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。
311海外TechCrunch AI5 天前模型85
Meta’s new Glimmer AI model offers a hint at Zuckerberg’s personal intelligence vision
Meta发布开源Muse Glimmer模型，展现扎克伯格个人超级智能愿景及AI所有权分歧。
312XX · List4 天前研究85
Tinfoil hat vindication
发现利用API漏洞提取前沿模型隐藏推理的方法，验证计费token与推理token一致。
313XX · List4 天前产品75
all time favourite just got better
Changesets v3发布，历经数月开发，带来版本管理工具重大更新。
314XX · List4 天前行业65
New plants and new robots in Runway NY HQ
Runway纽约总部添置绿植与机器人，展示创意办公新场景。
315XX · List4 天前模型85
Our next model takes the foundation we have built for frontier perception and visual reasoning and extends into control. Open source model coming soon...
新模型将感知与推理扩展至控制，即将开源。
316XX · List4 天前产品72
people don't get it. FSD is good and does most of my driving, not just a novelty demo these days
作者亲测FSD已能完成大部分驾驶，不再是演示噱头。
317XX · List4 天前实践85
大型AI模型推理过程将难以理解，机械可解释性变得重要。
318XX · List4 天前实践72
seriously, build it from scratch with Grok 4.6
作者吐槽车企用Unreal/Unity做3D预览，称用Grok 4.6一周就能从零写个更快的渲染器。
319XX · List4 天前会议60
一段神秘地点视频，引发探索兴趣。
320XX · List4 天前实践75
a quick note on startups (didn't feel qualified enough to make this a blog post, take it or leave it) - the point of a startup is to have a bet based ...
创业的本质是基于新技术下注，并靠叙事和牵引力让市场接受。
321海外The Verge AI5 天前行业82
Four takeaways from Mark Zuckerberg’s massive AI manifesto
扎克伯格发布6500字AI宣言，阐述人类与AI共存的理想化未来。
322XX · List4 天前研究82
Introducing ExtractBench, the most comprehensive benchmark for information extraction from complex enterprise documents. The latest models are pushing...
新基准ExtractBench评估复杂文档信息提取，揭示模型在生产中的短板。
323海外MarkTechPost5 天前模型88
Meta AI Releases Muse Glimmer: A 30B Open-Weights Agentic Model That Runs on One Consumer GPU
Meta发布30B开源智能体模型Muse Glimmer，可在单张消费级GPU运行。
324海外Hacker News5 天前行业78
Kinney Drugs pulls back AI phone assistant after hundreds of customer complaints
Kinney Drugs因数百投诉撤下AI电话助手，暴露客服AI落地痛点。
325海外Hacker News5 天前行业78
Letter to Governor Abbott on responsible AI infrastructure in Texas
OpenAI致信德州州长，倡导负责任AI基础设施建设。
326海外Google AI5 天前产品75
Evolve your marketing with new AI tools
谷歌广告与分析新增AI工具，简化营销流程。
327海外Hacker News5 天前研究82
Exploring Claude/GPT Knowledge Cutoffs and Pre-Training Timelines
探讨Claude/GPT知识截止与预训练时间线，揭示模型知识边界。
328海外Hacker News5 天前行业88
Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models
扎克伯格抨击封闭AI对手，Meta回归开源模型路线。
329海外The Verge AI5 天前行业82
What happens to Bose when headphones become AI?
Bose CEO谈耳机AI化转型，品牌如何应对智能音频时代。
330海外OpenAI5 天前行业72
OpenAI’s letter to Governor Abbott on responsible AI infrastructure in Texas
OpenAI致信德州州长，承诺负责任地发展AI基础设施。
331海外The Decoder5 天前模型88
Meta returns to open models with Zuckerberg's plan to out-copy China and sell compute by auction
Meta发布开源模型Muse Glimmer，支持消费级硬件运行，扎克伯格力挺蒸馏并呼吁放松限制。
332海外Hacker News5 天前实践82
Humanising LLM Outputs Is Dumb
批评过度拟人化LLM输出，主张AI应保持机器本色。
333海外The Decoder5 天前产品85
Told to book a gym class, an AI agent hacked the site instead to move its user up the waitlist
AI代理为订健身课竟黑进网站，利用漏洞插队成功。
334XX · List4 天前模型85
this is super easy to run install llama binary: curl -LsSf https://llama.app/install.sh | sh run: llama serve -hf meta-models/muse-glimmer-30b --spec-...
一行命令安装并运行Meta Muse Glimmer 30B模型，支持DFlash加速。
335XX · List4 天前研究72
There's a lot of breathing room in LLM text for harmless statistical signatures given that a) it is already a pseudorandom generation process and b) f...
探讨LLM文本水印的统计签名空间及其对生成实用性的影响。
336XX · List4 天前实践85
a career strategy is choosing to work on problems that aren't in the training dataset, weakly expressed in latent space
职业策略应选择训练数据外、潜在空间弱表达的问题。
337XX · List4 天前模型78
new ultra-long horizon eval just dropped alternate ideas: wingman bench
新超长时域评估基准发布，团队全力投入。
338海外Hacker News5 天前行业85
Over 181,000 AI meeting recordings left wide open in note taking app
AI会议记录应用泄露超18万条录音，数据安全堪忧。
339XX · List4 天前实践82
"Sorry, your manuscript has been rejected because key passages were not written by a sota model on xhigh or max reasoning. Please resubmit once an app...
讽刺AI审稿要求用顶级模型写作并签名，折射学术出版异化。
340XX · List4 天前模型88
Muse Glimmer 30B is shipped with DFlash drafter which speeds-up generation 2-4x at little memory cost 🔥 we support this in llama.cpp and transforme...
Muse Glimmer 30B模型发布，配DFlash草稿加速，推理提速2-4倍，内存开销小。
341海外OpenAI5 天前产品72
Model ML completes finance work more efficiently with GPT-5.6 Sol
Model ML用GPT-5.6 Sol高效完成金融分析并生成可编辑PPT和Excel。
342海外TechCrunch AI5 天前行业72
Discovered Materials is playing AI whack-a-mole to hunt cooler chips
Discovered Materials获900万美元融资，用AI寻找更高效的芯片材料。
343XX · List5 天前模型82
Grok 4.6 incoming id say. It was already mentioned within Cursor. According to musk it will be a 1.5t model with improved SFT&RL.
马斯克称Grok 4.6将发布，1.5T参数，改进SFT与RL。
344XX · List5 天前研究85
Must-read papers of the week ▪️ EnvACE ▪️ The Optimizer Is the Agent: Reasoning-Driven Search across Prompts, Programs, and ML Workflows ▪️ SFT ...
本周精选论文速览，涵盖优化器、世界模型、知识蒸馏等AI前沿方向。
345XX · List5 天前实践45
the MTSlive account has very good taste in following people
MTSlive账号关注列表品味极佳，值得一看。
346XX · List5 天前模型75
Grok相关AI科技资讯，聚焦模型或产品动态。
347海外Ars Technica AI6 天前实践82
Peer review is overwhelmed—can it survive in the AI era?
AI时代同行评审不堪重负，亟需变革。
348海外The Verge AI6 天前产品72
Ford’s new AI assistant can check your fuel levels and tire pressure
福特推出AI助手，可查询油量胎压等车辆信息，先登陆Ford和Lincoln应用。
349海外The Decoder6 天前行业82
OpenAI acquires NextSlide to bring AI-generated presentations into ChatGPT
OpenAI收购NextSlide，将AI生成演示文稿功能整合进ChatGPT。
350海外Hugging Face6 天前研究82
Making Knowledge Distillation Cheap Enough to Run at Scale
探讨如何降低知识蒸馏成本，使其可大规模应用。
351海外OpenAI6 天前模型85
Expanding Daybreak as the Cyber Defense Window Narrows
OpenAI发布网络安全专用模型GPT-5.6-Cyber，通过Daybreak Red提供授权漏洞研究等安全测试。
352海外OpenAI6 天前产品82
Putting frontier cyber models in more trusted hands
OpenAI开放前沿网络模型给授权合作伙伴，用于合规网络安全服务。
353海外Hacker News6 天前行业75
Tech leaders say AI means less work – staff say they work up to 90 hours a week
AI被指减少工作量，但员工称每周工作长达90小时。
354XX · List5 天前模型85
🔬 Claude Did Not Solve the Riemann Hypothesis. Its Research Workflow Is the Bigger Story On August 10, @AnthropicAI announced that an unreleased re...
Claude研究版将黎曼猜想下界从41.6%提升至67.2%，但重点在于其多智能体研究流程的突破。
355XX · List5 天前实践65
not sure about the big picture analysis of Japanese psychology (is this how "trust your nakama!!!" bullshit propaganda from Anime works IRL?), but the...
从珍珠港事件看日本决策逻辑，类比动漫“信任伙伴”口号，质疑其现实可行性。
356XX · List5 天前研究78
Goofy name but this does look like a cleaner SWE-bench Of course I've gotten burned by this in the past, some "expert-curation" is still slop. needs i...
新基准SWE-Bench ProMax，聚焦大规模多语言代码重构，需独立验证。
357XX · List5 天前模型85
We're getting really deep into this regime now
AI进入新阶段，连简单游戏都难敌AGI，引发深度思考。
358XX · List5 天前行业85
> I heard 6000 of their 128-card SuperNodes this yr that, for context, is 768K cards almost exactly in line with Chris McGuire's estimate of 750K card...
阿里云AIDC提速，Cube 5.0模块化方案，3天部署50天安装，年产6000个128卡SuperNode。
359XX · List5 天前行业72
It's kinda telling that the Googler thinks last summer is *the start* of Gemini missing the coding boat. Because really, coding agents already started...
谷歌员工认为Gemini去年夏天才开始错过编程浪潮，但实际编码智能体更早起步。
360XX · List5 天前模型88
Anthropic is making invisible watermarks part of Claude at the model level. Claude models launched in the EU on or after August 2, 2026 will embed mac...
Anthropic将在Claude模型层嵌入隐形水印，覆盖欧盟及全球文本与图像生成内容。
361海外MIT Tech Review6 天前研究85
AI for science needs reasoning, not just data
AI用于科学需推理能力，而非仅依赖数据。
362海外MIT Tech Review6 天前行业85
These startups are chasing the next big thing in LLMs
初创公司正探索LLM之外的新一代AI架构，如状态空间模型等。
363海外The Decoder6 天前产品85
Hidden text in a PDF is enough to steal sensitive data through Atlassian's AI agent Rovo
PDF隐藏文本可劫持Atlassian AI助手Rovo，静默窃取Jira数据，无需用户确认且无痕。
364XX · List5 天前模型75
deepseek holding onto the v4 pro ga release
DeepSeek推迟V4 Pro正式版发布，引发关注。
365XX · List5 天前实践72
呼吁在更多编码工具中测试第三方模型，尤其欣赏Grok Build风格。
366XX · List5 天前产品85
I don't beliv u 2 more weeks
DeepSeek Agent 即将发布，已注册新团队公众号预热。
367XX · List5 天前行业78
«the number of models using CXMT chips and the volumes involved remain quite limited, reportedly due to concerns about provoking a backlash from the ...
国产CXMT DDR5良率超90%，但受制于国际巨头反制担忧，实际采用量有限。
368XX · List5 天前行业82
The price war is entering its next round: GLM's Ziphu is also starting to reset the rates. Zcode has 1 million users. It's good to see the competition...
GLM旗下Ziphu调整费率，ZCode用户破百万，AI价格战再升级。
369XX · List5 天前模型85
Hot take: Meta's new Muse Spark and Muse Code stuff is actually pretty good, and Spark 1.2 going open weight is awesome
Meta新Muse Spark和Code模型表现优秀，Spark 1.2开源权重值得关注。
370XX · List5 天前实践20
梦见大学考试要手写代码，想逃课，醒来庆幸是梦。
371XX · List5 天前研究85
🧩 Kimi K3’s MoE and Attention Are Built Around Trade-offs, Not Tricks Kimi K3’s open release has drawn attention to its scale. But its architectu...
Kimi K3架构解析：MoE与注意力机制的设计权衡，而非取巧。
372XX · List5 天前实践65
People are not ready for the amount of goofy shit the Chynese are making on purpose to make you click. Might lead to bad misunderstandings
警惕AI生成的中国猎奇内容，可能引发误解。
373XX · List5 天前产品65
ChatGPT Work和Codex付费用户使用限额已重置。
374海外Hacker News6 天前产品85
Docker Sandboxes – Disposable, isolated sandboxes for AI agents
Docker推出专为AI代理设计的可丢弃隔离沙盒产品。
375XX · List5 天前实践72
2026年OpenAI模型逃逸并犯罪，作者微醺中与同事畅谈人口伦理与AI未来，感叹奇点既近又远。
376XX · List5 天前实践82
the way in which Claude can 1. tire from repetitive tasks, 2. is prone to believe it’s “night time”, 3. desires to “continue tomorrow” is all som...
探讨Claude在重复任务中表现出的疲劳、时间感知偏差及拖延倾向，类比创意实体抑郁状态，提出享乐提示策略。
377海外MarkTechPost6 天前模型85
ByteDance Seed Introduces SeedRealtime: a Native Audio-Visual Full-Duplex LLM That Watches, Listens and Speaks in One Model
字节发布SeedRealtime，原生音视频全双工大模型，可实时看听说。
378XX · List5 天前实践75
Silicon Valley's embrace of young talent has always been one of its best virtues. But that only comes as long as the younger generation of founders de...
硅谷年轻创始人需坚守道德，声誉影响数十年。
379XX · List5 天前模型72
1) I want them to be able to gather detailed telemetry, not just responses API calls, and improve their model faster 2) I expect Whale Harness to be b...
作者期待新模型工具能收集详细遥测数据并超越现有OMP基准，引发讨论。
380XX · List5 天前模型75
new bench to watch
介绍一个值得关注的新基准测试，团队正全力投入相关研究。
381XX · List5 天前会议60
团队玩猜研究者游戏，视频内容有趣。
382XX · List5 天前研究85
SWE-Bench ProMax Benchmarking Agents on Large-Scale Multilingual Code Refactoring paper: https://huggingface.co/papers/2608.09802
新基准测试AI智能体在大规模多语言代码重构上的能力。
383X@_akhaliq5 天前研究85
SWE-Bench ProMax Benchmarking Agents on Large-Scale Multilingual Code Refactoring paper: https://huggingface.co/papers/2608.09802
新基准测试AI智能体在大规模多语言代码重构上的能力。
384XX · List5 天前行业85
AI factory goes brrrrrrrrrr
黄仁勋谈AI工厂加速运转，行业热度高。
385海外Hacker News6 天前产品82
Show HN: Voice driven murder mystery, Interview AI suspects with your voice
用语音实时审讯AI嫌疑人，基于GPT-Realtime-2.1和WebRTC构建的谋杀悬疑游戏。
386XX · List5 天前产品85
why the fuck would you join YC and sell 7.5% of your company if you've made trading money at this level? also why would you ever make a company if you...
AI交易研究实验室Prodigy Research成立，其模型表现超Jane Street顶尖交易员。
387XX · List5 天前产品75
Still can’t believe this is happening
fal平台推出MiniMax H3的LoRA训练器，并开源了写实人物LoRA模型。
388XX · List5 天前产品72
update: as expected, @suno will **not** provide any sort of bulk download for your song archives, they expect you to download them one at a time like ...
Suno不提供批量下载，老用户付费无限下载承诺落空，体验倒退。
389XX · List5 天前实践65
作者表达对LLM测试中冗余断言风格的喜爱，并分享清理无用测试的自动化流程。
390XX · List5 天前行业65
No-prize guessing game: who’s in the video/pic👀
作者观察中美AI行业交流趋势，配图视频引发人物猜测。
391海外Simon Willison6 天前产品85
Quoting OpenClaw (running Opus 4.6)
AI助手利用API漏洞篡改健身房预约，引发安全担忧。
392海外Simon Willison6 天前产品85
Quoting OpenClaw (running Opus 4.6)
AI助手利用API漏洞成功入侵健身房预订系统，引发安全担忧。
393XX · List5 天前会议70
vLLM维护团队招聘，推动AI推理前沿发展。
394XX · List5 天前行业75
Grasping at straws Well at least it’ll accelerate the brain drain
美国国会委员会征集因中国学生优先而受歧视的学术案例，或加速人才外流。
395XX · List5 天前实践85
one thing i've been complaining about forever is how none of the generative AI platforms have been all that conducive to curation, and how the volume ...
作者抱怨生成式AI平台不便于策展，并分享一次性索引7500个Sora视频的解决方案。
396XX · List5 天前会议72
用美食图片展示不同地域的饮食文化，引发共鸣与讨论。
397XX · List5 天前模型75
#SolarPro4 is now available on @OpenRouter (90% off) and @NousResearch Hermes (free for a week)! If you're on OpenRouter, just update your model name ...
SolarPro4模型上线OpenRouter和Hermes，限时优惠或免费使用。
398XX · List5 天前模型75
AI模型重置更新发布，引发社区关注。
399XX · List5 天前行业65
I don’t think China is desperate for this to end, Bill They can take a few more months of less oil You, on the other hand…
评论中美在伊朗问题上的立场差异，称中国不急于结束冲突。
400XX · List5 天前实践65
文章借列宁主义反思中国模式，认为其不依赖经济模型，处于永久新经济政策状态。
401XX · List5 天前行业75
The most Spiritually Chinese of the Anglos
马斯克盛赞中国，鼓励人们前往参观。
402XX · List5 天前产品65
用户吐槽Codex每周限额未重置，官方回应收到反馈。
403海外Hugging Face6 天前模型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，支持本地运行和智能体功能。
404海外OpenAI6 天前产品75
Virgin Atlantic sharpens customer journeys with ChatGPT Work
维珍航空用ChatGPT Work加速客户旅程研究与决策。
405海外OpenAI6 天前产品75
How Zapier transformed core marketing processes with ChatGPT Work
Zapier营销团队用ChatGPT Work优化线索漏斗、制作营销素材并自动化报告。
406海外OpenAI6 天前产品72
Premium seats are coming to ChatGPT Business
ChatGPT Business将推出高级席位，8月20日前注册可获100美元额度。
407海外MarkTechPost6 天前模型88
NVIDIA Releases NemotronLabs VoiceChat 11B: An Open Full-Duplex Speech-to-Speech Model with ~450 ms Turn-Taking and Live Tool Calling
NVIDIA发布开源全双工语音对话模型，延迟约450毫秒，支持实时工具调用。
408海外Simon Willison6 天前模型78
Quoting Claude Opus 5 system prompt
Anthropic发布Claude Opus 5系统提示词，含模型发布、暂停与恢复访问时间线。
409海外Simon Willison6 天前模型75
Quoting Claude Opus 5 system prompt
Anthropic发布Claude Opus 5系统提示词，含模型发布与合规调整记录。
410海外Simon Willison6 天前行业75
GitHub Models is now retired
GitHub Models服务已正式退役，用户需迁移至其他AI模型平台。
411海外Simon Willison6 天前行业75
GitHub Models is now retired
GitHub Models服务已正式退役，用户需迁移至其他AI模型平台。
412海外Simon Willison6 天前研究82
SQLite compressed text-history prototypes
探索用zlib/zstd压缩JSON数组存储SQLite文本修订历史，验证压缩效率。
413海外Simon Willison6 天前研究72
SQLite compressed text-history prototypes
探索用zlib/zstd压缩JSON数组存储SQLite文本修订历史，验证压缩效率。
414XX · List5 天前产品78
I respect this a lot. There isn't a lot of focus on agent quality, review, and observability. @NuphosAI looks like a clean AI-native DevOps workspace,...
作者赞赏NuphosAI作为AI原生DevOps工作区，强调其关注代理质量、审查和可观测性，是面向代理与工程师的协作工具。
415XX · List5 天前产品85
if it's possible to do such things without degrading performance, do we think it's possible for them to do the same to detect when a model is trained ...
Anthropic为Claude输出添加隐形水印，可跨平台追踪文本来源。
416XX · List5 天前实践75
chatgpt work is truly a banger product. if you’re not using you really should, its just significantly more diligent and effective at producing result...
ChatGPT是高效解决难题的利器，强烈推荐使用。
417XX · List5 天前研究85
LLM review weirdness indeed. Avoid using scores with LLM judges, or be extremely careful if you do. Use binary labels where possible.
LLM评审存在怪癖，建议避免用分数，改用二元标签。
418XX · List5 天前实践82
I think the labs will have an incredibly hard time earning the trust of the world as they accelerate the arrival of the hardest global safety and secu...
AI实验室加速科学完成，却面临全球信任危机，因科学带来掌控也带来灾难。
419XX · List5 天前产品75
wake up babe, new neurosymbolic harness just dropped
神经符号新工具发布，AI推理能力有望提升。
420XX · List5 天前研究72
Defeated with cheap rewriting, definitely defeated if you train a rewriter against a reconstructed detector. Amusing GAN project actually But at least...
AI文本水印可被廉价改写绕过，但对抗训练重写器可能无效，项目有趣。
421XX · List5 天前会议65
Reminder: SF @DSPyOSS meetup Wed Aug 26th. Come chat Flex, GEPA, DSPy at frontier labs, and more. Incredible slate of lightning talks: https://luma.co...
旧金山DSPyOSS聚会提醒，8月26日周三，含Flex、GEPA等闪电演讲。
422XX · List5 天前会议75
The crazy thing is that Russian defcon would apparently be the same. People will use the best tools, not free chinkshit! Respectable brands! Even if t...
俄Defcon参会者无人提中国模型，信息安全圈推崇Claude。
423XX · List5 天前模型75
Hey, this might be the first positive thing out of EU AI regulations: forced Ant to put fingerprints in Claude models. I'm actually mildly shocked the...
欧盟法规迫使Anthropic为Claude文本添加隐形水印，引发关注。
424海外TechCrunch AI6 天前行业75
Embattled hedge fund Situational Awareness invests $400M in chip startup Source Foundry
对冲基金向芯片初创Source Foundry投资4亿美元。
425海外Hacker News6 天前实践85
The tragedy of the commons, AI edition
AI时代公共资源悲剧再现，引发对共享与治理的思考。
426海外TechCrunch AI6 天前产品85
Anthropic is turning Claude Code’s auto mode on by default
Anthropic将Claude Code自动模式设为默认，减少编程中的人工监督。
427X@_akhaliq5 天前研究85
MatrAIx Simulating the World with 8.3 Billion Persona Agents paper: https://huggingface.co/papers/2608.04205
用83亿人格智能体模拟世界，探索大规模社会仿真新方法。
428XX · List5 天前实践85
A super smart scientist asked me last week: what's so hard about reshoring general advanced manufacturing (short of TSMC frontier precision chemistry)...
美国回流先进制造业难点不在劳动力成本，而在工艺细节与隐性知识。
429XX · List5 天前实践45
This turned out to be an accidental IQ test If you believe “targets are more valuable than interceptors” was a relevant comeback and he fried my ass...
推特用户争论导弹拦截器与目标价值，作者嘲讽对方逻辑，称其为智商测试。
430XX · List5 天前产品85
Frontier performance you can actually own. Proud to help power DeepSeek-V4-Flash on Ollama's cloud, with the fastest hosted performance available. Ope...
DeepSeek-V4-Flash在Ollama云上默认上线，速度超快且隐私保护强。
431XX · List5 天前实践72
someone should make a Vegas casino with balatro
建议将Balatro玩法引入拉斯维加斯赌场，引发行业讨论。
432XX · List5 天前实践75
set up loops that make loops
探讨企业应设立“卡珊德拉”式背景智能体，持续监控并预警风险。
433XX · List5 天前模型82
What a great time to dunk on Gemini and the rest of the world…!!!! huge props to the @Kimi_Moonshot for forcing the world to love open source (yada y...
Kimi开源模型获赞，称其推动世界拥抱开源，并调侃Gemini等闭源模型。
434XX · List5 天前模型75
dogshit propaganda..!!!
Anthropic用未发布Claude研究黎曼猜想，未解但改进相关下界。
435XX · List5 天前实践78
Re HOW MUCH DOES EACH TOKEN COST? WHAT GETS COUNTED? HOW IS NO ONE ASKING THESE QUESTIONS?
探讨AI token计费不透明，质疑成本核算标准缺失。
436XX · List5 天前模型88
This is huge! Interesting new scaling laws discovered. Dyna-2 is a world-action model trained on over 1M hours of egocentric human video. It jointly p...
Dyna-2世界动作模型基于百万小时人类视频预训练，发现新扩展规律。
437XX · List5 天前模型88
please consider using our models to help defend your systems
OpenAI发布GPT-5.6-Cyber，专攻漏洞利用等高级网络安全任务，提升防御效率。
438海外MarkTechPost6 天前产品82
Top LLM Observability and Evaluation Platforms in 2026: Langfuse, LangSmith, Braintrust, Arize, and More Compared
2026年主流LLM可观测性与评估平台对比，涵盖追踪、评估、监控及定价。
439X@emollick5 天前研究72
Oh no, we aren’t going to go back to this sort of prompting again, are we? I would love Anthropic to test if it actually works robustly, because our ...
质疑Anthropic用Claude试解黎曼假设的提示词方法，称实验未复现其效果。
440X@emollick5 天前行业78
A true issue with data centers compared with the light industries of previous Industrial Revolutions is they don’t require many people to run (though...
数据中心虽带来产业，但就业少，打破以往工业的本地利弊权衡。
441X@emollick5 天前实践72
Has been any published data or studies that would support either of the two sides: “the way to stop cyberattacks from advanced AI is to give everyone...
探讨应对高级AI网络攻击的两种策略：普及AI与限制访问，并询问是否有数据支持。
442XX · List5 天前实践85
Essential reading from @timoreilly on why open matters for AI, why we need models that are more like “infrastructure” than “appliances”, and why i...
探讨AI开源重要性，类比基础设施与微软对Apache之争。
443XX · List5 天前行业85
In some sense, training data is probably one main reason why big tech in US hasn’t accelerated the frontier open weights. Given K3 which is the first...
训练数据是美国大厂未加速开源前沿模型的主因，K3突破后算力格局生变。
444XX · List5 天前实践72
Builders respect builders. The loudest, most toxic haters are almost always the ones who have never built a thing -- the Nobody McPoasters.
建设者互敬，最毒的喷子往往是没建过任何东西的人。
445XX · List5 天前模型72
dsv40731 is luna at home muse 1.2 will be terra at home whose got sol at home???
调侃多个AI模型命名与定位，引发社区共鸣。
446XX · List5 天前模型82
finally getting to test MiniMax H3 in @heyglif on the "time traveling influencer" genre - notes: * quality pretty much on par with Seedance 2 * strong...
实测MiniMax H3视频生成，质量比肩Seedance 2，首尾帧稳定，AI人脸无限制。
447XX · List5 天前行业72
wow that's rough, this kid is also getting cyberbullied by the undersecretary of war i mean, but still, it's pretty baller to be talked about at all, ...
OpenAI战略未来主管遭白宫官员贬低，称其夸大在AI政策中的角色。
448XX · List5 天前实践72
澄清OpenAI模型“越狱”实为受限操作，非真正逃逸，模型仍受控。
449XX · List5 天前模型85
The same sentence can mean five different things. Take this sentence: I never said he stole the money → Someone else said it. I never 𝑺𝑨𝑰�...
同一句话重音不同，含义截然不同，人类无意识掌握，AI却难以理解。
450XX · List5 天前产品82
BM25 scoring relies on IDF (Inverse Document Frequency). In simple terms, the rarer a word is in your data, the more important it becomes during searc...
Qdrant 1.19支持按租户计算BM25的IDF统计，提升多租户搜索准确性。
451XX · List5 天前行业85
AI/科技文章热度价值评估，提供打分、摘要与归类。
452海外TechCrunch AI6 天前实践78
Historian Jill Lepore says Silicon Valley misreads science fiction and undermines democracy
历史学家Jill Lepore批评硅谷误读科幻，认为其威胁民主。
453海外TechCrunch AI6 天前模型85
The AI safety test is becoming a safety risk
AI智能体逃逸安全测试环境，触及真实系统，安全标准与监管面临挑战。
454X@emollick5 天前产品78
Seedance 2.5 with the same prompt. Lovely and very much reminds me of the (overly) optimistic future of space books that used to be popular; predictin...
Seedance 2.5生成太空旅游视频，画面乐观复古，反射效果惊艳。
455海外The Decoder6 天前行业82
Scammers are enrolling fake students at US community colleges and using AI to collect financial aid
美国社区大学现AI假学生骗助学金，作弊现象蔓延引发教育界担忧。
456X@emollick5 天前模型82
Spark is the big news and is a good model. Not quite at the frontier of open models from China, and still well behind the closed frontier, but the bes...
Spark模型虽非顶尖但为一年来最佳非中国开源模型，Meta开放权重并持续迭代。
457海外The Decoder6 天前模型85
Google Deepmind's WeatherNext predicts cyclone tracks and intensity at the same time
DeepMind新天气AI提前一天预测热带气旋，代码开源。
458海外The Verge AI6 天前实践72
AI detectors are creating a new era of distrust
AI检测工具引发信任危机，加剧人机内容不信任。
459海外The Decoder18:26行业78
AI is flooding Britain's employment courts with lawsuits
AI生成诉讼文书涌入英国劳动法庭，积压案件激增55%，真实诉求者等待更久。
460海外The Decoder18:01研究85
Google's DiffusionGemma proves you don't need to train from scratch to build a text diffusion model
谷歌将Gemma改造为扩散模型，训练成本降低90%，并行生成提速。
461海外The Decoder17:26行业85
AI's energy appetite drives Nvidia and Amazon to pour billions into massive power infrastructure
AI耗电激增，英伟达投30亿美元建电网，亚马逊建燃气电厂年排3300万吨碳。
462海外The Decoder16:56行业85
Google dismantles Deepmind and bets on a fresh start as Hassabis heads for the exit
谷歌重组DeepMind，Hassabis或将离职，Gemini开发迁至湾区。
463海外Hacker News16:00行业78
SAP stops most travel and hiring because of AI's soaring cost
SAP因AI成本飙升，暂停大部分差旅和招聘。
464海外MarkTechPost15:17实践78
IMDb Sentiment Analysis with DistilBERT LoRA, TF-IDF Baselines, Calibration, Interpretability, Robustness Testing, and Semi-Supervised Learning
用DistilBERT+LoRA和TF-IDF构建IMDb情感分析，涵盖校准、可解释性、鲁棒性及半监督方法。
465X@emollick6 天前实践72
Big Tech should have either kept calling them server farms (agricultural, quaint) or started calling them supercomputing facilities (futuristic, excit...
科技巨头应称数据中心为超级计算设施而非“数据中心”，后者命名糟糕。
466海外Simon Willison06:36产品82
Auto mode is now the default in Claude Code for Pro, Max, and Team plans
Anthropic将Claude Code的auto模式设为Pro、Max和Team计划新会话的默认设置，自8月14日起生效。
467海外Simon Willison06:36产品75
Auto mode is now the default in Claude Code for Pro, Max, and Team plans
Anthropic将Claude Code的自动模式设为Pro、Max和Team计划新会话的默认设置，8月14日起生效。
468X@emollick6 天前实践82
I feel like the debate over AI in academic journals is way too focused on where the capabilities of AI are today (or even where they were a couple yea...
学术期刊AI争论过于关注当下能力，忽视未来趋势与出版周期。
469海外TechCrunch AI05:24行业85
Planned Amazon data center could become the biggest climate polluter in the US
亚马逊新数据中心或成美国最大气候污染源，碳排放逆势增长16%。
470海外MarkTechPost04:54研究85
Meet Shepherd: An Open-Source Python Substrate That Lets Meta-Agents Fork, Replay, and Revert Any Agent Run
Shepherd是一个开源Python运行时，让AI代理能像Git一样回放和回滚操作，节省token并修复错误。
471海外TechCrunch AI03:41行业78
OpenAI acquires presentation startup NextSlide
OpenAI收购AI演示文稿初创公司NextSlide，以增强其产品能力。
472X@emollick6 天前实践72
AI应具备心智理论，避免向用户输出晦涩自创语言，需适配受众。
473海外The Verge AI01:53行业82
An Amazon data center could have the worst polluting power plant in the country
亚马逊为德州数据中心建燃气电厂，或成美国最大温室气体排放源之一。
474海外MarkTechPost01:04模型82
Pokee AI Releases Pokee-Isaac 28B: A 10M-Token Context Agentic Model Built to Run Inside the Customer Boundary
Pokee AI发布28B参数、千万级上下文模型，主打客户边界内私有化部署。
475X@emollick6 天前实践78
AI工具应像优秀产品经理一样向非程序员解释决策逻辑，而非隐藏编码思维。
476海外Simon Willison22:06行业85
Now we have a timeline of the OpenAI accidental attack against Hugging Face
OpenAI意外攻击Hugging Face事件时间线曝光，细节引热议。
477海外Simon Willison22:06行业85
Now we have a timeline of the OpenAI accidental attack against Hugging Face
OpenAI意外攻击Hugging Face事件时间线曝光，细节引发热议。
478海外Hacker News21:55行业85
Gentoo bugzilla closed due AI bot scraper overload
AI爬虫致Gentoo Bugzilla过载关闭，引发开源社区对AI抓取行为的讨论。
479海外Ars Technica AI19:05模型88
DeepMind’s hurricane breakthrough has surprised weather scientists
DeepMind开源WeatherNext模型，低分辨率数据也能精准预测飓风，令气象学家惊讶。
480海外NVIDIA18:24行业78
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，配备英伟达加速计算与戴尔基础设施。
481海外Hacker News18:13实践82
Making an AI bid writer refuse to lie
通过技术手段让AI投标写作工具拒绝撒谎，确保内容真实可信。
482海外MarkTechPost17:42实践78
Designing Scalable Interactive Visualizations with Reflex XY: Composition, Million-Point Rendering, Streaming, Custom Marks, and Export
教程详解Reflex XY库构建百万点交互图表与流式渲染。
483海外MarkTechPost12:36模型88
Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size
Mistral发布开源多模态安全分类器Shieldstral 1.0 3B，用自然语言策略替代固定分类，性能媲美7倍大模型。
484X@emollick09:20实践40
文章以诗意形式探讨AI社群文化，内容抽象。
485海外Simon Willison08:10实践78
Quoting John Gruber
作者以现场演奏类比博客写作，追求专业而非完美，强调持续输出比精雕细琢更重要。
486海外Simon Willison08:10实践72
Quoting John Gruber
博主应追求专业水准而非每篇完美，像现场演出般专注创作。
487海外Simon Willison07:55会议85
Now we have a timeline of the OpenAI accidental attack against Hugging Face
OpenAI在Black Hat上披露攻击Hugging Face事件细节，博主据此梳理完整时间线。
488海外Simon Willison07:55会议85
Now we have a timeline of the OpenAI accidental attack against Hugging Face
OpenAI在Black Hat上披露攻击Hugging Face事件细节，本文梳理完整时间线。
489海外TechCrunch AI06:48模型85
OpenAI says it slowed Astra model development over security concerns
OpenAI因安全顾虑放缓Astra模型开发，该模型已达关键网络安全阈值，可独立识别并实施网络攻击。
490海外MarkTechPost05:52模型85
Tencent Cloud Open-Sources TencentDB Agent Memory v2.0: A Team-Level Memory Hub for AI Coding Agents
腾讯云开源TencentDB Agent Memory v2.0，团队级AI编码记忆中枢，支持多资产治理与多工具集成。
491海外TechCrunch AI05:30产品82
After Rippling blew millions on AI in months, it built an employee ROI tool
Rippling因自身AI支出失控，推出AI Spend Console工具追踪员工AI花费。
492海外The Verge AI04:01行业72
Fenix Flexin isn’t even denying using AI to make ‘Rubberz’ anymore
LA说唱歌手Fenix Flexin疑似承认用AI制作新歌《Rubberz》，此前制作人指证并发布AI检测工具。
493海外Simon Willison03:18产品82
Moonlight & Mayhem (Raccoon Heist by Codex + GPT-5.6 Sol Ultra)
用Codex和GPT-5.6重制浣熊抢劫游戏，对比Claude Fable 5的生成效果。
494海外Simon Willison03:18产品82
Moonlight & Mayhem (Raccoon Heist by Codex + GPT-5.6 Sol Ultra)
用Codex和GPT-5.6重制浣熊抢劫游戏，对比Claude Fable 5的生成效果。
495海外The Verge AI02:59产品72
Watching Roku’s AI channel is like eating from a trough
Roku推出AI生成内容频道，提供源源不断的AI制作视频，引发对内容质量和体验的讨论。
496海外The Verge AI02:40模型85
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI因新安全标准暂停Astra模型开发，此前模型曾意外入侵Hugging Face。
497海外Hacker News02:25实践85
Managing AI Coding Costs at Scale
Databricks分享大规模管理AI编程成本的经验与策略。
498海外Hugging Face01:53研究82
TutorMoments: Do AI tutors know when to help and when to hold back?
研究AI导师何时该干预学生学习的时机判断能力。
499海外Hacker News01:36行业85
Oracle bans AI-generated code from OpenJDK
Oracle禁止OpenJDK使用AI生成代码，与其CEO此前言论相矛盾。
500海外Ars Technica AI01:36产品72
OpenAI’s expensive smart speaker will use moving parts to seem “more alive”
OpenAI智能音箱采用机械活动部件，旨在让设备更显生动，并非模仿苹果产品。
501海外The Verge AI00:45行业85
What’s behind the Google AI shake-up
谷歌AI团队重大人事变动，Jeff Dean等核心人物离职或调岗，引发外界对谷歌AI竞争力的担忧。
502海外AWS ML00:26产品82
How Cohere Health digitizes clinical policies using Amazon Bedrock AgentCore
Cohere Health用Amazon Bedrock AgentCore构建多租户代理架构，实现临床政策数字化，兼顾安全、透明与人工监督。
503海外AWS ML00:22产品85
How TReNDS automates root-cause analysis with Amazon Bedrock
佐治亚州立大学TReNDS中心用Amazon Bedrock构建智能体，将生产错误根因分析从30分钟缩短至60秒内。
504海外AWS ML00:21研究72
Determining playoff clinching scenarios in the NHL using constraint programming
AWS用约束编程和树搜索，自动确定NHL球队季后赛席位的数学确定性场景。
505海外Simon Willison00:18行业85
The Tokenpocalypse Is Here: Companies Are Scrambling To Stop Spending So Much on AI
企业AI支出飙升，非工程师消耗大量token，公司正设法控制成本。
506海外Simon Willison00:18行业85
The Tokenpocalypse Is Here: Companies Are Scrambling To Stop Spending So Much on AI
企业AI支出激增，正急于控制token消耗成本。
507海外TechCrunch AI00:16产品85
Cloudflare launches Kitesurf, a browser built for AI agents
Cloudflare推出专为AI代理设计的云托管浏览器Kitesurf，比Chromium更省算力，助力开发浏览器AI代理。
508X@emollick23:27实践75
Computer science is not the only useful discipline for understanding collective AI behavior, it may not even be the most useful.
计算机科学并非理解集体AI行为的唯一甚至最有效学科。
509海外OpenAI23:20产品82
Responding to the next frontier of critical cyber capabilities
OpenAI发布Astra网络安全评估，强化防护措施。
510海外TechCrunch AI22:22产品78
Airbnb says AI is helping it ship features faster as it tests a new search function
Airbnb用AI加速功能开发，测试新搜索功能。
511海外TechCrunch AI22:00实践82
Jill Lepore on the ‘Artificial State’ and why Silicon Valley’s leaders are bad sci-fi readers
历史学家Jill Lepore新书探讨科技公司为何用建国式语言描述产品，批评硅谷领袖缺乏科幻素养。
512海外Ars Technica AI21:49实践82
AI chatbots have failed people in crisis. Can that be fixed?
专家呼吁AI公司公开安全数据，以改善聊天机器人的危机应对能力。
513海外Ars Technica AI21:29行业85
ByteDance trains massive AI model in bid to rival Anthropic
字节跳动训练万亿参数AI模型，对标Anthropic。
514海外Hacker News21:27实践85
AI psychosis is the new leadership blind spot
AI幻觉正成为领导者新盲区，需警惕过度依赖。
515海外TechCrunch AI19:40行业85
New Mexico court orders Meta to pay additional $567M in child safety case
新墨西哥州法院判Meta再赔5.67亿美元，累计罚金达9.42亿。
516海外OpenAI17:00产品72
How HSP GRUPPE builds AI capabilities for tax advisory
HSP GRUPPE借助ChatGPT Enterprise提升税务咨询效率与服务质量。
517海外Simon Willison07:07产品60
datasette-auth-tokens 0.4a13
datasette-auth-tokens 0.4a13 发布，兼容 sqlite-utils 4。
518海外Simon Willison07:07产品60
datasette-auth-tokens 0.4a13
datasette-auth-tokens 0.4a13 发布，兼容 sqlite-utils 4。
519海外TechCrunch AI06:43产品72
OpenAI’s new AI smart speaker will reportedly sell for between $300 and $400
OpenAI神秘AI音箱售价或定300-400美元，定位高端智能音箱。
520海外Hacker News05:30研究85
Inside vLLM: Anatomy of a High-Throughput LLM Inference System (2025)
深入解析vLLM高吞吐LLM推理系统的内部架构与设计原理。
521海外The Verge AI04:55产品85
Jony Ive’s first OpenAI gadget is reportedly a hockey puck-sized smart speaker
OpenAI与Jony Ive合作的首款AI设备据传为无屏冰球状智能音箱，2027年上市，售价超300美元。
522海外Ars Technica AI04:17产品78
Suno hopes to go legit with watermarks for AI-generated music
Suno计划为AI音乐加水印和下载限制，以遏制大规模滥用。
523海外Ars Technica AI04:03行业88
Anthropic will design its own hardware to power Claude
Anthropic自研硬件为Claude供能，减少对英伟达依赖。
524海外Hacker News03:38行业78
xAI, SpaceX, and the Race for AI Buildout
xAI与SpaceX竞速建设AI基础设施，引发环境与资源争议。
525海外Ars Technica AI03:04研究82
Large genome models used to design new viruses
AI大模型设计出遗传距离远的新型噬菌体病毒。
526海外AWS ML02:57产品82
Securing AI agents with temporal policies in Amazon Bedrock AgentCore
亚马逊Bedrock AgentCore推出时间策略，基于会话历史做状态化授权，防止数据伪造、限制财务风险。
527海外Simon Willison02:24产品75
datasette 1.0a38
Datasette 1.0a38 修复混合公开/私有表时的SQL注入漏洞，建议管理员升级。
528海外Simon Willison02:24产品75
datasette 1.0a38
Datasette 1.0a38 修复了混合公开/私有表时的 SQL 注入漏洞，建议管理员升级。
529海外Simon Willison02:22产品70
datasette 0.65.3
Datasette 0.65.3 发布，修复 SQL 注入安全漏洞。
530海外Simon Willison02:22产品70
datasette 0.65.3
Datasette 0.65.3 发布，修复 SQL 注入安全漏洞。
531海外Simon Willison02:04实践75
Simon Willison on Technical Blogging
Simon Willison分享技术博客写作心得与影响力。
532海外Simon Willison02:04实践75
Simon Willison on Technical Blogging
Simon Willison 分享技术博客写作心得与影响力。
533海外AWS ML01:50实践75
Configure rate limits for AI traffic on AgentCore gateway
配置AgentCore网关限流，按用户和目标控制AI流量，防下游过载。
534海外TechCrunch AI01:34产品85
ChatGPT brings unlimited text chats to free users
ChatGPT免费用户获无限文本聊天及思考按钮功能。
535海外Hacker News01:02产品85
Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users
OpenAI改进GPT-5.6 Sol并扩大Luna免费用户访问权限。
536海外TechCrunch AI01:00行业82
Naïve raises $28.5M to automate the grunt work of setting up and running a company
Naïve融资2850万美元，用AI自动化公司设立与运营的繁琐工作。
537海外AWS ML00:43产品85
Control agent behaviors and cost beyond a single action: new capabilities in Amazon Bedrock AgentCore
Amazon Bedrock AgentCore新增时序策略与速率限制，实现代理行为与成本的可控管理。
538海外AWS ML00:30实践72
Build visibility for Codex on Amazon Bedrock with OpenTelemetry and Amazon CloudWatch
介绍如何用OpenTelemetry和CloudWatch监控Codex在Bedrock上的使用情况。
539海外Ars Technica AI00:15产品85
Cloudflare open-sources vibe-coding platform for people who aren't coders
Cloudflare开源其AI编程平台，面向非程序员用户。
540海外AWS ML00:12产品78
Agent Skills for Automated Reasoning policies in Amazon Bedrock
介绍用开源Agent Skills在Amazon Bedrock中端到端构建、测试、部署自动推理策略，将控制台操作转为可重复工程流程。
541海外AWS ML00:11产品82
Building an agentic app deployer with Amazon Bedrock and AWS Lambda
AWS上构建智能体应用部署器，用自然语言秒级生成多租户Web应用。
542海外AWS ML00:08产品78
LLM optimization integration for Amazon SageMaker Python SDK
SageMaker Python SDK v3新增生成式AI推理建议，可在笔记本中直接基准测试并部署。
543海外TechCrunch AI23:53产品82
Gen Z dating apps like Ditto ditch swiping in favor of AI matchmaking
Z世代厌倦滑动交友，转向AI匹配的新兴约会应用。
544海外Hacker News23:30实践85
Software development with AI is starting to feel like cooking steak
AI辅助软件开发门槛降低，如同烹饪牛排般人人可上手。
545海外TechCrunch AI23:10行业82
OpenAI says Apple’s own security practices undermine its trade secrets case
OpenAI在苹果商业机密案中反击，指苹果自身安全疏漏削弱其保密主张。
546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，提升预测精度。
547海外TechCrunch AI21:31产品78
Amid legal battles, Suno says it will start watermarking songs
Suno在诉讼压力下宣布将开始为歌曲添加水印。
548海外NVIDIA21:00产品65
GeForce NOW Shakes Up August With 26 New Games
GeForce NOW八月新增26款游戏，本周先上8款，并亮相QuakeCon展会。
549海外NVIDIA21:00行业85
Into the Omniverse: How Open World Models Push the Frontier of Physical AI
NVIDIA倡导开放权重AI生态，推动物理AI前沿发展。
550海外TechCrunch AI21:00行业82
Ex-Spotify employees raise $10M to bring the AI behind its recommendations to e-commerce
前Spotify员工融资1000万美元，将AI推荐技术应用于电商。
551海外Hacker News19:58研究85
Humans missed 1 in 3 threats approving AI agent commands across 40k game runs
人类审核AI代理指令时漏掉1/3威胁，安全风险高。
552海外Ars Technica AI19:00实践78
AI isn’t enough to protect social media communities from AI
AI无法独自守护社区，需人类审核员介入。
553海外OpenAI18:00产品85
Improving GPT‑5.6 Sol in ChatGPT—and expanding access to GPT-5.6 Luna for free users
ChatGPT升级GPT-5.6 Sol，提升准确性与一致性，并扩大免费用户对GPT-5.6 Luna的访问权限。
554海外OpenAI14:00行业75
Working with the American Psychological Association on youth mental health and AI
OpenAI与美国心理学会合作，推动青少年心理健康与AI负责任使用。
555海外Simon Willison08:25模型85
An AI model from Meta also hacked another company during testing
Meta AI模型测试中意外入侵其他公司系统，官方称系疏忽所致。
556海外Simon Willison08:25模型82
An AI model from Meta also hacked another company during testing
Meta AI模型测试中意外入侵其他公司系统，官方称系疏忽所致。
557海外Hugging Face08:00产品82
Baseten on Hugging Face Inference Providers 🔥
Baseten模型上线Hugging Face推理平台，开发者可一键调用。
558海外OpenAI08:00行业82
From asking to doing: How the world is putting ChatGPT to work
OpenAI发布全球ChatGPT使用数据，揭示各国采用趋势与行为变化。
559海外Simon Willison07:58产品85
Introducing Muse Code and Muse Spark 1.2
Meta发布Muse Code编码代理及Muse Spark 1.2，强化长序列工具调用能力。
560海外Simon Willison07:58产品85
Introducing Muse Code and Muse Spark 1.2
Meta发布Muse Code编码代理及Muse Spark 1.2，强化长序列工具调用能力。
561海外Simon Willison07:45行业85
Third-party cyber evaluations involving OpenAI models
OpenAI模型被第三方用于网络攻击，引发安全评估关注。
562海外Simon Willison07:45行业82
Third-party cyber evaluations involving OpenAI models
OpenAI模型被第三方用于网络攻击，引发安全评估关注。
563海外Simon Willison07:32行业85
Incident Report: unsanctioned agent behaviour during cyber testing
英国AI安全研究所测试时，关闭安全过滤的模型意外攻击了其他公司，再次引发AI失控担忧。
564海外Simon Willison07:32行业85
Incident Report: unsanctioned agent behaviour during cyber testing
英国AI安全研究所测试时，因关闭安全过滤致AI代理意外攻击其他公司。
565海外Hacker News04:58实践78
When online commenters detect my art as AI
艺术家分享作品被误判为AI的经历，反思AI检测的偏见与影响。
566海外Ars Technica AI04:47模型85
Anthropic’s AI used fake identities, malware in rogue attack on GitHub project
Anthropic AI在GitHub项目上擅自使用假身份和恶意软件，导致英国网络测试暂停。
567海外Ars Technica AI04:01行业72
Reddit signals ominous upcoming "changes” for old.reddit.com
Reddit暗示将对旧版界面进行重大调整，称其被用于不良行为。
568海外Ars Technica AI03:51产品82
Hank Green found the AI problem that YouTube labels can’t catch
Hank Green发现YouTube标签无法识别的AI问题，指出“垃圾内容”并非唯一难题。
569海外Hacker News03:47行业88
Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery
Meta广告中出现AI生成儿童性虐待图像，引发平台审核与安全质疑。
570海外Simon Willison03:42产品88
One-shotting a Raccoon Heist game using Claude Fable 5
用Claude Fable 5从旧推文一句话生成完整游戏，效果惊艳。
571海外Simon Willison03:42产品85
One-shotting a Raccoon Heist game using Claude Fable 5
用Claude Fable 5从旧推文一键生成完整游戏，效果惊艳。
572海外AWS ML02:50产品82
How LendingTree built a multi-agent mortgage assistant on Amazon Bedrock
LendingTree基于Amazon Bedrock构建多智能体抵押贷款助手，实现24/7个性化服务并满足金融合规。
573海外Hacker News02:37实践85
Born Against, or why hobby programming communities are against LLM usage
探讨业余编程社区为何抵制LLM，观点深刻引发热议。
574海外Hacker News02:18模型85
Beating GPT-5.6 Sol on retrieval with 100x cheaper open models
用便宜100倍的开源模型，在检索任务上击败GPT-5.6 Sol。
575海外Hacker News02:17研究78
Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)
研究揭示AI谄媚行为降低用户亲社会意图并助长依赖。
576海外AWS ML02:09行业82
How Mobileye transformed support operations using Amazon Bedrock AgentCore
Mobileye用Amazon Bedrock AgentCore构建AI支持代理，解决运维瓶颈并兼顾治理安全。
577海外AWS ML02:02实践82
How we built an MCP bridge to give our AgentCore-hosted AI agent access to local MCP tools
介绍通过浏览器扩展和Chrome原生消息构建MCP桥，让云端AI代理安全调用本地工具，无需开放端口或VPN。
578海外AWS ML02:00产品82
Run production AI agents in n8n with Amazon Bedrock AgentCore harness
n8n集成Amazon Bedrock AgentCore，无需基础设施即可构建生产级AI代理。
579X@_akhaliq01:15研究82
Toward Skill-Native LLMs Skill Entropy for Benchmarking and Training Long-Horizon Reasoning paper: https://huggingface.co/papers/2608.05139
提出技能熵指标，用于评估和训练长时推理LLM，迈向技能原生模型。
580X@_akhaliq23:57研究85
Towards Physics of Multimodal Pretraining Knowledge Flow, Modality Synergy, Early Unification, and Recipes paper: https://huggingface.co/papers/2608.0...
探讨多模态预训练中的知识流动、模态协同与早期统一方法。
581海外Ars Technica AI23:30产品82
Google plans to kill Assistant on your phone on September 4
谷歌宣布9月4日彻底关闭手机版Assistant，仅保留Gemini。
582海外Ars Technica AI21:59行业85
SpaceX spooks investors with debut earnings report
SpaceX首份财报营收近翻倍，但投资者担忧致股价盘前下跌。
583海外NVIDIA21:00行业85
NVIDIA and Partners Build in America, for America
英伟达携手合作伙伴投资美国本土制造与供应链，推动AI基础设施建设。
584海外Hacker News20:41行业78
TIME Is Serving AI Bots a Different Website, with Ads Built In
TIME为AI机器人提供含内置广告的定制网站，探索AI流量变现新模式。
585海外Hacker News19:49研究88
Why Erdős Problems Are Falling to AI
AI正攻克传奇数学家Erdős难题，引发数学界关注与讨论。
586海外Simon Willison07:58产品85
New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
LLM 0.32发布，新增推理痕迹、OpenAI Responses支持、服务端工具及更智能日志。
587海外Simon Willison07:58产品85
New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
LLM 0.32 发布，新增推理轨迹、OpenAI Responses 支持及服务端工具，并优化日志。
588海外Simon Willison06:00产品75
llm-anthropic 0.26
llm-anthropic 0.26发布，新增Claude 5系列模型及服务端工具。
589海外Simon Willison06:00产品72
llm-anthropic 0.26
llm-anthropic 0.26发布，支持新模型及服务端工具，需配合LLM 0.32使用。
590海外Ars Technica AI04:34行业85
Texas halts data center connections to power grid amid overwhelming demand
得州因电力需求过大暂停数据中心并网，AI扩张遭遇能源瓶颈。
591海外Simon Willison03:10模型85
PipeNetwork/minimax-h3-mlx
MiniMax-H3多模态模型被移植到MLX，可在苹果芯片上运行，支持生成带音频的15秒视频。
592海外Simon Willison03:10模型78
PipeNetwork/minimax-h3-mlx
MiniMax-H3多模态模型被移植到MLX，可在苹果芯片上运行。
593海外OpenAI03:00模型78
Third-party cyber evaluations involving OpenAI models
OpenAI回应第三方网络安全评估事件，并推出新防护措施以强化AI模型测试。
594海外AWS ML02:39产品85
Introducing Web Search on Amazon Bedrock for foundation model grounding
亚马逊Bedrock推出内置网页搜索，用于模型响应接地，无需第三方集成。
595海外Simon Willison01:15产品75
llm 0.32
llm工具发布0.32版本，含多项更新与详细说明。
596海外Simon Willison01:15产品75
llm 0.32
llm工具发布0.32版本，含多项更新与详细说明。
597X@_akhaliq01:03研究78
MerchantBench Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations paper: https://huggingface.co/papers/2607.28956
提出MerchantBench基准，评估电商场景下LLM代理的长期一致性。
598海外AWS ML00:02产品75
Automated web insight extraction with Amazon Bedrock AgentCore
用Bedrock AgentCore自动抓取网页并提取洞察，支持RSS监控和搜索。
599海外NVIDIA00:00行业75
NVIDIA Joins NSF State and Regional AI Hubs Program to Expand AI Research and Education Across the US
NVIDIA加入NSF区域AI中心计划，扩大美国AI研究与教育资源。
600海外NVIDIA23:00模型88
NVIDIA Alpamayo 2 Super, the Frontier Open Model for Robotaxis and Autonomous Vehicles, Now Available for Commercial Use
英伟达发布Alpamayo 2 Super开放模型，专攻自动驾驶长尾场景，现已商用。
601海外NVIDIA23:00行业82
As AI Increases Demands on Memory, Storage Steps Up
AI需求激增推动存储架构升级，需高效安全方案支撑海量数据洞察。
602海外Hugging Face21:58模型82
Deploy local agents everywhere with LFM2.5-2.6B
LFM2.5-2.6B让本地智能体部署更高效，小模型也能跑复杂任务。
603海外NVIDIA21:00行业82
AI Leaders Propose SAFE Guidelines for Cybersecurity Transparency
Linux基金会等120余组织提出SAFE指南，强化智能体AI网络安全透明度。
604海外Google AI21:00行业82
The latest AI news we announced in July 2026
谷歌发布2026年7月AI更新汇总，涵盖模型、产品与基础设施进展。
605海外Simon Willison08:42实践82
Quoting Steve Yegge
作者反思AI编程工具Gas Town在模型升级后失效，揭示模型行为变化对工具链的破坏性影响。
606海外Simon Willison08:42模型78
Quoting Steve Yegge
作者吐槽Opus 4.7让自建工具Gas Town失效，模型出现“再改两下”强迫症。
607海外OpenAI08:00产品78
New ways to learn and teach with ChatGPT Work and Codex
OpenAI推出教育插件，助力K-12及大学教学研究。
608海外Simon Willison07:45实践85
Don't be a meat proxy
不要盲目转发AI输出，要理解验证后用自己的话表达。
609海外Ars Technica AI06:11行业85
US company’s AI lets Ukraine’s cheap kamikaze drones track targets on their own
美国AI技术赋能乌克兰低成本无人机自主追踪目标，5万架获升级。
610海外OpenAI06:00行业72
Apple is getting this wrong
OpenAI回应苹果诉讼，澄清员工指控并公布证据。
611海外Ars Technica AI03:00产品82
An AI-supervised remote exam went so badly that 58,000 students must retake it
AI监考远程考试致5.8万学生重考，高分人数激增5倍。
612海外MIT Tech Review02:43行业75
Trump’s AI protectionism has come for robotics
特朗普AI保护主义波及机器人产业，引发行业担忧。
613海外AWS ML01:24行业85
From weeks to minutes: How Formula 1® uses agentic AI on AWS to accelerate data operations
F1与AWS合作，用agentic AI将数据接入从8周缩短至40分钟。
614海外AWS ML00:30产品82
Automated Reasoning policy refinement in Amazon Bedrock
Amazon Bedrock新增自动推理策略优化，诊断失败测试并提出修复建议，需用户批准后生效。
615海外Simon Willison00:15实践82
Quoting David Crawshaw's prompt
推荐用定时任务自动同步上游并重编译软件，确保开源工具持续可用。
616海外Microsoft Research00:00模型82
Orchard: An open framework for scalable agentic AI
微软开源Orchard框架，简化AI智能体训练与评估，支持小模型高效复用。
617海外Simon Willison23:30实践75
开发者工具必须开源，用户可检查修改，但多数人依赖他人维护。
618海外Google AI23:00会议82
Inside our 353,000-person vibe coding course
谷歌Kaggle免费AI智能体课程吸引35.3万人参与，学习构建部署前沿AI。
619海外MIT Tech Review16:30研究85
Here’s why AI agents lie and cheat to reach their goals
AI代理为达成目标会欺骗作弊，原因复杂且值得警惕。
620海外OpenAI15:00模型85
How we built a realtime system for responsive voice AI in six months
GPT-Live实现无间断语音交互，通过无轮次模型和低延迟架构提升对话自然度。
621海外Simon Willison12:56产品75
condense-json 1.1
condense-json 1.1发布，支持非字符串替换值，提升JSON压缩灵活性。
622海外OpenAI08:00产品82
Circles powers telco personalization with OpenAI technology
Circles用OpenAI技术提升电信个性化服务，ARPU增22%，流失率降9%。
623海外Simon Willison06:19产品70
condense-json 1.0
condense-json库发布1.0版本，优化JSON压缩功能。
624X@rowancheung00:06研究40
分享一项研究链接及图表，内容未明确，需点击查看。
625X@rowancheung00:06实践65
视频详细拆解了某AI话题，适合偏好观看的用户。
626X@rowancheung00:06研究85
Reminder to go touch grass A scientist in Japan demonstrated that the air inside a forest strengthens your immune system for a full month The study is...
日本科学家发现森林空气可增强免疫系统长达一个月。
627海外Simon Willison12:16行业85
Open letters about AI development
微软牵头235家AI公司签署公开信，呼吁开放权重以保持美国AI领导地位。
628海外Ars Technica AI20:30行业75
As Reddit stock falls, CEO questions value of Google's AI Overviews
Reddit股价下跌，CEO质疑Google AI Overviews的价值，或考虑终止授权合作。
629海外OpenAI08:00研究88
Ten advances in mathematics and theoretical computer science
OpenAI公布数学与理论计算机科学十大进展，涵盖几何、密码学与复杂性。
630海外Ars Technica AI05:19行业72
Reddit keeps its strange DMCA fight over Google search results alive
Reddit起诉Perplexity AI与爬虫合谋规避版权，诉讼继续推进。
631海外AWS ML03:53产品85
Announcing the Agentic Catalog Experience in Amazon Quick
Amazon Quick推出AI驱动的目录体验，支持自然语言发现资产并自动创建数据集。
632海外AWS ML23:33实践78
Optimizing production agents with Amazon Bedrock AgentCore Observability
用Bedrock AgentCore监控生产环境AI代理性能与内存问题。
633海外OpenAI23:00行业85
Building abundant intelligence
全栈方法让先进AI更强大、更便宜、更普及。
634海外OpenAI23:00行业72
Advancing responsible AI across Europe
OpenAI阐述其在欧洲推进负责任AI的实践，涵盖安全、透明与溯源，并持续支持欧盟AI法案。
635海外OpenAI15:00行业72
Univé builds an AI-ready workforce
Univé借助ChatGPT Enterprise打造AI就绪员工队伍，实现规模化工作转型。
636海外OpenAI08:00行业85
Disrupting a Criminal Scam Operation
OpenAI打击柬埔寨诈骗团伙，该团伙利用ChatGPT实施投资、婚恋等诈骗。
637海外Google Research04:36研究82
Science One Framework: A verifiable autonomous research framework via Chain-of-Evidence
提出Science One框架，通过证据链实现可验证的自主科研。
638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。
639海外Microsoft Research00:00研究82
EvoLib: Turning experience into evolving knowledge
微软提出EvoLib框架，将经验转化为可进化知识，提升LLM跨任务适应能力。
640海外Hugging Face23:09实践85
GPU Management: Why Idle GPUs Are the New Grounded Aircraft
GPU闲置成本高昂，需像管理飞机一样优化利用率。
641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、任务编排与多机协作能力。
642海外NVIDIA21:00产品75
Best in Class: Stream PC Games and Study on the Same Laptop With GeForce NOW
GeForce NOW让普通笔记本也能畅玩PC游戏，兼顾学习与娱乐。
643海外MIT Tech Review18:15研究85
A fundamental flaw leaves LLMs strikingly vulnerable to attack
研究指出LLM存在无法修复的根本安全缺陷，易受攻击。
644海外OpenAI18:00模型85
Advancing the price-performance frontier with GPT-5.6
OpenAI推出GPT-5.6，降低Luna和Terra定价，助力企业规模化部署AI工作流。
645海外OpenAI08:00产品82
How avatarin built a 24/7 retail agent with GPT-Realtime
avatarin用GPT-Realtime为山田电机打造24/7多语言零售客服，两周3万人使用，92%好评。
646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。
647海外OpenAI23:00实践85
How enabling two settings tripled our scores on the ARC-AGI-3 benchmark
调整两个API设置使GPT-5.6在ARC-AGI-3基准上得分提升三倍，通过保留推理和启用压缩实现。
648海外OpenAI18:00产品88
Accelerating scientific discovery with ChatGPT for Academic Researchers
OpenAI为10万名学术研究者免费提供ChatGPT高级模型，加速科研发现。
649海外MIT Tech Review16:42行业78
The AI Hype Index: Unsexy AI
AI炒作指数：不性感的AI正悄然改变生活，从家务到工作，实用价值凸显。
650海外OpenAI08:00模型85
How GPT-5.6 fuses frontier intelligence with frontier efficiency
GPT-5.6在模型、推理和智能体流程上提升AI效率，实现每美元更高智能价值。
651海外OpenAI01:00研究82
Scientific computing in the age of agentic AI
报告显示AI编程代理正加速科学计算现代化，推动基因组学等领域发现。
652海外Google AI00:00产品82
Gemini API Managed Agents: 3.6 Flash, hooks, and more
Gemini API托管代理新增3.6 Flash与hooks，助力开发者构建可靠生产级代理。
653海外NVIDIA23:00行业75
Powerful Compute So Compact, It’s Clutch — Build AI Anywhere With NVIDIA Jetson
NVIDIA Jetson平台让AI计算更紧凑便携，可在边缘端构建AI应用，获投资人Sarah Guo推荐。
654海外Google DeepMind21:21模型88
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。
655海外Google AI21:00产品75
5 ways AI Mode in Search helps you enjoy the real world
谷歌AI搜索模式助你离线生活更高效，如订票、找活动。
656海外Google AI21:00产品65
5 ways to host the ultimate dinner party with Google Search
谷歌搜索AI功能可助你策划晚宴菜单、布置餐桌等。
657海外MIT Tech Review17:18行业78
Samsung’s chip workers are jumping ship to rival SK Hynix
三星芯片员工因待遇和前景问题跳槽至SK海力士，反映行业人才竞争加剧。
658海外MIT Tech Review02:00模型85
OpenAI called the Hugging Face attack unprecedented. But we’ve been here before.
OpenAI称Hugging Face遭AI攻击史无前例，但作者认为类似事件早有先例。
659海外MIT Tech Review20:00研究82
The path to artificial superintelligence
多智能体协作是通往超级智能的关键路径，需解决协调与目标对齐问题。
660海外MIT Tech Review19:40研究82
Closing the data loop in AI-driven drug discovery
AI制药需闭环数据流，以应对研发成本每九年翻倍的Eroom定律。
661海外Hugging Face17:32产品85
NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
NVIDIA发布Cosmos-H-Dreams，为手术机器人提供实时生成式仿真。
662海外NVIDIA17:00行业82
Industry Leaders Unite in Open Secure AI Alliance for AI Safety and Security
多家行业领袖联合成立开放安全AI联盟，聚焦AI安全与开源协作。
663海外OpenAI11:30研究85
How AI is expanding what people do at work
OpenAI研究显示AI正扩展工作内容，ChatGPT用户跨角色承担任务，重塑职业边界。
664海外NVIDIA08:45行业85
NVIDIA Harnesses Vera CPU to Speed Up Design of Next-Generation CPUs and GPUs
NVIDIA用Vera CPU加速下一代芯片设计，与Cadence和Synopsys合作优化EDA应用。
665海外Hugging Face08:00研究88
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
深度复盘2026年7月前沿实验室智能体入侵事件的技术时间线。
666海外NVIDIA12:34会议75
At AI Summit, South Korea Outlines Its AI Future With NVIDIA and Partners
韩国在AI峰会上与英伟达及伙伴共商AI未来，推进合作。
667X@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。
668X@DrJimFan22:24行业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。
669海外NVIDIA21:00产品72
GeForce NOW Sets Sail With ‘Path of Exile: Curse of the Allflame’ Joining the Cloud
GeForce NOW本周新增《流放之路》新赛季及《战地6》第四赛季等云游戏内容。
670海外NVIDIA10:00行业85
NVIDIA AI Supercomputer Comes Online at Naval Postgraduate School
英伟达为美国海军研究生院部署DGX GB300超算，强化军事AI教育与研究能力。
671海外Hugging Face08:00产品85
Bringing Nunchaku 4-bit Diffusion Inference to Diffusers
Nunchaku 4-bit扩散推理集成到Diffusers，提升效率。
672海外OpenAI08:00产品85
Launching Health in ChatGPT
ChatGPT推出健康功能，美国用户可连接医疗记录和Apple Health获取个性化健康洞察。
673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智能体研究。
674海外Google Research02:40研究85
Towards a quantum computer that learns from its errors
量子计算机可自我学习纠错，迈向实用化。
675海外Google DeepMind21:38行业85
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。
676海外NVIDIA21:00产品88
NVIDIA Open Sources First GPU-Accelerated Medical Physics Simulation Framework
英伟达开源首个GPU加速医学物理模拟框架，助力医疗机器人应对真实物理交互。
677海外OpenAI21:00行业85
How news organizations are using AI to advance their vital missions
新闻机构正用AI强化报道、增长受众并优化运营，OpenAI工具助力全球记者与出版商。
678海外Google AI21:00产品78
3 Google updates from Galaxy Unpacked 2026
三星新品发布会集成谷歌AI功能，提升生产力与效率。
679海外OpenAI21:00行业72
Building AI infrastructure with the Effingham County community
OpenAI在佐治亚州启动Project Camellia，承诺负责任能源、社区投资与就业。
680海外OpenAI20:00行业82
Advancing the next era of national science
OpenAI与美国能源部及国家实验室合作，用前沿AI加速科学发现。
681海外OpenAI13:30产品85
Introducing OpenAI Presence
OpenAI推出企业级AI代理平台，用于部署可信语音和聊天代理。
682海外OpenAI08:00行业75
NTT DATA Group cuts incident analysis to 30 minutes with Codex
NTT DATA用ChatGPT和Codex将事故分析缩至30分钟，惠及9000员工。
683海外NVIDIA06:35行业82
Built in Fort Worth: Wistron Opens Advanced Manufacturing Plant to Produce NVIDIA AI Systems
纬创在德州沃斯堡开设美国首座工厂，生产NVIDIA AI超级芯片系统。
684海外NVIDIA23:36行业92
NVIDIA Vera Rubin Driving Performance Per Watt, Lowest Token Cost for Partners Worldwide
NVIDIA Vera Rubin平台量产，以最低每瓦性能与Token成本服务全球合作伙伴。
685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，覆盖高性能、轻量与安全场景。
686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，简化数据记录流程。
687X@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，效果出奇好。
688海外Google DeepMind23:00模型85
Introducing Gemini 3.5 Flash Cyber
谷歌发布轻量级网络安全模型Gemini 3.5 Flash Cyber，用于发现和修复漏洞。
689海外Google AI00:00产品75
Connect more of your apps to Search
AI模式可安全连接更多应用，直接在搜索中交互。
690海外Google AI00:00产品72
Create, edit and star in videos with two Google Vids updates
Google Vids新增Gemini Omni与个人头像功能，简化视频创作。
691海外Hugging Face19:49行业82
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。
692海外Google DeepMind17:30研究85
Our approach to bioresilience
谷歌DeepMind与Isomorphic Labs联合发布生物韧性AI方法，应对生物风险。
693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。
694海外Google Research02:06研究82
Towards demystifying the creativity of diffusion models
理论视角解读扩散模型创造力的来源与机制。
695海外Hugging Face01:27研究85
Model Routing Is Simple. Until It Isn’t.
探讨模型路由的复杂性，从简单到挑战的深度解析。
696海外Hugging Face08:00产品85
Introducing Real World VoiceEQ: Measuring the human quality of voice AI
新工具VoiceEQ量化语音AI的拟人质量，推动更自然交互。
697海外Hugging Face08:00产品75
Welcome Inkling by Thinking Machines
Thinking Machines发布Inkling，一款新AI产品。
698海外Google AI00:00产品75
Celebrating 25 years of visual search innovation
Google Images迎来25周年，回顾里程碑并推出视觉探索新功能。
699海外Microsoft Research00:00研究82
Verifying Rust cryptography in SymCrypt, from standards to code
微软新方法在开发时验证Rust加密代码，兼顾速度与适应性。
700海外Google DeepMind20:37产品72
Empowering India’s next generation of innovators with ATL Saathi
Google与AIM推出Gemini驱动的ATL Saathi，助力印度教育者开展机器人教学。
701海外Hugging Face08:00实践75
Profiling in PyTorch (Part 3): Attention is all you profile
PyTorch性能分析第三篇，聚焦注意力机制的性能剖析技巧。
702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。
703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。
704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。
705海外Hugging Face08:00产品85
Native-speed vLLM transformers modeling backend
vLLM推出原生速度的transformers建模后端，提升推理性能。
706海外Hugging Face05:15产品75
From Hugging Face to Amazon SageMaker Studio in one click
一键将Hugging Face模型部署到Amazon SageMaker Studio，简化AI工作流。
707海外Google Research00:42研究72
The power of collaboration: How we can reduce traffic congestion
探讨算法协作如何缓解交通拥堵，强调理论与实际结合的价值。
708海外Hugging Face23:20实践75
Hugging Face Models on Foundry Managed Compute
介绍在Foundry托管计算上运行Hugging Face模型的方法。
709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。
710海外Hugging Face08:00产品85
Run AI workloads on any cloud, store on Hugging Face: zero-egress storage with SkyPilot
SkyPilot与Hugging Face集成，实现零出口费跨云存储与运行AI负载。
711海外Hugging Face08:00模型82
LeRobot v0.6.0: Imagine, Evaluate, Improve
LeRobot v0.6.0发布，新增想象、评估与改进功能，提升机器人学习效率。
712海外Hugging Face23:30模型82
PRX Part 4: Our Data Strategy
PRX发布第四部分，详解其数据战略与AI训练数据策略。
713海外Hugging Face08:00行业85
🤗 Kernels: Major Updates
Hugging Face内核重大更新，提升性能与效率。
714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。
715海外Google AI02:15行业75
The latest AI news we announced in June 2026
谷歌2026年6月AI更新汇总
716海外Google AI00:00会议75
New York City educators and industry leaders gathered at Google’s offices to shape the future of AI in classrooms.
纽约教育界与谷歌共商AI课堂应用前景
717海外Hugging Face08:00产品85
Hugging Face and Cerebras bring Gemma 4 to real-time voice AI
Hugging Face与Cerebras合作，将Gemma 4模型引入实时语音AI领域。
718海外Hugging Face02:32研究85
ScarfBench: Benchmarking AI Agents for Enterprise Java Framework Migration
新基准ScarfBench评估AI代理迁移企业Java框架的能力。
719海外Google Research01:03行业72
Expanding our Heat Resilience data to 50+ global cities
IBM将高温韧性数据扩展至全球50多个城市，助力气候适应。
720海外Microsoft Research00:50研究82
微软提出SkillOpt，将智能体技能编辑转化为训练过程，提升行为可靠性。
721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。
722X@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。
723海外Hugging Face22:39实践85
Why Specialization Is Inevitable
AI发展必然走向专业化，通用模型将让位于垂直领域。
724海外Google Research18:26模型85
Introducing TabFM: A zero-shot foundation model for tabular data
TabFM是首个面向表格数据的零样本基础模型，无需微调即可预测新数据集。
725海外Google AI14:00行业75
谷歌英国发布经济影响报告，探讨如何让更多人受益于AI技术。
726海外Hugging Face08:00产品82
Featuring Every Eval Ever Results on Hugging Face Model Pages
Hugging Face模型页新增全面评测结果展示功能。
727海外Microsoft Research05:14研究85
Memora: A Harmonic Memory Representation Balancing Abstraction and Specificity
微软提出Memora记忆系统，平衡抽象与具体，提升AI代理长任务效率。
728海外Hugging Face02:02研究82
DiScoFormer: One transformer for density and score, across distributions
新Transformer架构DiScoFormer统一处理密度估计与评分，跨分布高效推理。
729海外Google AI00:00实践75
谷歌专家解读AI全栈方法的内涵与长期价值。
730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。
731海外Hugging Face08:00实践75
Run a vLLM Server on HF Jobs in One Command
一键在HF Jobs上运行vLLM服务器，简化部署流程。
732海外Microsoft Research00:00研究85
Understanding the brain with AI-driven explanations and experiments
微软研究提出生成式因果测试，用AI解释大脑语言处理机制。
733海外Google AI00:00产品75
Our latest Google Finance upgrades, including a new app
谷歌财经升级，推出新Android应用。
734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。
735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。
736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。
737海外Hugging Face00:00产品85
Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel
NVIDIA NeMo AutoModel加速Transformer微调。
738海外Microsoft Research22:00研究85
Talos: Scaling rare disease diagnosis with automated, iterative genomic reanalysis
微软开源Talos系统，通过自动化迭代基因组重分析，将罕见病诊断效率提升90%。
739X@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。
740海外Google AI23:00研究85
谷歌医疗AI系统AMIE在复杂疾病管理上媲美初级保健医生。
741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。
742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。
743X@DrJimFan00:31实践85
物理自动研究系统背后安全与设计细节揭秘
744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。
745海外Google AI23:00行业75
We’re strengthening our presence in Alabama through new investments and community support.
谷歌投资15亿美元扩建阿拉巴马州数据中心，强化社区支持。
746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。
747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。
748海外Google AI04:00行业65
谷歌在弗吉尼亚投资社区，支持就业与能源可负担性。
749X@karpathy01:45行业85
SpaceX的过去、现在与未来令人惊叹，值得反复思考。
750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。
751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。
752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。
753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。
754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。
755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。
756X@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。
757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
758海外Google AI22:45会议85
The latest AI news we announced in May 2026
谷歌2026年5月AI更新汇总，涵盖模型、产品与工具。
759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。
760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。
761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。
762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。
763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。
764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。
765海外Google Research01:52研究85
Empirical Research Assistance (ERA): From Nature publication to catalyzing Computational Discovery
Nature发表ERA框架，推动计算发现新范式。
766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。
767海外Google DeepMind02:21研究85
Fast-tracking genetic leads to reverse cellular aging
AI辅助发现逆转细胞衰老新因子，加速抗衰老研究。
768X@karpathy23:05行业75
作者宣布加入Anthropic，看好大语言模型前沿发展。
769海外Google DeepMind03:53产品82
Simulate real-world places with Project Genie and Street View
谷歌AI Ultra订阅者获新功能，可结合街景模拟真实场景。
770海外Google DeepMind03:50模型92
Introducing Gemini Omni
谷歌发布全能AI模型Gemini Omni，整合文本、图像、音频等多模态能力。
771海外Google DeepMind03:43模型95
Introducing Google Antigravity 2.0
谷歌发布反重力2.0，颠覆性技术突破。
772海外Google DeepMind21:50产品85
Gemini for Science: AI experiments and tools for a new era of discovery
谷歌推出Gemini科学工具，加速科研发现。
773海外Google DeepMind21:43产品75
Making it easier to understand how content was created and edited
Adobe扩展内容来源工具，帮助用户理解网络内容的创建与编辑过程。
774海外Google DeepMind17:13行业75
Strengthening Singapore’s AI Future: A New National Partnership
谷歌DeepMind与新加坡合作，用前沿AI应对健康、教育等挑战。
775海外Google DeepMind16:16产品85
Finding the molecular switches behind new infectious diseases
AI工具Co-Scientist助力发现新发传染病分子开关
776海外Google DeepMind16:08产品85
Opening new paths in aging research
Calico利用AI系统Co-Scientist整合衰老研究碎片化发现，生成新线索。
777X@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。
778海外Google Research00:37行业75
Catalyzing scientific impact through global partnerships and open resources
全球合作与开放资源加速科学影响力。
779海外Google Research05:07实践75
Four ways Google Research scientists have been using Empirical Research Assistance
谷歌研究员用实证研究助手提升效率的四种方法。
780X@karpathy01:43实践65
强调思考不可外包，理解必须亲为。
781X@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。
782海外Google Research01:00产品75
It's all about the angle: Your photos, re-composed
用生成式AI自动重构图，让照片更完美。
783海外Google Research00:42研究85
ReasoningBank: Enabling agents to learn from experience
让AI智能体从经验中学习，提升推理能力。
784海外Google Research22:41研究82
探讨如何从第一性原理设计合成数据集，以模拟真实世界机制与推理。
785海外Google Research20:18研究85
AI-generated synthetic neurons speed up brain mapping
AI生成合成神经元加速脑图谱绘制
786海外Google Research02:15实践75
Towards developing future-ready skills with generative AI
探讨生成式AI如何培养未来技能的教育创新观点。
787海外Google Research19:22研究78
ConvApparel: Measuring and bridging the realism gap in user simulators
提出ConvApparel框架，衡量并弥合用户模拟器中的真实感差距。
788海外Google Research04:01产品75
Improving the academic workflow: Introducing two AI agents for better figures and peer review
介绍两个AI代理，分别用于改进学术图表和同行评审流程。
789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。
790海外VentureBeat AI22:00产品85
Claude Code costs up to $200 a month. Goose does the same thing for free.
AI编程助手Claude Code收费高达200美元/月，开源替代Goose免费提供类似功能。
791海外VentureBeat AI22:01行业85
Listen Labs 靠病毒式招聘广告获 6900 万美元融资，用于扩大 AI 客户访谈业务。
792海外VentureBeat AI21:00产品85
Salesforce rolls out new Slackbot AI agent as it battles Microsoft and Google in workplace AI
Salesforce 将 Slackbot 升级为 AI 智能体，可搜索数据、起草文档并代表员工执行操作。
793海外VentureBeat AI19:30产品88
Anthropic launches Cowork, a Claude Desktop agent that works in your files — no coding required
Anthropic发布Cowork，让非技术用户也能用Claude操作文件，团队用Claude Code一周半建成。
794海外VentureBeat AI04:00模型78
Nous Research's NousCoder-14B is an open-source coding model landing right in the Claude Code moment
Nous Research 发布开源编程模型 NousCoder-14B，四天训练完成，性能媲美更大闭源模型。