【AI 英文奏折】01-14 2026

👤 (2) X 上的 Rohan Paul：“This paper finds the job slowdown in AI exposed work began before ChatGPT, not after it. The authors test the popular story that ChatGPT caused the slump using unemployment claims, LinkedIn job histories, and about 3M course outlines. They call a job AI exposed when a big chunk https://t.co/Ce2TcXIqTK” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 14:43 >

AI岗位衰退早于ChatGPT >

研究通过失业、招聘及课程数据，发现AI相关岗位放缓始于2022年初，早于ChatGPT发布。

“ This paper finds the job slowdown in AI exposed work began before ChatGPT, not after it.

The authors test the popular story that ChatGPT caused the slump using unemployment claims… ”

💡 AI 洞察：技术变革的经济影响往往在关键产品亮相前已悄然显现。

阅读原文 →

👤 (2) X 上的 Sundar Pichai：“Today, Veo 3.1 Ingredients to Video is getting more expressive + portrait mode has arrived (we heard you!) You can now create vertical clips with livelier movement and better control directly in the @Geminiapp, @YouTube & Google Vids. All this plus state-of-the-art upscaling to https://t.co/vCvGEelldS” > @sundarpichai

深度观察者 | 2026-01-14 02:27 >

Veo 3.1升级发布 >

Veo视频生成功能增强，新增竖屏模式与高清升级，多平台可用。

“ Today, Veo 3.1 Ingredients to Video is getting more expressive + portrait mode has arrived (we heard you!)

You can now create vertical clips with livelier movement and better cont… ”

💡 AI 洞察：此次更新通过竖屏模式与高清化，精准回应了短视频创作的主流需求，展现了AI视频工具向实用性与平台生态融合的深化。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Google released a major update to their open-source MedGemma model with improved medical imaging support. Its mainly about expanding MedGemma from 2D medical images to higher-dimensional imaging, especially full 3D CT and MRI volumes. It does this by letting developers pass https://t.co/e6E5rJemSP” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-14 05:28 >

MedGemma升级医学影像分析 >

谷歌开源模型升级，支持3D影像、纵向比对与解剖定位，提升医疗文档理解。

“ Google released a major update to their open-source MedGemma model with improved medical imaging support.

Its mainly about expanding MedGemma from 2D medical images to higher-dim… ”

💡 AI 洞察：从2D到3D全景分析，标志着AI医疗正从“看单张片子”迈向“解构全病程”。

阅读原文 →

👤 (2) X 上的 Philipp Schmid：“New Gemma model! @GoogleDeepMind MedGemma 1.5 4B is an updated 4B variant expanding support for 3D imaging, longitudinal analysis, and localization while improving text accuracy over MedGemma 1. – Interprets 3D volumes for CT, MRI, and whole-slide histopathology. – Achieves 38% https://t.co/6kZ0wg2ilA” > @_philschmid

深度观察者 | 2026-01-14 05:09 >

MedGemma 1.5 4B发布 >

谷歌发布医疗AI模型，支持3D影像、纵向分析及定位，文本理解准确率大幅提升。

“ New Gemma model! @GoogleDeepMind MedGemma 1.5 4B is an updated 4B variant expanding support for 3D imaging, longitudinal analysis, and localization while improving text accuracy ov… ”

💡 AI 洞察：该模型通过多模态能力整合，将通用大模型技术向专业、复杂的临床工作流纵深推进。

阅读原文 →

👤 (2) X 上的 Amira Zairi：“Prompt share: Product Photography. 💬Prompt: Studio shot of a [PRODUCT], elegantly positioned on a [background], soft ambient shadows, smooth gradient backdrop, high-key lighting, shallow depth of field, ultra-sharp focus on the subject, subtle reflections, minimal aesthetic, https://t.co/kDVlcqqiyy” > @azed_ai

深度观察者 | 2026-01-13 18:57 >

产品摄影提示分享 >

分享高端产品摄影的提示词，包含布光、背景与细节要求。

“ Prompt share: Product Photography.

Prompt:
Studio shot of a [PRODUCT], elegantly positioned on a [background], soft ambient shadows, smooth gradient backdrop, high-key lighting, s… ”

💡 AI 洞察：提示词通过精准的视觉要素堆叠，构建了可复制的商业摄影工业化模板。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Anthropic launches Cowork, a Claude Code-like for general computing. Users share a specific folder, and Claude can read, edit, or create files inside it. It is a research preview for Claude Max subscribers, with optional connectors, skills, and Claude in Chrome for browser https://t.co/rbSSYPr1t0” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 08:26 >

Anthropic推出Cowork协作式AI代理 >

Claude推出文件操作代理模式，支持规划与执行任务，面向高级订阅用户。

“ Anthropic launches Cowork, a Claude Code-like for general computing.

Users share a specific folder, and Claude can read, edit, or create files inside it.

It is a research preview… ”

💡 AI 洞察：这是AI从对话工具迈向自主操作系统的关键一步，但安全与可控性成为新挑战。

阅读原文 →

👤 (2) X 上的 Google Gemini：“With the latest Veo updates, you now have even more creative control over how you prompt and what you make. Join us on Discord for our first community event of the year as Ale from the Gemini Product team gives you the director’s cut on the latest improvements, plus a chance to” > @geminiapp

Sharing insights to use AI in practical ways for you and your business • Follow me to learn and master AI & Tech tools • AI Educator & Writer @theprohumanai | 2026-01-14 04:50 >

Veo更新增强创作控制 >

Veo更新提升提示与生成控制，社区活动详解新功能。

“ With the latest Veo updates, you now have even more creative control over how you prompt and what you make.

Join us on Discord for our first community event of the year as Ale fr… ”

💡 AI 洞察：官方通过社区活动深入解读更新，旨在增强用户粘性与产品认知。

阅读原文 →

👤 (2) X 上的 Bindu Reddy：“🚨 Announcing An AI Agent That Combines Software and QA Engineering The best AI agents won’t just replace software engineers; they will replace entire software teams They will combine app building, testing, debugging, and monitoring to create a scalable production system https://t.co/2VCR2nbkfa” > @bindureddy

深度观察者 | 2026-01-14 04:42 >

AI代理将取代整个软件团队 >

Abacus AI推出DeepAgent，整合开发、测试与运维，实现自动化构建与扩展。

“ Announcing An AI Agent That Combines Software and QA Engineering

The best AI agents won’t just replace software engineers; they will replace entire software teams

They will comb… ”

💡 AI 洞察：AI代理正从替代个体工程师转向重构整个软件生产流程，标志着软件开发进入全栈自动化新阶段。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Google has updated its Veo 3.1 AI video model with a stronger Ingredients to Video workflow. Ingredients just means the input “parts” you give the model up front, usually a subject image, a background image, and a style or look reference. Veo uses those images as anchors while https://t.co/yN5uZGlbiE” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-14 04:11 >

Google升级Veo视频AI模型 >

Google更新Veo 3.1，通过“素材到视频”工作流提升视频生成的一致性与质量。

“ Google has updated its Veo 3.1 AI video model with a stronger Ingredients to Video workflow.

Ingredients just means the input “parts” you give the model up front, usually a subjec… ”

💡 AI 洞察：此次升级通过锚定输入素材实现跨镜头一致性，标志着AI视频生成从片段创作向实用工作流的关键转变。

阅读原文 →

👤 (2) X 上的 Santiago：“The hard part of building is no longer writing code. Code is easy. Nobody wins by writing more code or doing it faster. If you want to win, you need better thinking. There are a ton of startups focusing on using AI to write more code. I think they are missing the plot. We https://t.co/DSTdbUNA6j” > @svpino

深度观察者 | 2026-01-14 04:10 >

AI时代胜在思考而非编码 >

构建软件的核心挑战已从写代码转向深度思考和决策。

“ The hard part of building is no longer writing code.

Code is easy.

Nobody wins by writing more code or doing it faster. If you want to win, you need better thinking.

There are a… ”

💡 AI 洞察：当AI能轻易生成代码时，真正的竞争优势将来自人类的问题定义与战略思考能力。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Today’s edition of my newsletter just went out. 🔗 https://t.co/Sj2MMROnlP 🚨 Apple finally chooses Google’s Gemini for its AI 🧪 Anthropic Unveils “Claude for Healthcare” to Help Users Understand Medical Records 📡 1X is presenting a “video-to-action” world-model approach https://t.co/VgbEfru0Gg” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-14 04:04 >

苹果联手谷歌Gemini布局AI >

苹果选择谷歌Gemini，Anthropic推出医疗与办公AI，1X展示机器人新模型。

“ Today’s edition of my newsletter just went out.

https://rohan-paul.com/p/apple-finally-chooses-googles-gemini…

Apple finally chooses Google’s Gemini for its AI

Anthropic Unve… ”

💡 AI 洞察：科技巨头竞合揭示AI生态从规模扩张转向垂直场景深耕的战略转折。

阅读原文 →

👤 (2) X 上的 Google AI：“Veo update! Here’s what’s rolling out: — Vertical formats with a more expressive model *and* improved visual consistency across characters, objects, and backgrounds in Veo 3.1 Ingredients to Video — State-of-the-art video upscaling to 1080p and 4K across all Veo models —” > @googleai

深度观察者 | 2026-01-14 01:04 >

Veo 3.1更新发布 >

Veo推出竖屏视频、画质一致性增强、支持4K超分及AI生成视频验证功能。

“ Veo update! Here’s what’s rolling out:

— Vertical formats with a more expressive model *and* improved visual consistency across characters, objects, and backgrounds in Veo 3.1 Ing… ”

💡 AI 洞察：谷歌正通过提升画质、适配社交格式及强化认证，推动AI视频从“可生成”迈向“可用可信”。

阅读原文 →

👤 (2) X 上的 Ksenia_TuringPost：“Most companies treat an IPO as proof of success. In China’s AI market, it often means survival. @MiniMax_AI went public in Hong Kong on January 9, 2026. Its shares surged 109% on the first trading day – the strongest Hong Kong tech IPO performance in four years. It also became https://t.co/8zcbfFS0gQ” > @theturingpost

深度观察者 | 2026-01-13 06:50 >

MiniMax上市：AI生存战 >

MiniMax港股上市首日股价翻倍，IPO在中国AI市场是生存手段而非成功证明。

“ Most companies treat an IPO as proof of success. In China’s AI market, it often means survival.

@MiniMax_AI went public in Hong Kong on January 9, 2026. Its shares surged 109% on … ”

💡 AI 洞察：在资本与技术双重压力下，中国AI企业正将上市从荣耀勋章转变为续命氧气。

阅读原文 →

👤 (2) X 上的 @levelsio：“Does anyone know anybody who could reverse engineer or edit a Windows 3.1 VGA driver so that I can have it run at portrait instead of landscape? All the VGA drivers are like 800×600 or 1024×768 But it’d be nice to flip that to 600×800 so it become more useful to run on” > @levelsio

深度观察者 | 2026-01-14 03:46 >

Windows 3.1竖屏驱动求助 >

用户寻求修改Windows 3.1 VGA驱动，以实现竖屏显示用于智能手机。

“ Does anyone know anybody who could reverse engineer or edit a Windows 3.1 VGA driver so that I can have it run at portrait instead of landscape?

All the VGA drivers are like 800×6… ”

💡 AI 洞察：复古技术爱好者通过改造旧驱动来适配现代移动设备，体现了技术应用的跨时代创意。

阅读原文 →

👤 (2) X 上的 Aakash Gupta：“California’s inflation-adjusted GDP per capita grew about 35% in the same period. So spending growth outpaced economic growth by roughly 5x. Where did the money go? The state spent $24 billion on homelessness from 2019-2023. The homeless population went UP 30,000 to 181,000.” > @aakashgupta

深度观察者 | 2026-01-14 03:42 >

加州支出激增与效益缺失 >

人均GDP增35%，但支出增速5倍于此，无家者与路况问题随支出恶化。

“ California’s inflation-adjusted GDP per capita grew about 35% in the same period. So spending growth outpaced economic growth by roughly 5x.

Where did the money go?

The state spe… ”

💡 AI 洞察：缺乏问责的支出增长，是公共治理失效的核心症结。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Meta launched Meta Compute, a program to expand AI infrastructure. Meta frames the goal as personal superintelligence, where machines outthink humans. It puts data centers, suppliers, and capacity planning into 1 group. Zuckerberg says the target is 10s of gigawatts this https://t.co/nuEK0yV9xU” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 07:13 >

Meta启动AI基建计划 >

Meta整合资源，以千兆瓦级计算能力推动AI超级智能发展

“ Meta launched Meta Compute, a program to expand AI infrastructure.

Meta frames the goal as personal superintelligence, where machines outthink humans.

It puts data centers, suppl… ”

💡 AI 洞察：科技巨头以能源密集型基建追逐AGI，将引发资源与伦理的深层博弈。

阅读原文 →

👤 (2) X 上的 FlowbyGoogle：“Our new state-of-the-art video upscaling allows you to download videos in 1080p and, for the first time, in 4K! 💡 Pro Tip: Want the absolute highest fidelity? Combine our upscaling tools. 1. Download a generated image in 4K 2. Upload it as an Ingredient 3. Turn it into a https://t.co/uGQVy8CvVw” > @flowbygoogle

深度观察者 | 2026-01-14 02:39 >

AI视频超分实现4K下载 >

新视频超分技术支持1080P和4K下载，结合图像上传可优化输出质量。

“ Our new state-of-the-art video upscaling allows you to download videos in 1080p and, for the first time, in 4K!

Pro Tip: Want the absolute highest fidelity? Combine our upscaling… ”

💡 AI 洞察：技术通过“优质输入产生优质输出”的逻辑，将静态图像超分与动态视频生成链路打通，形成质量增强闭环。

阅读原文 →

👤 (2) X 上的 Aakash Gupta：“My take is different from most people covering this. Cowork looks like a product release. It’s actually a distribution play. Let me explain. Anthropic built Cowork in 10 days using Claude Code. That tells you everything about the difficulty of shipping this feature. The hard” > @aakashgupta

深度观察者 | 2026-01-14 02:44 >

Anthropic的Cowork是渠道战略 >

Claude用户用命令行处理日常任务，Cowork通过友好界面解决产品分发问题。

“ My take is different from most people covering this.

Cowork looks like a product release. It’s actually a distribution play.

Let me explain.

Anthropic built Cowork in 10 days us… ”

💡 AI 洞察：企业AI应用的关键往往不是技术突破，而是降低使用门槛和精准把握市场时机。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“Meta is discussing job cuts in Reality Labs, the unit behind its metaverse and wearables. NYT reports. Around 10% of the roughly 15,000-person division may be cut. Reality Labs has lost more than $60B since 2020 under Zuckerberg’s metaverse bet. Reality Labs ships Quest https://t.co/YzwPgKLxOV” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 07:10 >

Meta或裁员元宇宙部门 >

Meta拟裁撤约1500人，其元宇宙部门已累计亏损超600亿美元。

“ Meta is discussing job cuts in Reality Labs, the unit behind its metaverse and wearables.

NYT reports.
Around 10% of the roughly 15,000-person division may be cut.

Reality Labs… ”

💡 AI 洞察：在巨额亏损压力下，Meta的元宇宙战略正经历残酷的现实检验与收缩。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“This paper teaches LLMs to avoid showing personally identifiable information (PII) in chain of thought, without hurting answers much. The problem is that chain of thought, meaning the model’s step by step notes, can spill names, health details, or account data even if the final https://t.co/prE7PLSwE5” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 10:28 >

LLMs防范思维链隐私泄露 >

论文提出用占位符替代敏感信息，防止思维链泄露个人隐私

“ This paper teaches LLMs to avoid showing personally identifiable information (PII) in chain of thought, without hurting answers much.

The problem is that chain of thought, meaning… ”

💡 AI 洞察：通过前置干预而非后置清洗，从根本上重塑了隐私保护的技术路径。

阅读原文 →

👤 (2) X 上的 Rohan Paul：“New MIT paper proposes using a hypergraph knowledge map so LLM agents can follow real multi item links in papers and generate better grounded scientific hypotheses. A hypergraph lets 1 link connect many concepts, giving agents a clearer path from evidence to hypothesis. LLMs https://t.co/ZIJYBwJGd3” > @rohanpaul_ai

Compiling in real-time, the race towards AGI. The Largest Show on X for AI. \uD83D\uDDDE️ Get my daily AI analysis newsletter to your email \uD83D\uDC49 https://t.co/6LBxO8215l | 2026-01-13 09:29 >

超图知识图谱提升LLM科学假设生成 >

MIT研究用超图知识图谱替代传统知识图谱，使LLM能追踪论文中多元概念关联，生成更可靠的科学假设。

“ New MIT paper proposes using a hypergraph knowledge map so LLM agents can follow real multi item links in papers and generate better grounded scientific hypotheses.

A hypergraph l… ”

💡 AI 洞察：该研究通过超图结构保留科学主张的原始群体语义，从本质上解决了传统知识图谱因二元链接拆分导致的语义丢失和幻觉问题。

阅读原文 →