Today's
Briefing

Rival Nations Seize On Choke Points to Counter Trump

From Iran to China, President Trump’s global aggression has encouraged other countries to search for new ways to pressure the U.S. economy.

Trump Hails Destruction of Iranian Bridge, Warning ‘More to Follow’

Fars, a semiofficial news agency, said eight people had been killed in the attack and dozens wounded.

Trump Says He’ll End the Iran War in Weeks. What If He Can’t?

President Trump faces the possibility that at the end of his own two-to-three week window for wrapping up the war in Iran, nothing much will have changed.

Trump removes US Attorney General Pam Bondi

Bondi's time as America's top law enforcement officer was overshadowed by the justice department's handling of the Jeffrey Epstein investigation.

特朗普解聘检察长邦迪，因司法部对 Epstein 案件处理失当而受关注。

'Be serious... don't speak every day': Macron criticises Trump approach to Iran war

The comments were an apparent jab at the US leader's sometimes contradictory stance on the US-Israeli war against Iran.

马克龙批评特朗普对伊朗的轻佻言论，呼吁其严肃处理，不频繁发表言论。

Funeral fraudster who kept bodies and ashes was 'living beyond his means', ex-worker says

Robert Bush conned the bereaved while living in luxury and enjoying foreign holidays.

Some on-air claims about Dominion Voting Systems were false, Fox News acknowledges in statement after deal is announced

Fox承认部分Dominion说法不实，随交易公布。

Dominion still has pending lawsuits against election deniers such as Rudy Giuliani and Sidney Powell

Dominion仍对吉安尼、鲍威尔等否认者提起诉讼。

Here are the 20 specific Fox broadcasts and tweets Dominion says were defamatory

• Fox-Dominion trial delay 'is not unusual,' judge says • Fox News' defamation battle isn't stopping Trump's election lies

How to Build a Secure WhatsApp AI Assistant with Arcade and Claude Code

Article URL: https://www.arcade.dev/blog/secure-openclaw-alternative-arcade-claude-code/

Comments U

利用Arcade框架和ClaudeCode实现安全的WhatsApp AI助手

Show HN: SharpSkill – AI writes your code. We teach you to understand it

Gx HN,

AI makes it easy to generate code, but much harder to truly understand it especially under pressure.

SharpSkill simulates real technical interviews : solve problems live, explain your t

ON1 Restore AI Turns Old Family Photos into Grotesque Nightmare Fuel

Article URL: https://petapixel.com/2026/03/25/on1-restore-ai-turns-old-family-photo

Gemma 4: Byte for byte, the most capable open models

Gemma 4: Our most intelligent open models to date, purpose-built for advanced reasoning and agentic workflows.

Gemma 4发布，专为高级推理和智能体工作流打造，成为最强开源模型。

Gemini 3.1 Flash Live: Making audio AI more natural and reliable

Our latest voice model has improved precision and lower latency to make voice interactions more fluid, natural and precise.

Gemini 3.1 Flash Live提升语音模型精度、降低延迟，交互更流畅自然。

Protecting people from harmful manipulation

Google DeepMind researches AI's harmful manipulation risks across areas like finance and health, leading to new safety measures.

谷歌大脑研发AI有害操控风险，推出新安全措施。

How Emotion Shapes the Behavior of LLMs and Agents: A Mechanistic Study

arXiv:2604.00005v1 Announce Type: new Abstract: Emotion plays an important role in human cognition and performance. Motivated by this, we investigate whether analogous emotional signals can shape the

情感信号可塑造大语言模型与Agent的行为模式。

One Panel Does Not Fit All: Case-Adaptive Multi-Agent Deliberation for Clinical Prediction

arXiv:2604.00085v1 Announce Type: new Abstract: Large language models applied to clinical prediction exhibit case-level heterogeneity: simple cases yield consistent outputs, while complex cases produ

单一面板无法通用，需针对各病例自适应多智能体协商预测。

Open, Reliable, and Collective: A Community-Driven Framework for Tool-Using AI Agents

arXiv:2604.00137v1 Announce Type: new Abstract: Tool-integrated LLMs can retrieve, compute, and take real-world actions via external tools, but reliability remains a key bottleneck. We argue that fai

提出开放、可靠、协作的AI工具框架，显著提升大模型工具调用的可靠性。

Two-Stage Optimizer-Aware Online Data Selection for Large Language Models

arXiv:2604.00001v1 Announce Type: new Abstract: Gradient-based data selection offers a principled framework for estimating sample utility in large language model (LLM) fine-tuning, but existing metho

本文提出两阶段优化器感知的在线数据选择方法，提升大语言模型的在线微调效果。

Task-Centric Personalized Federated Fine-Tuning of Language Models

arXiv:2604.00050v1 Announce Type: new Abstract: Federated Learning (FL) has emerged as a promising technique for training language models on distributed and private datasets of diverse tasks. However

研究提出基于任务的个性化联邦微调，解决异构任务聚合导致性能下降的问题。

Evolution Strategies for Deep RL pretraining

arXiv:2604.00066v1 Announce Type: new Abstract: Although Deep Reinforcement Learning has proven highly effective for complex decision-making problems, it demands significant computational resources a

利用进化策略进行深度强化学习预训练，可显著降低资源消耗与参数调优难度。

Converge Bio raises $25M, backed by Bessemer and execs from Meta, OpenAI, Wiz

AI drug discovery startup Converge Bio raised $25 million in a Series A led by Bessemer Venture Partners, with additional backing from executives at Meta, OpenAI, and Wiz.

Converge Bio获Bessemer领投2500万美元A轮，Meta等高管支持。

Meta bought 1 GW of solar this week

The social media company inked three deals in the U.S. to power its data centers and offset its carbon footprint.

Meta本周签署美国三项光伏协议，购入1吉瓦清洁能源，为数据中心减碳。

How one AI startup is helping rice farmers battle climate change

Mitti Labs is working with The Nature Conservancy to expand the use of climate-friendly rice farming practices in India. The startup uses its AI to verify reductions in methane emissions.

MittiLabs与自然保护协会合作，推广印度气候友好水稻种植，AI核实甲烷减排。

[object Object]

[object Object]

抱歉，我无法对“[object Object]”进行总结，因为缺少具体的新闻内容，请提供真实的标题和正文。

[object Object]

[object Object]

抱歉，未提供具体新闻内容，无法进行总结。

[object Object]

[object Object]

内容缺失，无法提供有效总结。请补充新闻标题与内容信息。

‘Uncanny Valley’: Iran’s Threats on US Tech, Trump’s Plans for Midterms, and Polymarket’s Pop-up Flop

In this episode, we discuss Iran’s threats to target US tech firms, gear up for the midterm elections, and get a scene report from the Polymarket pop-up bar in DC.

The Google Pixel 10 Is $150 Off

Amazon marks Google's flagship down to just $650 for 128 GB or $750 for 256 GB.

亚马逊Pixel 10降至650美元（128GB）或750美元（256GB），优惠150美元。

Tech Companies Are Trying to Neuter Colorado’s Landmark Right-to-Repair Law

A bill in Colorado is a glimpse into the future of how corporations are working to limit the freedom people have to make their own fixes and upgrades.

汇聚热爱，「2050 大会」少数派线下活动邀请你来参与

「2050大会」是国内科技圈里一个非常独特的存在。这是由中国工程院院士、阿里云创始人王坚博士，通过杭州市云栖科技创新基金会，发起的一项非营利科技活动。每年4月最后一个完整周末，来自世界各地的科技爱好者 ...查看全文

派晚报：2026 年愚人节产品大赏

假作真时真亦假，明天晚上就放假。查看全文

被咬一口的世界，苹果 50 年的疯狂与奇迹

科技应该怎么改变人类的生活，Apple 花 50 周年给出了自己的答卷。查看全文

科技爱好者周刊（第 390 期）：没有语料，大模型就是智障

这里记录每周值得分享的科技内容，周五发布。

本杂志开源，欢迎投稿。另有

套壳中国大模型撑起500亿美元估值？扒一扒 Cursor 的"套壳"疑云

1、

本周末，有一条最热闹的 AI 新闻，震动了太平洋两岸，连马斯克都关注了。

昨天，AI 编程工具 Cursor 推出了"自己的"模型 Composer 2。

科技爱好者周刊（第 389 期）：未来如何招聘程序员

这里记录每周值得分享的科技内容，周五发布。

本杂志开源，欢迎投稿。另有

[北京] 2026 Bybit 大陆用户注册教程 开卡已放水中！! !

[问与答] 为什么皮夹式纸巾（钱包式）现在很少看到了？

[新手求助] 现在交易所新用户有什么福利？

The March jobs report will be released on Friday. Here's what to expect

The U.S. economy is projected to show job gains of 59,000 in March, with the unemployment rate holding at 4.4%.

三月就业报告将于周五公布，预计新增5.9万岗位，失业率维持在4.4%。

Trump fires Attorney General Pam Bondi

President Trump was said to be unhappy with Pam Bondi's handling of DOJ files about Jeffrey Epstein and the DOJ's failure to prosecute his political enemies.

Iran and Oman drafting protocol to 'monitor' Hormuz Strait traffic: IRNA

The Strait of Hormuz, the vital artery for global oil transit, has been effectively closed since the Iran war started in late February.

伊朗与阿曼拟制定洞庭海交通监管协议，因晚二月伊朗战争导致航道封闭。

AIPI: From Capped Upside To Rangebound Income Opportunity

AIPI从上限突破转为区间震荡，呈现收益机会。

Texas Pacific Land Corporation: Still Too Expensive

德州太平洋土地公司仍被认为贵得不合理。

Workday: SaaSpocalypse Selloff Creates A Golden Opportunity

Workday股价因SaaS板块抛售下跌，但被视为黄金买入机会。

“Liberation Day” has reshaped trade—but not as Donald Trump hoped

In many ways, global commerce has strengthened

Can a country get too rich?

Norway shows the potential pitfalls of uncommon prosperity

挪威的富庶背后隐藏了经济与环境的潜在风险。 （28字）

Coal is back in fashion

An LNG crunch is good news for the world’s dirtiest fuel

9 best silk pajamas, according to extensive testing

After testing silk pajama sets and nightgowns from Eberjey, Quince, Lunya, Clementine, and more, we found the best silk pajamas for women, including washable, luxury, and budget-friendly picks.

根据测试，确定9款最佳丝绸睡衣，涵盖可洗、奢华与亲民选项。

Amazon is selling TurboTax for $56 in time for the approaching deadline

The deadline to file your taxes this year is April 15, and time is running out to get your TurboTax program ordered. Here's how to save now.

Best TVs of 2026

These are the best TVs we've reviewed, whether you want a premium OLED upgrade or a budget-friendly QLED.

2026年最佳电视推荐包括高端OLED和性价比QLED等多款精选。

Federal Reserve Board announces it has made the joint findings with the Office of the Comptroller of the Currency required for the OCC to approve a request by Morgan Stanley Bank, N.A., for an exemption under section 23A of the Federal Reserve Act

Federal Reserve Board announces it has made the joint findings with the Office of the Comptroller of the Currency required for the OCC to approve a request by Morgan Stanley Bank, N.A., for an exempti

美联储完成审查，OCC批准摩根大通国银依据23条A款的豁免申请。（48字）

Federal Reserve Board releases annual audited financial statements

Federal Reserve Board releases annual audited financial statements

美联储公布年度审计财务报表。

Federal Reserve Board issues enforcement actions with former employee of Ally Bank and former employee of Regions Bank

Federal Reserve Board issues enforcement actions with former employee of Ally Bank and former employee of Regions Bank

美联储对前Ally银行、Regions银行员工实施强制纪律处置。

一个周末 + 1100 美元，干完 5 人 6 个月的活：Cloudflare 用 AI“复刻”Next.js，已跑进生产环境

从数据留底到隐身进开源，Claude Code 泄露的代码里，处处写着：这家公司人品不行

AI 原生时代，如何提供可用、易用、高效的算力服务？

36氪晚报｜AI大模型产品“奇趣”完成千万级融资；阶跃星辰上线Step 3.5 Flash 2603模型；沃尔沃汽车第一季度全球销量同比下降11%

大公司：

滴滴租车推出“无车必赔”机制，无车率下降15%

为提升履约确定性，滴滴租车推出“无车必赔”机

轻舟、元戎秘交上市资料，和Momenta冲刺港股IPO | 36氪独家

泡泡玛特无需下一个LABUBU

3月25日，泡泡玛特交出了2025年的答卷。过去一年，泡泡玛特的收入为人民币371.2亿元，同比增长185%，净利润为130.1亿元，同比增长293%。

围绕泡泡玛特的讨论再次沸腾起来，只是相较于去年中报发布之际的欢欣鼓舞，此次与泡泡玛特亮眼财报成绩一并而至的，是资本市场的负面情绪。

对此泡泡玛特没有坐以待毙，连续选择四个交易日斥资12亿港元

不只是押注利率，Rocket的股票也已蓄势待发，准备起飞

Rocket 正在不断扩大市场份额，并加深其在抵押贷款相关各个方面的参与力度。

认识新一代股息贵族

这六家公司已连续至少15年提高股息，使得持有它们的理由更为充分。

必需消费品板块被抛弃了，高露洁以及另外5只股票值得关注

高露洁-棕榄预计今年将在北美实现增长，部分将得益于其口腔护理和皮肤护理产品的新版本。

Import AI 451: Political superintelligence; Google's society of minds, and a robot drummer

Are there any genies that can be put back in the bottle?

AI进展能否被重新封印？

Import AI 450: China's electronic warfare model; traumatized LLMs; and a scaling law for cyberattacks

How will timeless minds value time?

AI时代的电子战模型、受创模型及网络攻击规模规律。

ImportAI 449: LLMs training other LLMs; 72B distributed training run; computer vision is harder than generative text

Will AI cause a political interregnum

AI可能引发政治空白期。

AI Weekly Issue #477: Jensen Huang says we've achieved AGI. The benchmarks say 0.37%.

💡 Insights AI is superhuman at exams but can't figure out a simple game. ARC-AGI-3 gave frontier models interactive environments with no rules and no goals — just figure it out. Humans solve 100%. Th

AI Weekly Issue #476: Weekly Intelligence Briefing: Tech, AI & Policy

A federal judge dismantled the Pentagon's attempt to blacklist Anthropic. OpenAI pulled the plug on Sora, writing off $15 million per day in inference burn. And a data leak revealed Anthropic's next-g

法官阻止五角大楼封禁Anthropic；OpenAI停掉Sora并削损失；Claude泄漏新模型。

AI Weekly Issue #475: 100 years from now : The Case for Artificial Stupidity

This is 100 Years From Now, a weekly series. Once a week, we skip ahead a century and imagine ordinary life in a world that's had a hundred years to absorb the things we're only beginning to build. No

OpenAI acquires TBPN, the buzzy founder-led business talk show

TBPN, Silicon Valley's cult-favorite tech podcast, will operate independently, even as it's overseen by chief political operative Chris Lehane.

OpenAI收购TBPN科技播客，将独立运营，由Lehane监督。

Microsoft takes on AI rivals with three new foundational models

MAI released models that can transcribe voice into text as well as generate audio and images after the group's formation six months ago.

微软推出三款多模态基础模型，可转写语音并生成音频、图像。

Google now lets you direct avatars through prompts in its Vids app

Google is adding a way to customize and instruct avatars for video creation in the Vids app.

Google Vids新增自定义并指挥角色的指令功能。

Gemma 4: Byte for byte, the most capable open models

Gemma 4: Our most intelligent open models to date, purpose-built for advanced reasoning and agentic workflows.

Gemma 4发布，专为高级推理和智能体工作流打造，成为最强开源模型。

Gemini 3.1 Flash Live: Making audio AI more natural and reliable

Our latest voice model has improved precision and lower latency to make voice interactions more fluid, natural and precise.

Gemini 3.1 Flash Live提升语音模型精度、降低延迟，交互更流畅自然。

Protecting people from harmful manipulation

Google DeepMind researches AI's harmful manipulation risks across areas like finance and health, leading to new safety measures.

谷歌大脑研发AI有害操控风险，推出新安全措施。