AI News

AI changes every day,
catch up on it all on one page

The latest updates from AI companies, research institutions, and media around the world, automatically selected for you and updated around the clock.

Updated 11/16 18:05 (UTC+8) · refreshed hourly

Today11/16 · Monday6 items

Tencent Hunyuan: ResearchModelsAI Score73

Tencent Hunyuan Releases Hy4 flash: A Cost-Effective Everyday Model with 224B Total Parameters and a 1M Context Window

Tencent Hunyuan introduces Hy4 flash, with 224B total parameters (including 64B Engram parameters) and 12B activated parameters. It supports multimodal input and a 1M context window. API pricing is much lower than Hy4 (0.5 yuan for input and 2 yuan for output per million tokens).

Why it's recommended:The original article provides the architecture, pricing table, cost comparisons, and evaluation data across multiple business scenarios, helping readers assess cost-effective options for everyday tasks.

Also reported by 3 other sources
Arena.ai@arenaModelsAI Score82

Qwen4-Max tops the Chinese model rankings in Agent Arena, with Alibaba rising to No. 14 overall

The Arena rankings show Qwen4-Max at No. 14 overall in Agent Arena with a +7.5% net improvement, making it the top Chinese model. Alibaba consequently becomes the highest-ranked Chinese lab and No. 5 overall, behind only Anthropic, Google, OpenAI, and Meta.

Why it's recommended:Based on 14.7k real Agent sessions, the rankings provide placements and a breakdown of four signals, allowing readers to compare Qwen4-Max's actual performance with other Chinese models.

Also reported by 1 other source
Qwen@Alibaba_QwenModelsAI Score73

Alibaba Qwen Releases Qwen4 Max, Scoring 9 Points Higher Than Qwen3.8 Max on the Intelligence Index at Lower Cost

Alibaba Qwen announced the release of Qwen4 Max. According to Artificial Analysis, it scores 43 on the Intelligence Index, 9 points higher than Qwen3.8 Max (xhigh) and 3 points behind Kimi K3.1 Pro. Its cost per task is $3.27, about one-third that of Qwen3.8 Max, while output token usage is reduced by 39%. It has a 1M-token context window and is priced at $1.50/$4.50 per 1M input/output tokens. It is available through Alibaba's first-party API.

Why it's recommended:The original article uses Artificial Analysis data to compare Qwen4 Max's scores and cost changes, helping readers assess its value relative to its predecessor and competitors.

Also reported by 1 other source
Qwen:Blog RetrievalModelsAI Score80

Qwen releases Qwen4-Max: a 2.4T-parameter backbone, a million-token context window, and long-horizon Agent capabilities

The Qwen team released Qwen4-Max, based on the new Qwen4 architecture, with 2.4T backbone parameters, 71B parameters activated per token, and 328B N-gram Embedding parameters. It has native vision capabilities and a 1 million-token context window.

Why it's recommended:The official release also provides architectural details, cost comparisons, and real-world industry cases, allowing readers to assess the actual changes in Qwen4-Max compared with the previous generation.

Reported by 5 additional sources
Apple:NewsroomPapersAI Score61

Apple partners with OpenEvidence to launch the Apple Health AI Usage Study exploring health AI use

Apple announced the launch of the Apple Health AI Usage Study in the United States, partnering with OpenEvidence to study how users interact with conversational health AI tools.

Why it's recommended:The original article provides the study scale, participation requirements, and privacy arrangements, offering insight into how Apple and OpenEvidence are collaborating to evaluate the real-world use of health AI.

Also reported by 2 other sources
WeChat Official Account: Digital Life KazikeTutorialsAI Score72

An AI-made 1.4M pure-HTML 3D game goes viral, as Kazik breaks down Three.js procedural generation principles and provides a prompt template

The author breaks down the 1.4M pure-HTML 3D game that recently spread rapidly in online groups, noting that about 700KB consists of Three.js bundled into the file, while everything else is procedurally generated by code, with no images, models, or recordings.

Why it's recommended:The original article breaks down the size composition and procedural generation principles of the 1.4M pure-HTML 3D game, and provides a directly reusable prompt template and self-checking iteration methods.

Yesterday11/15 · Sunday2 items

TechCrunch:AIIndustryAI Score84

Musk and other executives co-sign a petition demanding that Nvidia reinstate Huang Renxun as CEO

The petition demanding that Nvidia reinstate ousted CEO Huang Renxun had received more than 115,000 signatures by Saturday afternoon, including 5,965 current employees, 1,108 former employees, and 63,063 shareholders. Musk, Replit CEO Amjad Masad, and Epic Games CEO Tim Sweeney have all co-signed it.

Why it's recommended:The petition's breakdown of signatories and background to the ouster can help readers understand the reactions from various parties and the subsequent direction following changes in Nvidia's management.

The Decoder:AI NewsIndustryAI Score78

Anthropic CEO Amodei acknowledges huge uncertainty next year ahead of IPO, with compute representing the company's deepest business dilemma

At a JPMorgan investor event, Anthropic CEO Dario Amodei said there is huge uncertainty about how next year will unfold. Compute is the deepest business dilemma facing the company: compute must be purchased one to two years in advance; buying too little affects customers and revenue, while buying too much could lead to excessive financial expansion.

Why it's recommended:On the eve of a record IPO, Amodei bluntly said that investment in compute is the company's deepest business dilemma and outlined specific approaches involving gross margins and cash buffers.

Also reported by 1 other source

11/14 · Saturday17 items

The Verge:AIIndustryAI Score84

Huang Renxun leaves Nvidia: the era of visionary founder CEOs comes to an end

The Verge's feature looks back at Huang Renxun's departure from Nvidia and examines what the departure of the founder who led the first $5 trillion company, accounting for about 6% of the S&P 500, means. The article reviews Huang Renxun's key decisions, from founding Nvidia to betting on CUDA, and combines the views of multiple analysts to assess the company's roadmap and risks after his departure.

Why it's recommended:The article reviews Huang Renxun's key decisions, from founding Nvidia to betting on CUDA, and combines the views of multiple analysts to assess the company's roadmap and risks after his departure.

Nathan Lambert:InterconnectsOpinionsAI Score80

Nathan Lambert analyzes the vacuum in the U.S. open-source model ecosystem after Jensen Huang's resignation

Nathan Lambert analyzes the impact of Nvidia CEO Huang Renxun's resignation on the US open-source model ecosystem, arguing that Nvidia's financial support for open-source models is very likely to dry up with Huang Renxun's departure. The company's primary goal is to stabilize the situation and deliver to customers.

Why it's recommended:The author has long tracked the open-source model ecosystem and provides a comprehensive analytical framework covering Nvidia's internal support mechanisms, regulatory risks, and the issue of the United States' absence.

Cursor BlogIndustryAI Score72

Cursor discloses the results of its security incident investigation; malicious Claude Code extensions affected up to 5,000 users

Cursor releases a complete investigation into the security incident: attackers used exposed Fireworks access keys to read coding agent traces used for model training, exposing API credentials belonging to fewer than 0.01% of customers.

Why it's recommended:The official disclosure provides a complete account of the attack chain, affected scope, and response measures, allowing readers to assess whether they need to rotate credentials or reinstall the extension.

Also reported by 1 other source
Rohan Paul@rohanpaul_aiIndustryAI Score77

Federal Reserve financial stability report: AI becomes Wall Street's top financial risk

The Federal Reserve released a financial stability report. A survey by the Federal Reserve Bank of New York found that 87% of market participants listed AI as their top financial risk, up from 50% in the spring. The report said market enthusiasm for superintelligence has driven up stock valuations; AI infrastructure debt exceeded $680 billion as of October, about 5 times the full-year 2023 level and approximately 31% of investment-grade corporate borrowing; and the largest banks' exposure to big tech, data center construction, and financing vehicles approached $1 trillion.

Why it's recommended:The original article compiles survey data from the Federal Reserve’s Financial Stability Report identifying AI as the top risk, along with the scale of AI infrastructure debt, helping readers understand AI financing risks from a regulatory perspective.

TechCrunch:AIIndustryAI Score88

Jensen Huang’s departure could leave a major gap in AI startup financing, worrying investors and founders

Nvidia CEO Jensen Huang has officially resigned. The venture capital community believes his successor, Colette Kress, will be more conservative, making it difficult to sustain his pace of investing tens of billions of dollars annually, including participation in 105 rounds so far this year.

Why it's recommended:The original article brings together views from multiple investors and founders, outlining the potential capital shortfall and alternative landscape that AI startups may face in fundraising after Jensen Huang’s departure.

Rohan Paul@rohanpaul_aiIndustryAI Score78

NVIDIA Vice President Bryan Catanzaro clarifies that a letter calling for Jensen’s reinstatement was forged

Rohan Paul reposted a statement from Bryan Catanzaro, NVIDIA’s vice president of applied deep learning research: A letter circulating online called for the NVIDIA board to reinstate Jensen Huang and claimed to have his involvement. Catanzaro explicitly said the letter was forged and that he would not have signed it even if it were genuine. He said he was saddened by Jensen’s departure from NVIDIA while fully supporting the board, adding that he had worked with Colette for many years and fully supported her.

Why it's recommended:The reposted clarification from NVIDIA Vice President Bryan Catanzaro explains that the circulating joint letter was forged, helping readers verify whether the rumor is genuine.

Rohan Paul@rohanpaul_aiIndustryAI Score84

Dario Amodei responds to market turmoil: Before Anthropic’s IPO, he says he does not care much about the timing

At a JPMorgan event in New York, Dario Amodei spoke with Jamie Dimon and said he was unfazed by the market turmoil during Anthropic’s IPO roadshow, adding that he did not care much about the timing of the listing or market conditions.

Why it's recommended:The original article compiles key figures on the IPO size, cash reserves, and computing procurement strategy, helping readers understand the financial logic behind Anthropic’s listing.

Google Research:BlogProductsAI Score78

Google showcases Antigravity /teamwork multi-agent collaboration results and makes them available to researchers

Google Research describes how its teams and academic partners used Antigravity’s /teamwork multi-agent workflow together with the Gemini 4 Argon, Carbon, and Hydrogen models to achieve multiple results in information theory, formal verification, molecular structure, kernel optimization, and other fields.

Why it's recommended:The original article presents concrete results from the /teamwork multi-agent workflow in mathematical proofs, formal verification, and system optimization, enabling readers to assess its applicability to their own research or engineering work.

Ado@adocompleteProductsAI Score66

Claude Duet sends invitations to all US Pro, Max, and Ultra subscribers on the waiting list

Anthropic’s Claude Duet update: As of yesterday, all US Pro, Max, and Ultra subscribers who were on the waiting list had received invitations. The waiting-list page is https://claude.com/product/claude-duet. The research preview includes quota resets: 3 for Pro and 1 for Max and Ultra. The company plans to expand access to more countries outside the US soon and will continue collecting feedback.

Why it's recommended:The original article provides the Claude Duet waiting list’s availability, quota reset counts, and international plans, allowing users to understand their available usage quota.

Newcomer in-depth news articleIndustryAI Score84

On-site observations from the Cerebral Valley AI Summit on the eve of Nvidia’s crisis and Anthropic’s IPO

Newcomer reports from the Cerebral Valley AI Summit in San Francisco: Jensen Huang resigned after the disclosure of a second extramarital affair, while Reuters reported a third that same day. Attendees generally believed Nvidia would face no immediate problems but that its ecosystem was under pressure; the Databricks CEO said it was a blow to the entire ecosystem.

Why it's recommended:The firsthand account places Nvidia’s crisis and Anthropic’s IPO in the same setting, offering a direct view of industry leaders’ immediate assessments of the impact on the ecosystem.

The Decoder:AI NewsIndustryAI Score84

WSJ reports that Huang Renxun’s removal was driven more by Nvidia’s high-risk businesses than his personal life

The Decoder, citing a WSJ report, analyzes how Huang Renxun’s removal was driven more by Nvidia’s high-risk businesses than rumors about his personal life. Under his leadership, Nvidia provided tens of billions of dollars in financial guarantees to AI cloud providers and data center developers in exchange for a share of revenue. The largest guarantee involved $250 billion for OpenAI’s Ohio data center. After the news was announced, the stock fell 5%, and interim CEO Colette Kress cut the commitment in half.

Why it's recommended:The report cites WSJ details to examine the guarantee-backed credit, customer concentration, and other operational risks behind Huang Renxun’s departure, helping readers see the company-level causes beyond the personal rumors.

Claude:BlogTutorialsAI Score62

Anthropic explains the five core architectural primitives of always-on agents

Anthropic published a blog post breaking down the architecture behind always-on agents. These agents are stateful and capable of acting proactively. Their rise has been enabled by improvements in model judgment and falling intelligence costs; for example, on long-horizon coding tasks, Opus 5.5 costs 78% less per completed task than Opus 5.

Why it's recommended:Anthropic breaks always-on agents down into five practical architectural primitives and provides implementation details and best practices, enabling readers to build their own solutions accordingly.

Load more

Finished reading the news? Let’s talk about your business.

Leave your contact details, and I’ll get in touch with you personally.

WECHAT

Add me on WeChat

Press and hold or select the WeChat ID below to copy it.