Latest

Every published item, in order · 15368 items
Nov 14
IndustryUS vs ChinaMulti-source04:31
Chinese contentAnthropic 更新 IPO 文件:预期模型能力进步速度翻倍

Anthropic 把'模型能力进步速度翻倍'写进了 IPO 文件,这等于向投资人承诺自家用模型加速研发已经见效。对比 OpenAI 还没上市、Google 不单独披露研发节奏,这种量化表态很少见。想跟踪 Claude 迭代速度或打算参与其 IPO 的人应该看原文。

Event
The Information@theinformationSource
IndustryUS vs ChinaMulti-source91°04:28
Chinese content黄仁勋离开 Nvidia 或因激进商业操作而非个人原因

WSJ 挖了黄仁勋走人的内幕:不是私事,是 Nvidia 敢给云厂商几百亿美元担保、还规划给 OpenAI 2500 亿美元芯片订单,CFO 想踩刹车被他嫌弃。关心 Nvidia 现金流和股价的投资者、跟踪 GPU 供应链的人应该看看。

Event
Official accountDecoder@Matthias BastianSource
IndustryUS vs ChinaMulti-sourceSelected04:24
Daebak Cafe Disputes OpenAI Blackwell GPU Fingerprint Report

The teardown shows how a statistical fingerprint of initialisation hardware was stretched into a headline about massive sanctioned training clusters, a useful lesson in reading benchmark and forensics claims past the summary. Export-control watchers and enterprise GPU buyers in Southeast Asia should note that legally renting offshore Blackwell capacity in Singapore, Japan or the Middle East produces the same fingerprint as smuggled hardware. Anyone consuming US-China chip narratives should ask for the raw evidence before the headline numbers.

Event
宝玉@doteySource
IndustryBuildersMulti-source04:06
Chinese content两周内15家厂商密集发布"决策模型",Vercel报告称26%请求已由其路由

有人整理了一份完整时间线,两周内 Together AI、OpenAI、Google 等15家全上了自己的决策模型,连命名都互相抄。Vercel 那个26%路由占比数据挺值得琢磨,做 agent 架构的可以对照各家参数量和发布节奏看看选型。

Event
Official accountSebastian Raschka@rasbtSource
IndustryUS vs ChinaMulti-source84°04:01
Chinese contentWill Depue 称中国四家实验室用 Blackwell 训练模型,但报告自认不能证明

这条值得细看,因为推文和报告原文差距很大:报告自己写着"不能证明",推文里却变成了"确凿证据"。做开源模型、用云算力的团队都会被这类指纹指控波及,先搞清证据边界比转发结论有用。

Event
宝玉@doteySource
IndustryBuildersMulti-source03:24
Chinese content英伟达借贷财团陷入变动,至少一家贷款方考虑退出

Gary Marcus 转发了 NYT 关于 NVIDIA 借贷财团的报道,六家贷款方里至少一家在讨论退出,一笔贷款都还没成交。参与名单里有 GS、Apollo、BlackRock 这些大机构,如果你在关注 AI 资本开支和算力融资,这条值得看一眼。

Event
Gary Marcus@GaryMarcusSource
IndustryMulti-source03:20
Chinese contentAnthropic CEO Dario Amodei 对 Jamie Dimon 谈 IPO 策略:"underpromise and overdeliver"

Dario Amodei 亲自向 Jamie Dimon 确认 Anthropic 要 IPO,还把策略说成"少承诺多兑现"。对比 OpenAI 一直没松口上市,Anthropic 这话算是给资本市场递了个信号。想判断它什么时候上市、怎么估值的朋友值得看看这段对话。

Event
PolymarketMoney@PolymarketMoneySource
Chinese contentRadiant CEO:AI 基础设施建设远不止算力,四大支柱缺一不可

radiance 这个话题很多人只盯着 GPU,Radiant 的 CEO 把带电土地、资本、软件和算力四个环节摆在一起讲,Brookfield 背景也让它有底气。做 AI 部署决策的工程和采购团队可以看看别人怎么补电力这块短板。

The Information@theinformationSource
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
IndustryBuildersOfficial03:12
Chinese contentGoogle Research 与 Google DeepMind 在剑桥大学举办 AI Research Inspiration Day

Google 两支研究团队把工具带到剑桥实验室现场测试,聊的是多智能体集群和 48 小时编码流程怎么压缩成快工作流。做科研自动化或 agent 开发的可以看看这些一线研究人员怎么用 AI,比看产品发布会实在。

Primary sourceGoogle Research@GoogleResearchSource
IndustryUS vs ChinaMulti-source03:01
Chinese content黄仁勋卸任 Nvidia CEO,DeepSeek 间接投资方曝光

The Information 独家爆料,黄仁勋不再执掌 Nvidia,还顺带挖出 DeepSeek 间接投资方里有 Alice Huang 关联公司。两条线都指向 AI 资本圈的人脉网络,关心 Nvidia 接班人安排和 DeepSeek 股东结构的人都该看看原文。

Event
The Information@theinformationSource
Chinese contentLenny Rachitsky 报告:15% 科技岗位集中在湾区,远程岗位创历史新低

这份半年一更的就业报告有个反直觉数据:湾区一个地区的在招岗位已经超过全球远程岗位总和,而三年前远程还领先近 2 倍。想找远程工作或考虑搬去湾区的人应该看看,报告里 PM、工程、设计三条曲线走势完全不同。

Lenny Rachitsky@lennysanSource
IndustryBuildersSelected02:26
Databricks Takes 2nd in Meta Muse Spark Kernel Challenge

Autonomous agent loops are now producing production-grade GPU kernels that beat mainstream inference-framework baselines, which lowers the cost floor for serving AI workloads. Engineering teams running high-volume inference can treat LLM-driven kernel optimization as a practical pipeline rather than a research demo. The finding that Opus 5.5 outperforms other models on kernel optimization also gives practitioners a concrete model choice for similar code-optimization agents.

Yuchen Jin@Yuchenj_UWSource
IndustryUS vs ChinaMulti-source02:12
Chinese content英伟达上海数据中心被曝存有被禁芯片,黄仁勋被指撒谎

彭博记者挖到备案文件,证明上海 Megaspeed 数据中心真有被禁的英伟达芯片,而英伟达一年前对媒体口径是"没什么可看的"。Gary Marcus 还挖出黄仁勋对妻子和法务撒谎的旧料。关注出口管制执行和美国对华芯片监管的人值得看看这条线索链。

Event
Gary Marcus@GaryMarcusSource
Chinese contentLangChain 完成 3 亿美元 C 轮融资,估值达 30 亿美元

LangChain 这轮拿了 3 亿美元,估值 30 亿,钱主要砸向 LangSmith 这个智能体运行和治理平台。它不绑定自家框架,用别的工具链也能接入。做智能体应用的开发者和团队选观测、评测方案时可以关注一下它跟同类工具的差异。

Harrison Chase@hwchase17Source
IndustryBuildersMulti-source01:52
Chinese contentSpaceX 团队一个月合并 1.1 万个 PR,大部分不经人工审查

SpaceX 一个月合了 1.1 万个 PR,大部分不靠人看,这事对做工程流程的人很有参考价值。你可以对照自己团队的 review 瓶颈想想哪些环节能自动化。不过原文没讲具体工具链,当成流程信号看就好。

Event
AI Will@FinanceYF5Source
IndustryBuildersMulti-source01:49
Chinese contentAnthropic 用 1,000 个 Claude agent 在 28.7 小时内完成费马大定理证明

1,000 个 Claude agent 接力跑 28.7 小时证完费马大定理,这个规模的多智能体协作实验很少见。文中还讲了 Claude 的记忆和「做梦」机制怎么设计,17 张幻灯片都在长文里。做 agent 编排的人可以看看别人怎么管理超长任务上下文。

Event
AI Will@FinanceYF5Source
IndustryBuildersSelected01:04
Code Arena data shows October models beat April's best at 20x lower cost

The compression of capability-to-price ratios within six months means teams buying coding-model capacity on old benchmarks and contracts are overpaying. Engineering leads and procurement teams in ASEAN and China markets should re-run head-to-head evaluations before renewing API commitments, since mid-priced options like MiMo-V2.6-Flash now match former frontier quality. The dataset offers a reusable method for tracking value, not just raw scores.

lmarena.ai@lmarena_aiSource
IndustryUS vs ChinaMulti-source00:58
Chinese contentAlice Huang 关联公司经新加坡代名人向上海走私 Nvidia 禁运 AI 芯片

美国制裁之下,Nvidia 的高端芯片还是经新加坡代名人渠道流进了上海的数据中心,这条走私链路被完整挖出来了。做跨境算力采购或出口合规的人应该看看:新加坡中转不是安全区,芯片的最终落地地点才是审查重点。

Event
Techsauce@ Techsauce Team Source
Chinese content25 岁前 Google X 工程师 Cecilia Shen 创立 AI 电影工作室 Utopai Studios

前 Google X 女工程师 25 岁就辞职,把自己做的 Cybever 变成了 Utopai Studios,从工具直接切进 AI 电影制作全流程,还拿到了好莱坞投资人的钱。想了解 AI 怎么改影视制作流程、创业公司怎么从工具走向内容生产的,这个故事挺有参考价值。

Techsauce@ Techsauce Team Source
IndustryBuildersSelected00:54
Sakana AI CEO argues model orchestration beats single giant models

The piece reframes AI sovereignty as supply-chain resilience, which changes how governments and enterprises should evaluate national AI strategies beyond blanket localisation. It matters most for policy teams and AI builders in cost-constrained markets: orchestration of smaller specialised models offers a competitive path without frontier-scale budgets. Japan's positioning contrasts with China and US scale-first labs, and the argument carries for Southeast Asian countries drafting their own AI sovereignty policies.

Official accountSakana AI@SakanaAILabsSource

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