01-01-Daily AI News Daily

Today’s Digest

Zhihui Jun just dropped his small-scale humanoid robot, Q1, featuring full-body force control. This is a big deal, marking the official entry of a domestic player into the personal robot race! Meanwhile, South Korea’s government went all-in, open-sourcing three massive AI models in just one week. They’re treating open-source AI as a national strategy, no joke. Even on the very last day of 2025, the AI scene was still buzzing. Andrew Ng’s annual summary is a goldmine – definitely save it for your learning roadmap next year!

⚡ Quick Navigation

💡 Pro Tip: Dying to try out the latest AI models like Claude 4.5, GPT, or Gemini 3 Pro right away? No account? No sweat! Head over to Aivora to grab an account. You’ll be up and running in a minute, with zero fuss and awesome support. 🚀

Today’s AI News

👀 Just One Sentence

Zhihui Jun’s Q1 robot just dropped, South Korea open-sourced three big models in one go, and the AI scene was still grinding hard on the very last day of 2025.

🔑 3 Key Takeaways

#RobotsAreHere #SouthKoreaOpenSourceSpree #YearEndReview

🔥 Top 10 Heavy Hitters

1. Zhihui Jun Releases Small-Scale Full-Body Force-Controlled Humanoid Robot Q1

Zhihui Jun, remember that genius who built the “electronic spine”? Well, he’s at it again! This time, he’s unveiled the Prime Quester1 (Q1), a small-scale, full-body force-controlled humanoid robot. Don’t let its size fool you; this little guy isn’t a stiff mechanical arm. It can interact with its environment as smoothly and flexibly as a human. The official slogan, “Your first personal robot is HERE,” clearly shows their big ambitions. This marks the official entry of a domestic player into the personal robot arena.

2. South Korea Open-Sources Three Large Models at Once, Government Backs “Sovereign AI”

South Korea’s open-source AI ecosystem is exploding, as the Hugging Face CEO tweeted. This week, South Korea dropped three major open-source models: LG’s K-Exagone (236B MoE), Upstage’s Solar Open (102B), and SK Telecom’s A.X K1 (519B, released Jan 4). This surge is backed by the government’s “Sovereign AI Foundation Model” project. While others are still debating whether to open-source, South Korea has already made it a national strategy. Expect the open-source AI competition to heat up even more in 2026.

3. Alibaba Open-Sources Qwen-Image-2512: AI-Generated Images Finally Look Less “AI-ish”

Alibaba’s Qwen team just released their next-gen image generation model, Qwen-Image-2512. The big news? Human faces finally look real! Remember that uncanny “plastic” vibe from old AI-generated portraits? The official word is they’ve done a ton of optimization, leading to richer facial details, finer natural textures (think landscapes and hair), and improved text rendering. They even ran over 10,000 rounds of blind tests on AI Arena, claiming it’s the strongest open-source image model out there, capable of going head-to-head with closed-source ones. Wanna give it a spin? You can try it directly on Qwen Chat or Hugging Face.

4. Andrew Ng’s Annual AI Summary: 2025 is the First Year of a New Industrial Era

Andrew Ng just dropped his annual AI summary, highlighting four key points: 1) Model inference capabilities are now standard, no longer a privilege for a select few. 2) AI talent salaries are skyrocketing, with Meta leading the charge, offering “superstar contracts.” 3) Data centers are becoming the “steel mills” of a new era, signaling the arrival of a capital-intensive age. 4) AI Agents are reshaping programming, with code completion rates commonly exceeding 80%. His advice to developers? Don’t just watch from the sidelines; dive in and learn! Courses, hands-on practice, and papers—don’t skip any of them.

5. Google Gemini Open-Sources “Computer Use” Preview

Google quietly dropped a project called computer-use-preview on GitHub, already racking up over 2,300 stars. While there’s no official description, the name suggests it’s a Gemini-powered “computer control” capability preview—similar to Anthropic’s earlier Claude Computer Use. AI isn’t just for chatting anymore; it’s about to start helping you operate your computer. This space is heating up fast!

6. TrendRadar: AI Helps You Understand Hot Topics Across 35 Platforms

TrendRadar is an open-source project aiming to solve information overload, a common modern ailment. It monitors hot topics across 35 platforms like Douyin, Zhihu, Bilibili, and Wall Street News, using AI to filter, analyze, and push updates to you. It supports various notification channels, including WeChat Work, Feishu, DingTalk, and Telegram, and can even perform in-depth natural language analysis (trend tracking, sentiment analysis, etc.). Deploy it in 30 seconds, get notified in 1 minute—perfect for media professionals and investors who need to stay on top of trends. With over 40,000 stars, its popularity is well-deserved.

7. Chatterbox: Open-Source Real-time Voice Conversation Framework

Resemble AI has open-sourced Chatterbox, a real-time voice framework designed to eliminate the “laggy” feeling in AI voice interactions. It optimizes the pipeline for Automatic Speech Recognition (ASR), Large Language Model (LLM) processing, and Text-to-Speech (TTS). The magic? TTS starts generating the beginning of the audio even before the LLM has finished its full response. This streaming approach makes conversations sound much more natural. Developers looking to build virtual humans or smart customer service agents should definitely check it out. It’s already nearing 20,000 stars!

8. Vibe-Kanban: A Kanban Tool Ready for the AI Era

“Vibe Coding” is a hot term right now, and even Kanban tools are getting in on the “vibe.” Vibe-Kanban claims it can boost the efficiency of coding agents like Claude Code and Codex by tenfold. Instead of manually typing out task cards, it deeply integrates AI to help you break down vague ideas into actionable tasks. It’s lighter than Jira and understands developers better than Notion. With 10,000 stars, it’s perfect for fast-paced small teams and indie developers.

9. Nvidia Acquires Israeli AI Startup for $2 Billion

Nvidia’s Jensen Huang (老黄) is at it again! Nvidia is reportedly acquiring an Israeli AI startup for over $2 billion. The specific company hasn’t been officially announced, but this “acquire-to-hire” strategy is something Nvidia has mastered. Why wait for talent to slowly grow when you can just buy the whole team? In today’s red-hot AI talent war, having deep pockets means you can do whatever you want.

10. 2026 AI Predictions: Edge AI, Agent Harness, Human Authentication

Philipp Schmid of Google DeepMind just dropped his 8 big predictions for 2026: 1) Generative UIs will rise. 2) Personal agents will run on edge devices. 3) Smart homes will finally deliver on their promises. 4) “Agent Harness” will become the new benchmark for AI labs. 5) Vibe Coding will mature. 6) Social media will get “human-created” labels to fight AI content. 7) Biometrics will prove “you are human.” These predictions are pretty interesting and worth saving for a year-end check to see how they pan out.

📌 Worth Keeping an Eye On

  • [Product] NoCoDB - An open-source Airtable alternative, boasting 60K stars, it’s a godsend for building your own databases.
  • [Product] Organic Maps - A privacy-first offline map app with no ads or tracking, an absolute must-have for hiking and cycling.
  • [Open Source] OpenAI Cookbook Update - The official API usage guide with 70K stars, always fresh and worth revisiting.
  • [Open Source] Harvard Machine Learning Systems Textbook - With 14K stars, this is a hefty read for anyone serious about learning ML systematically.
  • [Tool] SpotiFLAC - Grab FLAC versions of Spotify tracks from Tidal/Qobuz, no account needed.
  • [Tool] pg-aiguide - An MCP server that helps AI write better PostgreSQL code.

❓ Related Questions

How to Try Alibaba’s Qwen-Image Image Generation Model?

Alibaba’s Qwen-Image-2512 is their latest open-source image generation model, renowned for its more realistic human portrait rendering and delicate natural textures. You can currently try it out for free on platforms like Qwen Chat, Hugging Face, and ModelScope.

If you’re looking for a more stable AI image generation service: * Aivora ( aivora.cn ) offers ready-to-use accounts for popular AI art tools like Midjourney and DALL-E. * Get instant delivery, use it right after ordering, and skip the hassle of overseas payments. * Enjoy stable, dedicated accounts with worry-free after-sales support.

Visit aivora.cn to check out their full list of AI account services.

Today’s Digest 

Zhihui Jun just dropped his small-scale humanoid robot, Q1, featuring full-body force control, marking the official entry of a domestic player into the personal robot race!
Meanwhile, South Korea's government went all-in, open-sourcing three massive AI models in just one week. They're treating open-source AI as a national strategy, no joke.
Even on the very last day of 2025, the AI scene was still buzzing. Andrew Ng's annual summary is a goldmine – definitely save it for your learning roadmap next year!

⚡ Quick Navigation

💡 Pro Tip: Dying to try out the latest AI models like Claude 4.5, GPT, or Gemini 3 Pro right away? No account? No sweat! Head over to Aivora to grab an account. You’ll be up and running in a minute, with zero fuss and awesome support. 🚀

Today’s AI News

👀 Just One Sentence

Zhihui Jun’s Q1 robot just dropped, South Korea open-sourced three big models in one go, and the AI scene was still grinding hard on the very last day of 2025.

🔑 3 Key Takeaways

#RobotsAreHere #SouthKoreaOpenSourceSpree #YearEndReview

🔥 Top 10 Heavy Hitters

1. Zhihui Jun Releases Small-Scale Full-Body Force-Controlled Humanoid Robot Q1

Zhihui Jun, remember that genius who built the “electronic spine”? Well, he’s at it again! This time, he’s unveiled the Prime Quester1 (Q1), a small-scale, full-body force-controlled humanoid robot. Don’t let its size fool you; this little guy isn’t a stiff mechanical arm. It can interact with its environment as smoothly and flexibly as a human. The official slogan, “Your first personal robot is HERE,” clearly shows their big ambitions. This marks the official entry of a domestic player into the personal robot arena.

2. South Korea Open-Sources Three Large Models at Once, Government Backs “Sovereign AI”

South Korea’s open-source AI ecosystem is exploding, as the Hugging Face CEO tweeted. This week, South Korea dropped three major open-source models: LG’s K-Exagone (236B MoE), Upstage’s Solar Open (102B), and SK Telecom’s A.X K1 (519B, released Jan 4). This surge is backed by the government’s “Sovereign AI Foundation Model” project. While others are still debating whether to open-source, South Korea has already made it a national strategy. Expect the open-source AI competition to heat up even more in 2026.

3. Alibaba Open-Sources Qwen-Image-2512: AI-Generated Images Finally Look Less “AI-ish”

Alibaba’s Qwen team just released their next-gen image generation model, Qwen-Image-2512. The big news? Human faces finally look real! Remember that uncanny “plastic” vibe from old AI-generated portraits? The official word is they’ve done a ton of optimization, leading to richer facial details, finer natural textures (think landscapes and hair), and improved text rendering. They even ran over 10,000 rounds of blind tests on AI Arena, claiming it’s the strongest open-source image model out there, capable of going head-to-head with closed-source ones. Wanna give it a spin? You can try it directly on Qwen Chat or Hugging Face.

4. Andrew Ng’s Annual AI Summary: 2025 is the First Year of a New Industrial Era

Andrew Ng just dropped his annual AI summary, highlighting four key points: 1) Model inference capabilities are now standard, no longer a privilege for a select few. 2) AI talent salaries are skyrocketing, with Meta leading the charge, offering “superstar contracts.” 3) Data centers are becoming the “steel mills” of a new era, signaling the arrival of a capital-intensive age. 4) AI Agents are reshaping programming, with code completion rates commonly exceeding 80%. His advice to developers? Don’t just watch from the sidelines; dive in and learn! Courses, hands-on practice, and papers—don’t skip any of them.

5. Google Gemini Open-Sources “Computer Use” Preview

Google quietly dropped a project called computer-use-preview on GitHub, already racking up over 2,300 stars. While there’s no official description, the name suggests it’s a Gemini-powered “computer control” capability preview—similar to Anthropic’s earlier Claude Computer Use. AI isn’t just for chatting anymore; it’s about to start helping you operate your computer. This space is heating up fast!

6. TrendRadar: AI Helps You Understand Hot Topics Across 35 Platforms

TrendRadar is an open-source project aiming to solve information overload, a common modern ailment. It monitors hot topics across 35 platforms like Douyin, Zhihu, Bilibili, and Wall Street News, using AI to filter, analyze, and push updates to you. It supports various notification channels, including WeChat Work, Feishu, DingTalk, and Telegram, and can even perform in-depth natural language analysis (trend tracking, sentiment analysis, etc.). Deploy it in 30 seconds, get notified in 1 minute—perfect for media professionals and investors who need to stay on top of trends. With over 40,000 stars, its popularity is well-deserved.

7. Chatterbox: Open-Source Real-time Voice Conversation Framework

Resemble AI has open-sourced Chatterbox, a real-time voice framework designed to eliminate the “laggy” feeling in AI voice interactions. It optimizes the pipeline for Automatic Speech Recognition (ASR), Large Language Model (LLM) processing, and Text-to-Speech (TTS). The magic? TTS starts generating the beginning of the audio even before the LLM has finished its full response. This streaming approach makes conversations sound much more natural. Developers looking to build virtual humans or smart customer service agents should definitely check it out. It’s already nearing 20,000 stars!

8. Vibe-Kanban: A Kanban Tool Ready for the AI Era

“Vibe Coding” is a hot term right now, and even Kanban tools are getting in on the “vibe.” Vibe-Kanban claims it can boost the efficiency of coding agents like Claude Code and Codex by tenfold. Instead of manually typing out task cards, it deeply integrates AI to help you break down vague ideas into actionable tasks. It’s lighter than Jira and understands developers better than Notion. With 10,000 stars, it’s perfect for fast-paced small teams and indie developers.

9. Nvidia Acquires Israeli AI Startup for $2 Billion

Nvidia’s Jensen Huang is at it again! Nvidia is reportedly acquiring an Israeli AI startup for over $2 billion. The specific company hasn’t been officially announced, but this “acquire-to-hire” strategy is something Nvidia has mastered. Why wait for talent to slowly grow when you can just buy the whole team? In today’s red-hot AI talent war, having deep pockets means you can do whatever you want.

10. 2026 AI Predictions: Edge AI, Agent Harness, Human Authentication

Philipp Schmid of Google DeepMind just dropped his 8 big predictions for 2026: 1) Generative UIs will rise. 2) Personal agents will run on edge devices. 3) Smart homes will finally deliver on their promises. 4) “Agent Harness” will become the new benchmark for AI labs. 5) Vibe Coding will mature. 6) Social media will get “human-created” labels to fight AI content. 7) Biometrics will prove “you are human.” These predictions are pretty interesting and worth saving for a year-end check to see how they pan out.

📌 Worth Keeping an Eye On

  • [Product] NoCoDB - An open-source Airtable alternative, boasting 60K stars, it’s a godsend for building your own databases.
  • [Product] Organic Maps - A privacy-first offline map app with no ads or tracking, an absolute must-have for hiking and cycling.
  • [Open Source] OpenAI Cookbook Update - The official API usage guide with 70K stars, always fresh and worth revisiting.
  • [Open Source] Harvard Machine Learning Systems Textbook - With 14K stars, this is a hefty read for anyone serious about learning ML systematically.
  • [Tool] SpotiFLAC - Grab FLAC versions of Spotify tracks from Tidal/Qobuz, no account needed.
  • [Tool] pg-aiguide - An MCP server that helps AI write better PostgreSQL code.

❓ Related Questions

How to Try Alibaba’s Qwen-Image Image Generation Model?

Alibaba’s Qwen-Image-2512 is their latest open-source image generation model, renowned for its more realistic human portrait rendering and delicate natural textures. You can currently try it out for free on platforms like Qwen Chat, Hugging Face, and ModelScope.

If you’re looking for a more stable AI image generation service:

  • Aivora offers ready-to-use accounts for popular AI art tools like Midjourney and DALL-E.
  • Get instant delivery, use it right after ordering, and skip the hassle of overseas payments.
  • Enjoy stable, dedicated accounts with worry-free after-sales support.

Visit aivora.cn to check out their full list of AI account services.

Last updated on
01-02-Daily