Latest

Every published item, in order · 29223 items
Nov 13
ProductsBuildersMulti-source71°03:08
Chinese contentGoogle Gemini 智能眼镜开启预售,12月10日发货

Google 的 Gemini 眼镜终于能买了,12 月 10 日发货,镜框来自 Gentle Monster 和 Warby Parker,不用挂着科技感的怪样子。跑在 Android XR 上,和 Meta Ray-Ban 直接打对台。想试试免提用 Gemini 的人可以去门店体验一下。

Event
Official accountDemis Hassabis@demishassabisSource
ModelsUS vs ChinaMulti-source71°03:04
Chinese contentGemini 4 Hydrogen (High) 亮相 Agent Arena:速度与成本占优

如果你在挑智能体模型,Hydrogen 值得放进候选清单:速度是 Claude Fable 5.2 的 3 倍,成本不到 GPT-6 Astra 的三分之一,可控性还排第 2。代价是输出 token 烧得多,Code 任务只排第 9。跑高频对话类智能体、对成本敏感的团队可以优先试它。

Event
lmarena.ai@lmarena_aiSource
ProductsBuildersMulti-sourceSelected03:03
CoreWeave webinar on hot-swapping RL checkpoints in production

The non-obvious point is that RL cadence, not compute size, becomes the serving bottleneck once weights update every few hours, and hot-loading removes the downtime tradeoff. ML platform engineers and inference teams running post-training pipelines should attend to copy the Dynamo and RL Rollouts setup directly, including cost figures. Comparing on-policy weight transition approaches before committing to an orchestration stack could save a costly rework later.

Event
CoreWeave@CoreWeaveSource
PapersBuildersSelected03:03
SteerDistill Quantizes LLM Agents via Closed-Loop Teacher Steering

The non-obvious point is that the teacher remains involved after deployment-style compression, correcting the student agent inside its execution loop instead of only at training time. Teams deploying agents on consumer GPUs or edge devices should evaluate whether this steering loop cuts their inference bill without breaking tool-use chains. Compared with static quantization pipelines like plain GPTQ-style post-training, the closed-loop approach targets the multi-step drift problem specific to agents.

AK@_akhaliqSource
ProductsBuildersMulti-source03:02
Chinese contentReplit Agent 新增购买能力,结合 Link 和 Visa 完成支付

Replit 把 Agent 从帮你写应用升级到帮你花钱,购物和支付交给 Link 和 Visa,你只批一个总额。比起其他只能干活的编程智能体,它开始管经营了。中小企业主和独立开发者可以直接试试这条采购自动化路径。

Event
Official accountReplit@ReplitSource
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
ProductsBuildersMulti-source72°02:57
Chinese contentWorld Labs 向所有用户开放 3D 世界模型 Atlas

李飞飞创办的 World Labs 把 Atlas 世界模型全量开放了,浏览器里就能用,API 也一并上线。它输出的不只是视频,还有能直接进 3D 渲染管线的高斯泼溅资产,和 Sora、Veo 这类纯视频产品思路不同。做游戏场景、虚拟空间或仿真数据的开发者值得上手试试。

Event
Official accountFei-Fei Li@drfeifeiSource
IndustryBuildersMulti-sourceSelected02:57
SGLang Consortium formed with NVIDIA, AMD, Google and Meta

Hardware vendors co-funding a single open-source inference engine reduces the risk that serving optimizations remain locked to one chip ecosystem, which matters for teams planning multi-vendor GPU fleets. Engineering leads running large-scale inference can track the consortium's roadmap to anticipate support for new models and distributed serving features. Organizations in Southeast Asia building inference capacity on heterogeneous hardware will benefit from upstream optimizations landing in SGLang rather than vendor-specific forks.

Event
Official accountLMSYS Org (SGLang)@lmsysorgSource
ProductsBuildersSelected02:43
LMSYS releases SGLang v1.0 and Miles v0.2 at SGLang Summit

The pairing matters: a hardened v1.0 inference engine plus a production-focused RL framework gives teams a single open-source alternative to closed serving and post-training stacks. Engineers building agent or chat products on self-hosted GPUs can benchmark SGLang v1.0 against vLLM for throughput and memory gains. Teams running RL post-training can evaluate Miles v0.2 for their specific hardware mix rather than stitching together ad-hoc pipelines.

Official accountLMSYS Org (SGLang)@lmsysorgSource
ProductsBuildersMulti-source02:42
Chinese contentGoogle 推出 Hydrogen:用更少的提示词快速生成可交互原型

Google 在 GeminiApp 上演示了 Hydrogen 的原型生成能力,接上 Gmail、Airtable、Dropbox 这些你已经在用的数据源,少几句提示词就能跑出可交互的布局。做前端或产品的人可以拿它试想法,比从零写 demo 快不少。

Event
Google Gemini App@GeminiAppSource
ProductsBuildersMulti-source02:41
Chinese contentFish Audio 推出 Fish Agent:无代码搭建可接打电话的语音智能体

Fish Audio 出了个无代码语音智能体平台,写句话就能部署一个能接打电话的 Agent,中位部署只要 32 秒,底层用自家 Live 3 全双工模型支持随时打断。能接自己的 LLM 和工具,走 API 集成,做客服或电话自动化的团队可以直接试试。

Event
Fish Audio@FishAudioSource
ModelsUS vs ChinaMulti-source88°02:41
Chinese contentGoogle 发布 Gemini 4 Hydrogen,主打快和省,免费层即可调用

Google 把 Hydrogen 定价压到 Sonnet 5.5 的四分之三,免费层就能调,日常写代码和跑 agent 的成本会明显下来。做 agent 和编程工具的开发者可以直接在 AI Studio 试,比 Terra 贵但更快,跟 Carbon 搭配用额度更省。

Event
宝玉@doteySource
Chinese contentGoogle Hydrogen 展示自然语言实时生成可交互代码

Google 用一个 3D 体素游戏 Geminisland 演示了 Hydrogen 的实时编码能力:说一句话,几秒生成能跑的交互代码,多人还能往同一个岛上实时添加结构。做实时生成类应用或协作工具的开发者可以看看这个思路,Gemini API 直接能用。

Official accountGoogle AI@GoogleAISource
ModelsBuildersMulti-source02:37
Chinese contentmillionwaysai 发布 Thorsten-4:主打识人动机的 Large Psychology Model

一家叫 millionwaysai 的公司发了个不走寻常路的模型,数据不是爬网页,而是心理学家面访了几千人标出来的。做销售、招聘类 AI 智能体的开发者可以看看,它专门补 LLM 读不懂人的短板。

Event
rohanpaul_ai@rohanpaul_aiSource
ModelsUS vs ChinaMulti-source74°02:36
Chinese contentGoogle 发布 Gemini 4 Hydrogen:速度快 2.5 倍的日常编码模型

Google 给 Gemini 4 家族添了个跑得快的成员,速度是 Gemini 3.8 Flash 的 2.5 倍还能打赢 Argon,思考档位随便调,跟 Carbon 搭着用在 Antigravity 里干编码和智能体活。做 agent 应用、被延迟卡过脖子的开发者可以盯一下。

Event
Official accountGoogle AI@GoogleAISource
ModelsUS vs ChinaMulti-source02:36
Chinese contentGemini 4 Hydrogen 今日起向全球所有用户推送

Gemini 4 Hydrogen 今天开始全球推送,桌面、手机、网页三端都能用。免费用户靠 auto model selection 就能被自动路由到这个模型,不用付订阅费。在意免费额度和模型档位选择的可以看看官方说明文档。

Event
Google Gemini App@GeminiAppSource
IndustryUS vs ChinaMulti-source02:30
Chinese contentBindu Reddy评Google vs OpenAI:Gemini Hydrogen对标Sonnet

Bindu Reddy把Gemini Hydrogen和Sonnet的价格、速度摆在一起对比,还说OpenAI的Astra 6.2迟迟不发让她开始担心。做模型选型的开发者可以看看这个价格差,关注OpenAI动态的人也值得看一眼她的判断。

Event
Bindu Reddy@bindureddySource
ModelsUS vs ChinaMulti-source02:29
Chinese contentGoogle 发布 Gemini 4 Hydrogen,主打高速编码与智能体场景

Google 这次把 Gemini 4 的能力下放到所有订阅层级和 API 通道,写代码和跑智能体的开发者可以在 AI Studio 里直接试用。和之前需要高端订阅才能用的 Gemini 4 比门槛低了不少,做智能体工作流的团队值得测一下速度表现。

Event
Official accountDemis Hassabis@demishassabisSource
ModelsUS vs ChinaMulti-source79°02:26
Chinese contentGemini 4 Hydrogen (High) 登上 Agent Arena 第 5 名,主打速度与成本

Google DeepMind 新模型 Gemini 4 Hydrogen 上线,Agent Arena 第 5 名,速度是 Claude Fable 5.2 的 3 倍,成本只有 GPT-6 Astra 的零头。跑智能体工作流、对 API 账单敏感的团队可以直接看下价格对比。

Event
lmarena.ai@lmarena_aiSource
ModelsUS vs ChinaMulti-source02:26
Chinese contentGemini 4 Hydrogen 成为 Agent Arena 速度最快的模型

做 Agent 应用的可以看看这份实测数据:Gemini 4 Hydrogen 单步只要 7.3 秒,10 分钟完成 82% 任务,速度比 Claude Sonnet 5.5 和 GPT-6.1 Sol 快不少,质量分数还没掉。对任务等待时间敏感的团队值得对比一下。

Event
lmarena.ai@lmarena_aiSource
ModelsUS vs ChinaMulti-source88°02:26
Chinese contentGoogle 发布 Gemini 4 Hydrogen,已在 GeminiApp 上线

Google 把新模型 Gemini 4 Hydrogen 直接塞进了 GeminiApp,主打速度快、智力提升明显。想尝鲜的可以直接打开 GeminiApp 试用,看看和 Claude、GPT 系列相比表现如何,尤其适合日常问答和快速任务场景。

Event
Josh Woodward@joshwoodwardSource
ModelsUS vs ChinaMulti-source88°02:24
Chinese contentGoogle 发布 Gemini 4 Hydrogen,面向所有用户层级开放

Google 把 Gemini 4 Hydrogen 铺到了 AI Studio、API、Gemini 应用甚至 Google Search,免费用户也能用。做开发的可以直接在 Gemini API 里调,付费用户还能在 Google Antigravity 里用。用 Gemini 生态的都该看看。

Event
Official accountLogan Kilpatrick@OfficialLoganKSource

Only the latest 2000 items are listed; for older ones see the daily archive