Live

AI Trending

The 10 things the AI world talked about most in the last 48 hours

15 Nov, 01:00 updated · Sorted by discussion heat

NO.01↓ 22%

Huang Renxun resigns as NVIDIA CEO, CFO Kress takes interim role

NVIDIA's board has approved the resignation of co-founder Jen-Hsun Huang from the positions of President, CEO, and Director, effective immediately; Executive Vice President and CFO Collette Kress appointed as interim CEO and continues to hold the position of CFO until a permanent successor is found. Board Chairman Harvey Jones stated that the resignation occurred after the board launched an investigation into multiple extramarital affairs involving Huang with a direct subordinate and a person accused of being a Chinese spy. The news triggered a sharp decline in global AI concept stocks, with NVIDIA's stock price falling 8.2% in the morning, with its market value falling below $400 billion for the first time since July 2025. The U.S. Department of Justice is conducting multiple investigations into NVIDIA, including one involving antitrust, and two others involving Megaspeed and Supermicro Computers; currently, there is no indication that the Department of Justice plans to criminally prosecute NVIDIA or Huang. Huang's departure comes as Anthropic is advancing its IPO: its subscription demand is about 6 times the number of shares available for sale, the company and its co-founders plan to raise up to $64.4 billion through the IPO and $26 billion through private placement, with a target valuation of about $1.4 trillion, NVIDIA plans to subscribe to about $10 billion in the private placement.

LatestHuang Renxun's departure from NVIDIA marks the end of the visionary showman founder-CEO era.

The Verge · AI, TechCrunch · AI and 6 more outlets·8 voices

Heat index

NO.02↓ 16%

Google Gemini 4 Carbon has expanded from cybersecurity to the public

On November 10, Google expanded Gemini 4 Carbon from a limited release to all paid developers and enterprises, and gradually opened it to Google AI Ultra subscribers. Two weeks ago, the model was only open to some cybersecurity defense personnel and trusted partners of the Fairwind program. Google claims Carbon is its strongest model for cybersecurity capabilities and long-range intelligent workflows, ranking first with 59 points on the Artificial Analysis Intelligence Index v5.0. The pricing follows the launch discount of Argon: $2 per million input tokens, $10 per million output tokens, with 95% off for cached input; according to Google AI Studio documentation, the discount will last until November 30, 2026, and then revert to $4 per million input tokens, $20 per million output tokens. Google claims Carbon complements rather than replaces Argon. The scope of opening is still limited: The Verge said it is unclear when lower-tier subscribers or free users can use it; Google claims the improvements of Carbon in cybersecurity include a 200% increase in the probability of generating bug-free code compared to Argon, all of which are Google's statements.

9 outlets · 9 voices

NO.03↓ 16%

OpenAI has launched ChatGPT Wallet for peer-to-peer payments.

OpenAI launched ChatGPT Wallet, allowing ChatGPT Work, Codex, and dots agents to complete purchases under user authorization. After entering bank card information once, users can let dot agents find products; each transaction requires security verification such as Face ID or fingerprint, and first displays the product, total amount, and selected card. OpenAI states that agents will not access real card numbers, and the payment network will generate dedicated payment tokens for limited use. Existing card rewards and bank protection still apply. This feature is initially only open to ChatGPT Pro subscribers aged 18 or over in the United States, requiring a U.S. billing address and eligible U.S.-issued bank cards (American Express, Chase, Mastercard, Visa), and no transaction fees are charged. OpenAI developed it in collaboration with Mastercard, Visa, and American Express, and plans to gradually expand the range of supported retailers and user control options. The Verge reports that in the future, users can set parameters to fully authorize agents to make purchases without needing to confirm each transaction.

5 outlets · 5 voices

Keep reading No.04–10

Voices · last 24 hours · heat index

04
Anthropic's progress in retraining the Mythos modelNew
Anthropic can retrain its Mythos model with 'a small part of the original computing resources,' indicating an accelerated improvement in algorithmic efficiency. This is further progress for the model after achieving a major leap in capabilities in the spring.
2 outlets17New
17New
05
NVIDIA employees petition for Huang Renxun's return
3 outlets17↓ 16%
17↓ 16%
06
Meta Muse AI launches a feature to make calls on behalf of US businessesRising
3 outlets17↑ 85%
17↑ 85%
07
OpenAI has reorganized its embodied intelligence team.Rising
OpenAI has reorganized its embodied intelligence team, integrating researchers into the core research team, while the application of robots and hardware team has been transferred to a larger application department. The hope is to directly use general AI like GPT-6 Astra for some robot tasks, making them the 'brain' of the robots. Astra achieved a 74.8% success rate in simulated operation tests, higher than the 51.0% of dedicated models. However, using general models to fully support robots still faces practical challenges, such as limitations in operating speed and cloud connectivity requirements, as well as the value of on-site experience and data accumulated by robot companies.
2 outlets15↑ 106%
15↑ 106%
08
Microsoft 365 service global outage
Major outage affecting core cloud services of Microsoft 365, impacting Outlook, Exchange Online, Microsoft Teams, OneDrive, SharePoint, Microsoft 365 Copilot, and more, with users in multiple global regions unable to access the services normally. Microsoft initially stated they were investigating issues with users in the Americas accessing Exchange Online email via various connections. Subsequently, the company expanded the scope of the impact, confirming that users in any region could be affected. Charles Ramanna, executive vice president of Microsoft, publicly apologized for the outage of services like Outlook and Exchange Online, stating the team is working hard to resolve the issue and will continue to update the latest status. The outage initially affected Teams, then spread to Exchange Online and Outlook, as well as some features in SharePoint Online, Microsoft 365 Copilot, and other services dependent on Exchange. Microsoft is working hard to restore access to some office software, stating that users may still be unable to connect to email and chat services. Microsoft said its flagship software services have fully recovered, and the outage previously caused governments in many parts of the world to...
2 outlets12↓ 16%
12↓ 16%
09
DeepSeek has opened the full suite of application APIs.
2 outlets11↓ 16%
11↓ 16%
10
Amazon plans to build a product graph
Amazon plans to integrate the data of all products currently on sale globally, regardless of inventory, to enhance the depth of product data on the platform and resist competitors. Amazon claims that its internal file data is outdated, and its AI search agent has an accuracy of about 97%. However, the plan aims to keep consumer demand within the Amazon platform. Amazon's AI search agent currently often provides inaccurate information. Once the 'World Product Graph' is fully implemented, with its vast amount of data, Amazon will attract consumers from competitor e-commerce platforms with data scarcity. The target release date for this plan is October 1, 2027. Amazon hopes to reduce consumers' tendency to turn to Reddit or Google for product research through the 'World Product Graph', while also potentially attracting consumers who search for products on other platforms to eventually visit Amazon, even relying on its data-driven AI agents. The project will also aim to fix AI issues by providing more data.
2 outlets11↓ 16%
11↓ 16%