Latest

Every published item, in order · 55860 items
Nov 12
ProductsBuildersMulti-source06:45
Chinese contentSemiAnalysis 使用 Perplexity Computer 管理持久化编码智能体

Perplexity Computer 这个名字来自 SemiAnalysis 的博客,Dylan Patel 说自己 AI 支出 45% 花在持久化智能体上,还顺手吐槽了 Claude Tags 和 OpenAI 的 Slack 智能体。做智能体产品或工具选型的可以看看这条对比。

Event
Aravind Srinivas@AravSrinivasSource
Chinese content2026 科技岗位盘点:前端工程师职位一年减少 51%,AI 工程师增长 46%

这是 Lenny Rachitsky 第五份半年就业报告,数据很扎眼:前端、iOS/Android 职位一年掉了近一半,Forward Deployed Engineer 涨了 4.6 倍。正在找工作的工程师和设计师可以看看哪些方向还有机会,做职业规划别只盯着客户端了。

Lenny Rachitsky@lennysanSource
ProductsBuildersOfficial06:43
Chinese contentAmazon Bedrock 与 AgentCore 10 月更新汇总

这份 AWS 官方月度汇总把 10 月 Bedrock、AgentCore 和 Strands 的新功能整理在了一起。智能体稳定性、模型按任务选型和成本拆账,这三件事都是企业开发者在落地时最头疼的。正在用 Bedrock 搭智能体的团队可以直接对照清单排优先级。

Primary sourceAWS Machine Learning Blog@Tymofii KryvtsunSource
Chinese contentElevenLabs 谈语音智能体愿景:让用户不再按 0 找人工

ElevenLabs 的高管在访谈里聊了个很实在的判断:等人工客服排队的人,会慢慢改成点名要和 AI agent 谈。做客服、外呼产品的开发者可以看看他们怎么讲低延迟语音对话的价值,对比 Retell、Vapi 这类同赛道产品也有些参考。

Official accountElevenLabs@elevenlabsioSource
Chinese contentMicrosoft Research 播客解析红队、蓝队与紫队网络安全实践

Microsoft Research 的人聊 AI 时代怎么做安全测试,把红队、蓝队、紫队各自干什么讲得很清楚。做安全或想给自家系统做攻防演练的工程师可以听听,比单纯讲攻防对抗的视角更完整。

Event
Official accountMicrosoft Research@MSFTResearchSource
IndustryUS vs ChinaMulti-source06:40
Chinese contenta16z 对谈前 Cisco 并购主管 Mike Volpi:AI 基建规模已达 GDP 约 2%

在 Cisco 做过 70 多起收购的 Volpi,把 AI 基建和当年互联网浪潮放在一起对比:投资强度从 GDP 的千分之几涨到 2%,但他反而更乐观。聊到 Nvidia 该怎么应对定制芯片挑战、GPU 为何保值,做基础设施投资的和创业的人都该听听。

Event
a16z@a16zSource
ModelsBuildersMulti-sourceSelected06:39
Gemini Decision 1 Joins Jev and OpenAI Decision Models

Continuous probability scores instead of hard 1.0/0.0 answers let downstream systems threshold confidence on their own, which matters for media moderation and content-routing pipelines. Engineering teams building video or audio classifiers should benchmark Gemini Decision 1 against OpenAI's GPT-6 Luna, since the two cost the same but differ in modality reach. Budget-sensitive text-only workloads still favour Jev at 4.2 cents per million tokens.

Event
Official accountSimon Willison’s WeblogSource
ModelsBuildersSelected06:39
Google DeepMind releases AlphaGenome 2 for regulatory mutation prediction

The practical gain is roughly doubled sensitivity to non-coding regulatory variants, which is where most GWAS hits and rare-disease candidates live, making the model directly useful for variant prioritisation. Genomics researchers, clinical genetics teams, and biotech pipelines evaluating open genome models can fold it into existing annotation workflows. Teams working with tissue-specific datasets should check whether their cell types fall within the 68 covered tissues.

Official accountGoogle DeepMind@GoogleDeepMindSource
PapersBuildersMulti-sourceSelected06:21
Continual-ARC benchmark shows agent harnesses fail to learn

The actionable shift is that harness-side memory may be a dead end for continual learning, with weight updates on small models doing the real work. Agent builders and ML engineers evaluating frameworks like OpenClaw or Hermes should treat Continual-ARC as a screening tool before committing to a stack. Teams chasing domain-specific performance can weigh fine-tuning a compact model against prompt and memory engineering. AI researcher Elvis Saravia highlighted the work as aligning with his ongoing research into harness-model co-evolution.

Event
elvis@omarsar0Source
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
Chinese contentLogan Kilpatrick 预告 Google AI Studio 新功能已上线

Google AI Studio 负责人 Logan Kilpatrick 发帖官宣团队落地了一项新能力,已在 Google 多个产品中铺开,具体细节还没公布。关注 Gemini API 和 AI Studio 的开发者可以先去控制台看看有什么变化,等官方文档再确认。

Official accountLogan Kilpatrick@OfficialLoganKSource
ProductsBuildersMulti-source05:39
Chinese contentLovart 接入 GPT Image 2.6,一张照片一条提示词即可生成动图

做设计或社媒素材的朋友可以试试:Lovart 现在接入了 GPT Image 2.6,上传一张照片,一条提示词就能换背景再直接输出动图,不用再开 Runway 分两步做。升级会员最高送 31 天不限量生成,成本也算得过来。

Event
LovartAI@lovart_aiSource
Chinese contentMicrosoft 与菲律宾 DepEd 启动 2000 名教师 AI 素养培训

Microsoft 拉上菲律宾教育部,给 2000 名公立学校老师上 AI 课,教他们怎么把 AI 用进课堂备课和评估。跟普通企业培训不同,这是政府体系内的教师职业发展项目,关注东南亚教育数字化的朋友和做教育类 AI 产品的团队都该看看。

Inquirer: Technology@Jem SagcalSource
ProductsBuildersMulti-source05:38
Chinese contentCognition 为使用 Devin 的初创团队提供 65,000 美元额度与专属支持

Cognition 给用 Devin 的创业团队发 65k 美元额度还配专属支持,够用好一阵子。要是你们在纠结 Devin 和 Cursor、Copilot 哪个更适合自家开发流程,先用免费额度跑几个真实任务再决定,成本几乎为零。

Event
Cognition@cognition_labsSource
How-toBuildersOfficial05:27
Chinese contentAWS re:Invent 2026 AI/ML 议程参会指南:140+ 场次怎么选

re:Invent 的 AI 场次多到看花眼,这篇官方指南直接按角色和目标帮你筛好了路线。140 多场里分 breakout、workshop、chalk talk 几种形式,讲芯片、推理、Agent 这些方向,去现场的工程师和架构师值得按它排日程。

Primary sourceAWS Machine Learning Blog@Markus AdhiwiyogoSource
PapersUS vs ChinaMulti-sourceSelected05:23
OpenAI research finds chain-of-thought monitorability declines as models scale

The non-obvious takeaway is that safety monitoring based on reading a model's reasoning traces may silently weaken as capability grows, meaning audits that work today may not work on next year's models. Safety researchers, alignment teams and enterprises using CoT inspection for oversight should benchmark monitorability on their own fine-tunes. The training-lifecycle framing gives practitioners a concrete methodology for measuring when and where this degradation starts.

Event
elvis@omarsar0Source
ProductsBuildersSelected05:06
ElevenLabs Announces Fine-tuning, Interaction Engine and ElevenAI Assistant

The routing approach in the Interaction Engine — mixing cheap and frontier LLMs within one call — signals where voice agent economics are heading and gives enterprises a template for cutting token costs. Teams building customer-support or phone-based agents should evaluate the fine-tuning programme, since tailored speech models on proprietary data remain rare. The announcements also mark a direct rivalry with OpenAI's realtime voice stack and general work assistants like Granola or Wispr Flow.

Official accountElevenLabs@elevenlabsioSource
ProductsBuildersOfficial05:03
Chinese contentDXC Technology 基于 Amazon Bedrock 搭建多模型 AI 底座

DXC 在 70 国做 IT 托管服务,把 AI 架构拆成 Bedrock 多模型底座加自建治理层,agent 的编排、权限、审计全走一套 harness。做企业级 agent 落地的开发者值得看它的分层设计,尤其受监管行业的团队。

Primary sourceAWS Machine Learning Blog@Asif FouziSource
How-toBuildersOfficialSelected05:03
AWS shows push-based pipeline to automate Agent Registry updates

The non-obvious point is that registry freshness can be solved with event routing rather than polling jobs or stored credentials, using CloudTrail as the change source of truth. Platform and MLOps engineers building on Amazon Bedrock AgentCore should care, since the same EventBridge-to-Lambda wiring generalises to other agent metadata stores. Teams in multi-account AWS organisations get a reference pattern that works without cross-account secret sharing.

Primary sourceAWS Machine Learning Blog@Prachi KulkarniSource
Chinese content许多已宣布的数据中心项目可能根本不会建成

很多人以为数据中心建得飞快,实际上不少项目连开工都难。这条视频讲清了当地反对、审批拖延和抢电力三个卡点,看懂后你再看到巨额投资新闻就不会照单全收了,做基建相关决策的人更该看看。

Event
techcrunch@Theresa Loconsolo et al. (3)Source
Chinese content播客对谈:AI 智能体趋势与欧洲视角的 2027 年展望

这是一期英文播客对谈,聊的是 AI agents 接下来会怎么发展、欧洲在整个格局里处什么位置,还有对 2027 年的具体预测。如果你平时听的 mostly 是硅谷那边的人聊 agent,换个欧洲从业者的视角挺有意思,音频已上线可以直接听。

kimmonismus@kimmonismusSource

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