Latest

Every published item, in order · 55824 items
Nov 13
ModelsBuildersMulti-source06:00
Chinese contentGoogle 发布 Gemini 4 Hydrogen,改进复杂问题理解能力

Google 把 Gemini 4 Hydrogen 直接塞进了搜索框,能拆解带好几个子问题的提问,连打错的错别字也能看懂。和只答一半的旧搜索 AI 模式相比,明显更聪明,经常用搜索的人今天就能试到。

Event
Official accounttheverge@Stevie BonifieldSource
ProductsBuildersMulti-source78°05:54
Chinese contentAnthropic 推出 Claude Meetings 测试版,可进入 Google Meet 参会并现场产出文档

Anthropic 把 Claude 拉进 Google Meet 当参会者了,边开会边写文档、画白板、查公司数据。跟 Gemini、Zoom 这些纪要助手比,它是实时跟着大家改东西,还能接你连好的工具。用 Team 或 Enterprise 套餐的公司可以试试,客户多的团队记得先看看默认设置。

Event
宝玉@doteySource
PapersBuildersSelected05:54
Arena.ai studies Interaction-as-a-Judge evaluation for computer-use agents

Evaluation methodology is quietly converging with agent capabilities: the better a model operates a GUI, the better it can grade outputs, creating a feedback loop between computer-use progress and benchmark quality. Teams building coding agents, game-generation tools, or computer-use products can download the two open trace datasets to train or validate their own judges. Compared with static human-vote arenas, interaction-based judging offers a cheaper way to scale evaluation as generated apps grow more complex.

lmarena.ai@lmarena_aiSource
ModelsBuildersSelected05:54
LMArena open-sources two web development evaluation datasets

Preference data of this scale for web app generation is scarce, and the pairing of subjective preference labels with verified functional requirements gives training and evaluation teams two complementary signal types. Model developers working on coding agents can benchmark against human judgment rather than proxy metrics. Comparison with existing web-dev benchmarks becomes straightforward since both datasets and the accompanying paper are public.

lmarena.ai@lmarena_aiSource
PapersBuildersSelected05:53
Arena.ai Open-Sources Datasets for Interaction-as-a-Judge Evaluation

Evaluation shifts from comparing static code screenshots to judging how agents actually behave inside an app, which changes how generation quality can be measured for web development agents. Builders of coding agents and eval pipelines can use the two open-sourced trace datasets to train reward models or benchmark correctness without collecting their own user complaints data. Compared with human-preference-only arenas, this adds an automated signal that scales with computer-use capability.

lmarena.ai@lmarena_aiSource
PapersBuildersSelected05:53
Wayne Chi Explains Interaction-as-a-Judge for Agent Evaluation

The core takeaway is that evaluation signal can be harvested from real user interactions instead of hand-written test suites, shifting agent benchmarking toward live platform data. Researchers and teams building computer-use agents can apply this pattern to score traces with LLM judges or reward models trained on user complaints. Code Arena and Agent Arena already supply the battle and complaint data, giving practitioners a concrete starting point for evaluation pipelines.

lmarena.ai@lmarena_aiSource
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
ProductsBuildersMulti-source05:48
Chinese contentOpenAI 为 Dots 加入自定义指令并加快响应速度

如果你每天都要跟 Dots 开一堆会话,这个更新能帮你把常用的设定一次写好,以后不用每次重新交代,回复也变快了。用过 ChatGPT 自定义指令的话会很容易上手,做智能体工作流的人可以试试。

Event
TestingCatalog@testingcatalogSource
IndustryMulti-source05:42
Chinese content从 Apple 到 OpenAI:创始人被罢免的历史重演

把 1985 年的 Jobs、2017 年的 Kalanick 和 2023 年的 Altman 摆在一条时间线上看,你会发现董事会炒创始人这事几乎每十年重演一次。搞创业或投资的都值得花一分钟琢磨这个规律:创始人是不是可替代的,答案可能一直是不能。

Event
rohanpaul_ai@rohanpaul_aiSource
Chinese contentbrowser-use 接入 1Password:智能体登录无需接触密码

browser-use 和 1Password 合作出了个集成:智能体碰到登录页时你在 1Password 里点批准,密码由它填,智能体全程看不到密码明文。做浏览器自动化、RPA 或者要跑真实网站任务的开发者可以直接接上,省去自己管凭证的麻烦。

Browser Use@browser_useSource
Chinese content思科 2027 财年 Q1 净利润 45.52 亿美元,同比增长 59%

思科这季财报数据很硬:营收增速 24% 是 15 年来最高,网络业务涨了 43%,全靠云厂商买 AI 数据中心设备。它还把全年 AI 基建收入预期提到 90 亿美元,直接翻倍。关注 AI 算力产业链的投资者和做企业网络的工程师都该看看这份财报细节。

IT之家Source
ModelsBuildersMulti-source05:35
Chinese contentGemini 4 Hydrogen 全球推送,免费用户可通过自动模型选择使用

Google 把 Gemini 4 Hydrogen 推到了免费用户的自动路由里,桌面、手机、网页全都有。不用付费就能在合适的 prompt 下用到它,想知道路由怎么判断的可以看官方文档。用 Gemini 的朋友今天打开 app 就能碰运气。

Event
Google Gemini App@GeminiAppSource
ProductsBuildersMulti-source72°05:35
Chinese contentAnthropic 推出 Claude Meetings,视频会议中实时生成文档与白板

Anthropic 把 Claude 直接拉进视频会议里,边开会边帮你写文档、做白板,省掉会后整理纪要的功夫。目前只在 Team 和 Enterprise 计划开 beta,用 Claude 办公的团队可以关注自己账号有没有收到入口。

Event
Alex Albert@alexalbert__Source
Chinese contentOcten 发布开源低延迟搜索引擎 Octen Pulse,主打智能体高频调用场景

10ms 延迟的开源搜索,还包月 20 美元不限调用量,专门给智能体循环里几十次检索的场景设计的。跟按次计费的 Brave、Tavily 这些比,跑高频任务的 Agent 开发者能省不少钱,还能自己部署,做智能体工具链的可以试试。

elvis@omarsar0Source
ModelsBuildersMulti-sourceSelected05:32
Inkling 2 Small Runs on vLLM From Day Zero

Day-zero vLLM support plus single-node GB200 fit means operational cost of serving this model is far lower than typical 594B-scale deployments. Engineering teams running self-hosted inference should benchmark the 630 tok/s TP4 figure against their current MoE serving setups, since the DSpark speculator and NVFP4 experts are the levers behind it. Teams needing 1M-token multimodal context without multi-node clusters have a concrete new option to evaluate.

Event
Official accountvLLM@vllm_projectSource
IndustryBuildersMulti-source81°05:18
Chinese contentReuters:SpaceX 计划投入超 740 亿美元在德州和亚利桑那建两座 AI 超算园区

SpaceX 这笔 740 亿美元的算力投入比很多国家一年的 AI 基建都大,Starbase 自建燃气电厂不走公共电网的做法很罕见。亚利桑那站点的干式冷却和电费下降 20% 的承诺也值得关注,做基础设施或算力采购的人应该看看细节。

Event
rohanpaul_ai@rohanpaul_aiSource
How-toBuildersOfficial05:15
Chinese contentAmazon Bedrock 推理省钱的 4 个请求级参数

很多人调完提示词发现 Bedrock 账单还在涨,这篇讲的就是原因:缓存时长、service tier、推理力度和 maxTokens 这 4 个请求参数在偷偷决定你的费率。每个都附了测试和验证方法,接了 Bedrock 的开发者直接照着改就行。

Primary sourceAWS Machine Learning Blog@Sai Pavan KasaniSource
Chinese contentReplit 入选 Newcomer 与 Bessemer 首届 2026 AI 100 榜单

如果你关注 AI 编程工具,可以留意这个:Newcomer 和 Bessemer 第一次办 AI 100 榜单,Replit 就在首批名单里。和 Cursor、GitHub Copilot 这些本地或插件式工具不同,Replit 走的是浏览器内开发加 AI 的路线,对不想配环境的开发者挺友好。

Official accountReplit@ReplitSource
How-toBuildersMulti-source05:14
Chinese contentGemini 4 Hydrogen 接入 AI SDK:速度较 3.8 Flash 提升 2.5 倍

Google DeepMind 的新模型 Gemini 4 Hydrogen,比 3.8 Flash 快 2.5 倍,编码和多模态都覆盖了。如果你在用 Vercel 的 AI SDK,改一行模型名就能切过去试试,搞智能体和聊天应用的开发者可以直接上手。

Event
AI SDK@aisdkSource
ModelsBuildersMulti-source84°05:13
Chinese contentThinking Machines 的 inkling 2 small 上线 Together AI

Mira Murati 的 Thinking Machines 把 inkling 2 small 放上了 Together AI,594B 参数只激活 17B,还有 1M 上下文。写代码、跑长任务、看图都能覆盖,推理成本却按小模型算。做 agent 或长文档处理的开发者可以直接去 Together AI 试。

Event
Official accountTogether AI@togethercomputeSource
ModelsBuildersMulti-sourceSelected05:10
Thinky Machines releases Inkling 2 Small weights on HuggingFace

License terms, not model size, are the real differentiator here: Apache 2.0 lets enterprises ship commercial products without the compliance overhead that Llama's custom license imposes. Teams evaluating self-hosted models for cost control or data privacy can fine-tune and deploy Inkling 2 Small with no royalties and no obligation to release source code. Engineering leads comparing open-weight options should shortlist it alongside other permissively licensed releases rather than treating all "open" models as equivalent.

Event
Official accountThinking Machines Lab@thinkymachinesSource

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