Shut Down for Five Years, Xiami Resurrects via AI: Alibaba Launches HappyShrimp for 30 Yuan

HappyShrimp, developed by Alibaba's ATH (TokenHub) business group, has officially launched on the iOS and Android app stores, and has been updated to version 0.2.14 on the App Store.

It is not simply a re-release of an old product; it is the third model product in Alibaba's "Happy Series"—the first two are the video generation model HappyHorse and the 3D interactive video model HappyOyster. Users can describe emotions, stories, or musical styles in plain language, and the AI completes lyric writing, composition, arrangement, and vocal performance in one go. It focuses on end-to-end full-song generation rather than piecing steps together, and offers two output modes, "Smart Lyrics" and "Instrumental," plus reuse features such as "Music Recipe" and "Make a Similar Track."

Pricing for ordinary users is quite modest: the standard membership in China costs about 30 yuan per month and can generate 150 songs, which works out to about 0.2 yuan per song; overseas plans start at $14.90 per month, with credit packs starting at $9.90. The store also offers credit and membership tiers ranging from 89 yuan to 1,149 yuan, giving heavy users room to upgrade.

Yufei International Chinese Language Education Vertical Large Model Released: Deeply Integrating Professional Teaching Scenarios to Provide Intelligent Services for Global Teachers, Students, and Institutions

The integration of artificial intelligence and vertical fields is advancing in depth. According to CCTV News, the 2026 World Chinese Language Conference opened in Beijing on November 13, where the Yufei International Chinese Language Education Vertical Large Model was released. This model serves Chinese language learners, frontline teachers, and educational institutions worldwide, providing comprehensive intelligent services, deeply adapting to the entire teaching and learning chain, covering three core groups, and promoting the deep integration of AI with the actual needs of international Chinese language education.

Nvidia Employees Launch Campaign to Retain Jensen Huang; Internal Poll Shows Over 97% Support His Staying On

WIRED reports that Nvidia employees spontaneously launched a campaign to retain Jensen Huang. An internal poll shows that more than 97% of participants want Jensen Huang to continue serving as CEO, and several employees have asked the board to reconsider related personnel decisions. A senior engineer posted through the internal suggestion box, calling on the board to fully consider employee opinions before taking action. The post received more than 1,600 comments, and many employees support Jensen Huang and hope he returns.

Tencent Hunyuan unveils real-time voice model Hy Realtime: it can chat, sing, and search the internet on its own

Tencent Hunyuan has released a brand-new real-time voice model, Hy Realtime. Focusing on real-time performance, it excels in basic conversational capabilities such as speech understanding, instruction following, and human-like expression, and newly adds Agentic tool calling and singing capabilities. It upgrades from simple responses to executable tasks and is closer to real human conversation. The model has been applied to products such as Yuanbao voice calls and Tencent Customer Service.

Even More Powerful Than High-End Models Yet Cheaper: Google Bucks the Trend with Gemini 4 Hydrogen

Google launched Gemini4Hydrogen (Hydrogen), focusing on speed and execution efficiency, but unexpectedly beat the higher-positioned Argon: on a third-party intelligence index, Hydrogen scored 51, with its Xhigh tier scoring 54, higher than Argon's 50. Its default tier is 2.5 times faster than the previous generation Flash, reduces hallucinations by over 80%, and its API price is also lower than Argon's. Google also launched Antigravity Orchestrator, which can, within the same

Claude Assists in Completing 190,000 Lines of Code, Successfully Formalizing the Ternary Goldbach Conjecture

AI is rewriting basic scientific research through formal proofs. Independent developer Charlie Reams, with the help of Claude Code, completed a formal proof of the ternary Goldbach conjecture, producing a total of 194,000 lines of Lean4 code with zero `sorry`, and all mathematical reasoning was rigorously verified by machine. During the process, the AI demonstrated rigor beyond traditional human peer review, autonomously identifying and patching 11 flaws from 13 years ago, highlighting its powerful capability in mathematical verification.