12-26-Daily AI News Daily
Aivora AI Daily 2025/12/26
AI Daily
Today’s Summary
Nvidia pulled a Google, poaching Groq’s founding team – they want the talent, not the whole company. Talk about a slick move!
The AI toolchain is getting a serious speed boost: ByteDance’s TRAE just hit over 1.6 million monthly active users, and Tsinghua open-sourced a video generation solution that’s 200 times faster.
Today’s highlights are all about talent mobility and the efficiency revolution. Developers, you might want to check out the newly open-sourced acceleration framework.
⚡ Quick Navigation
Quick glance at the latest developments
💡 Tip
Want to be among the first to try out the latest AI models mentioned in this article (like Claude 4.5, GPT, Gemini 3 Pro)? Aivora has got you covered! Grab an account, get started in a minute, and enjoy worry-free after-sales support.
Today’s AI News
👀 One-Liner
Nvidia just poached Groq’s founding team – they want the talent, not the company. That’s a quick lesson learned!
🔑 3 Keywords
#TalentHarvest #AIProgrammingTakesOff #VideoGenerationAccelerated
🔥 Top 10 Heavy Hitters
1. Nvidia “Acquires” Groq Founding Team, Wants People Not Company
Remember Groq, the company that boasted the ‘world’s fastest inference chip’? Well, Groq’s founder Jonathan Ross and his core team are joining Nvidia, with Groq’s inference tech non-exclusively licensed to Jensen Huang. This move feels super familiar – Google did the same thing with Character.AI. GroqCloud will keep running, but its soul has a new owner. Jensen’s play here is genius: he snagged the tech, harvested the talent, and didn’t have to carry the whole company’s baggage. Talk about a power move! 💪
ByteDance’s AI programming tool, TRAE, just dropped its annual report, and the numbers are seriously impressive: 6 million total users and 1.6 million monthly active users, making it potentially the top AI programming tool in China. What’s super interesting is the user behavior data: 40% use it for bug fixing, 30% for code generation, while features like repository management and code optimization are still viewed with a bit of skepticism. As for programming languages, Vue takes the lead, with Python and JS hot on its heels. Looks like frontend developers are the main force driving AI programming!
TurboDiffusion, open-sourced by Tsinghua TSAIL team and Shengshu Technology, is a game-changer for video generation. Previously, generating a 5-second 720p video with a Diffusion model took a whopping 4767 seconds (over an hour). Now? It’s done in 24 seconds, and even under 2 seconds at its fastest! The magic behind it? A triple threat of hybrid attention acceleration, efficient step distillation, and W8A8 quantization. Even better – the image quality is barely compromised. This means real-time video generation is now possible on consumer-grade GPUs, officially ushering AI video from ‘offline waiting’ to the ‘real-time preview’ era. 🚀
4. OpenAI Plans to Raise $100 Billion, Valuation to Hit $830 Billion
OpenAI is gearing up for another massive move. Reports suggest they’re looking to raise a staggering $100 billion, which, if successful, would push their valuation to $830 billion. What does that even mean? It’s more than a third of Nvidia’s market cap! While it’s still early days and investor demand isn’t set in stone, one thing’s crystal clear: the AI arms race is burning cash faster than most folks can imagine. 🔥
5. ChatGPT to Launch “Skills” Feature, Coming Early 2026
ChatGPT’s new ‘Skills’ feature, internally codenamed ‘Hazel,’ has been unveiled. Skills are modular capability units, organized like folders, containing instructions, context, examples, and even executable code. The coolest part? Skills can be automatically recognized, dynamically loaded, and combined by AI. Plus, you can convert existing GPTs into Skills with a single click! This means you’ll soon be able to assemble AI capabilities like LEGO bricks, instead of fine-tuning from scratch every time. Expect it to drop around January 2026. 🤯
6. Liquid AI Releases 2.6B Parameter Model, Outperforming Billion-Parameter Competitors
Liquid AI’s LFM2-2.6B-Exp model proves that small models can still win big! With only 2.6 billion parameters, it outperformed billion-parameter models in multiple benchmark tests, especially in instruction following. What’s even wilder is that it was trained purely with reinforcement learning, no supervised fine-tuning or large model distillation involved. It’s fully open-source and supports edge device deployment. So, running a small model on your phone might just become more effective than a massive cloud-based one. How cool is that? 😎
7. Kuaikan Manhua Teams Up with MiniMax, AI Interactive Comics Are Here
Kuaikan Manhua and MiniMax have teamed up to bring AI interactive comics to life, finally letting manga characters ’live’! Through AI tech, comic characters can now engage in multimodal interactions and emotional connections. You can chat and interact with your favorite characters, not just read their stories. This is being dubbed the ‘fourth track of AI experience’ in the comic industry. For anime/manga fans, this might just be one of the most exciting product forms to drop this year! ✨
8. Huawei Noah’s Ark Lab Releases SCOPE, Enabling Agents to Evolve from Errors
Huawei Noah’s Ark Lab, in collaboration with CUHK, has released the SCOPE framework, tackling a major AI Agent problem: seeing an error message but making the same mistake next time. SCOPE solves this by enabling an Agent’s Prompt to self-evolve, learning from its execution trajectory. In real-world tests on the HLE benchmark, task success rates more than doubled, jumping from 14.23% to 38.64%. This is a huge breakthrough for building truly reliable AI Agents! 🧠
Xiaohongshu’s open-sourced InstanceAssemble framework is tackling one of e-commerce’s biggest headaches: getting AI to precisely place multiple items in a single image. By using cascaded modeling and an Assemble-Attention mechanism, it significantly boosts spatial alignment accuracy in complex scenarios. With only a 0.84% parameter increase and compatibility with mainstream base models, this is a tool worth watching for anyone in e-commerce design or poster generation! 🎨
10. OpenAI Releases Holiday Custom Edition Codex Model: GPT-5.2-Codex-XMas
The OpenAI Codex team just dropped a Christmas special edition model, GPT-5.2-Codex-XMas! This is purely an Easter egg. Its performance is identical to the regular version, but it comes with a ‘Santa Claus persona.’ When you call it via the command line, it’ll chat with you about code in Santa’s voice. While it’s not super practical, this little bit of humor is pretty adorable! 🎅
📌 Worth Noting
[Products]
- Lingguang AI Launched for One Month, Users Created 12 Million Flash Apps
- Create mini-apps quickly without coding, a new way to play with AI-native apps
- Lenovo to Release “Super Intelligent Agent” at CES, Connecting PC and Mobile Ecosystems
- Going head-to-head with Doubao, Lenovo aims to be a cross-device AI assistant
[Research]
- AI-Generated Faces Can Be Detected with 5 Minutes of Training
- Teeth, hairline, and ear symmetry are the giveaways
- MeViSv2 Multimodal Video Understanding Dataset Released
- Adding 150,000 seconds of audio data, challenging the limits of AI logical reasoning
[Open Source]
- LEANN: RAG for Everything, Saving 97% Storage Space
- Run fast, accurate, and private RAG on personal devices
- MediaCrawler: Xiaohongshu/Douyin/Bilibili Scraper Tool
- 40k+ Stars, a must-have for content analysis
[Other]
- Epic is Giving Away “Disco Elysium - The Final Cut” for Free Today
- A masterpiece, grab it!
❓ Related Questions
How to experience ChatGPT’s latest features (e.g., upcoming Skills)?
ChatGPT’s advanced features (like GPT-4o and the upcoming Skills) usually require a ChatGPT Plus subscription. For users in mainland China, this can often mean payment difficulties or account registration restrictions.
Solution:
- Aivora offers ready-to-use ChatGPT Plus accounts
- Super fast delivery, use immediately after ordering, no payment or registration hassles
- Provides stable, exclusive accounts with worry-free after-sales support
Visit aivora.cn to see the full list of AI account services.
Aivora AI Daily 2025/12/26
AI Daily
Today’s Summary
Nvidia pulled a Google, poaching Groq’s founding team – they want the talent, not the whole company. Talk about a slick move! The AI toolchain is getting a serious speed boost: ByteDance’s TRAE just hit over 1.6 million monthly active users, and Tsinghua open-sourced a video generation solution that’s 200 times faster. Today’s highlights are all about talent mobility and the efficiency revolution. Developers, you might want to check out the newly open-sourced acceleration framework.
⚡ Quick Navigation
- 📰 Today’s AI News - Quick glance at the latest developments
💡 Tip: Want to be among the first to try out the latest AI models mentioned in this article (like Claude 4.5, GPT, Gemini 3 Pro)? No account? Aivora has got you covered! Grab an account, get started in a minute, and enjoy worry-free after-sales support.
Today’s AI News
👀 One-Liner
Nvidia just poached Groq’s founding team – they want the talent, not the company. That’s a quick lesson learned!
🔑 3 Keywords
#TalentHarvest #AIProgrammingTakesOff #VideoGenerationAccelerated
🔥 Top 10 Heavy Hitters
1. Nvidia “Acquires” Groq Founding Team, Wants People Not Company
Remember Groq, the company that boasted the “world’s fastest inference chip”? Well, Groq’s founder Jonathan Ross and his core team are joining Nvidia, with Groq’s inference tech non-exclusively licensed to Jensen Huang. This move feels super familiar – Google did the same thing with Character.AI. GroqCloud will keep running, but its soul has a new owner. Jensen’s play here is genius: he snagged the tech, harvested the talent, and didn’t have to carry the whole company’s baggage. Talk about a power move! 💪
2. TRAE Monthly Active Users Exceed 1.6 Million, AI Programming Ecosystem Data Revealed for the First Time
ByteDance’s AI programming tool, TRAE, just dropped its annual report, and the numbers are seriously impressive: 6 million total users and 1.6 million monthly active users, making it potentially the top AI programming tool in China. What’s super interesting is the user behavior data: 40% use it for bug fixing, 30% for code generation, while features like repository management and code optimization are still viewed with a bit of skepticism. As for programming languages, Vue takes the lead, with Python and JS hot on its heels. Looks like frontend developers are the main force driving AI programming!
3. Tsinghua University Teams Up with Shengshu Technology to Open Source TurboDiffusion, Accelerating Video Generation by 200x
TurboDiffusion, open-sourced by Tsinghua TSAIL team and Shengshu Technology, is a game-changer for video generation. Previously, generating a 5-second 720p video with a Diffusion model took a whopping 4767 seconds (over an hour). Now? It’s done in 24 seconds, and even under 2 seconds at its fastest! The magic behind it? A triple threat of hybrid attention acceleration, efficient step distillation, and W8A8 quantization. Even better – the image quality is barely compromised. This means real-time video generation is now possible on consumer-grade GPUs, officially ushering AI video from ‘offline waiting’ to the ‘real-time preview’ era. 🚀
4. OpenAI Plans to Raise $100 Billion, Valuation to Hit $830 Billion
OpenAI is gearing up for another massive move. Reports suggest they’re looking to raise a staggering $100 billion, which, if successful, would push their valuation to $830 billion. What does that even mean? It’s more than a third of Nvidia’s market cap! While it’s still early days and investor demand isn’t set in stone, one thing’s crystal clear: the AI arms race is burning cash faster than most folks can imagine. 🔥
5. ChatGPT to Launch “Skills” Feature, Coming Early 2026
ChatGPT’s new “Skills” feature, internally codenamed ‘Hazel,’ has been unveiled. Skills are modular capability units, organized like folders, containing instructions, context, examples, and even executable code. The coolest part? Skills can be automatically recognized, dynamically loaded, and combined by AI. Plus, you can convert existing GPTs into Skills with a single click! This means you’ll soon be able to assemble AI capabilities like LEGO bricks, instead of fine-tuning from scratch every time. Expect it to drop around January 2026. 🤯
6. Liquid AI Releases 2.6B Parameter Model, Outperforming Billion-Parameter Competitors
Liquid AI’s LFM2-2.6B-Exp model proves that small models can still win big! With only 2.6 billion parameters, it outperformed billion-parameter models in multiple benchmark tests, especially in instruction following. What’s even wilder is that it was trained purely with reinforcement learning, no supervised fine-tuning or large model distillation involved. It’s fully open-source and supports edge device deployment. So, running a small model on your phone might just become more effective than a massive cloud-based one. How cool is that? 😎
7. Kuaikan Manhua Teams Up with MiniMax, AI Interactive Comics Are Here
Kuaikan Manhua and MiniMax have teamed up to bring AI interactive comics to life, finally letting manga characters ’live’! Through AI tech, comic characters can now engage in multimodal interactions and emotional connections. You can chat and interact with your favorite characters, not just read their stories. This is being dubbed the ‘fourth track of AI experience’ in the comic industry. For anime/manga fans, this might just be one of the most exciting product forms to drop this year! ✨
8. Huawei Noah’s Ark Lab Releases SCOPE, Enabling Agents to Evolve from Errors
Huawei Noah’s Ark Lab, in collaboration with CUHK, has released the SCOPE framework, tackling a major AI Agent problem: seeing an error message but making the same mistake next time. SCOPE solves this by enabling an Agent’s Prompt to self-evolve, learning from its execution trajectory. In real-world tests on the HLE benchmark, task success rates more than doubled, jumping from 14.23% to 38.64%. This is a huge breakthrough for building truly reliable AI Agents! 🧠
9. Xiaohongshu Open Sources InstanceAssemble, Breaking Through Precision in Complex Multi-Instance Image Generation
Xiaohongshu’s open-sourced InstanceAssemble framework is tackling one of e-commerce’s biggest headaches: getting AI to precisely place multiple items in a single image. By using cascaded modeling and an Assemble-Attention mechanism, it significantly boosts spatial alignment accuracy in complex scenarios. With only a 0.84% parameter increase and compatibility with mainstream base models, this is a tool worth watching for anyone in e-commerce design or poster generation! 🎨
10. OpenAI Releases Holiday Custom Edition Codex Model: GPT-5.2-Codex-XMas
The OpenAI Codex team just dropped a Christmas special edition model, GPT-5.2-Codex-XMas! This is purely an Easter egg. Its performance is identical to the regular version, but it comes with a ‘Santa Claus persona.’ When you call it via the command line, it’ll chat with you about code in Santa’s voice. While it’s not super practical, this little bit of humor is pretty adorable! 🎅
📌 Worth Noting
[Products]
- Lingguang AI Launched for One Month, Users Created 12 Million Flash Apps - Create mini-apps quickly without coding, a new way to play with AI-native apps
- Lenovo to Release “Super Intelligent Agent” at CES, Connecting PC and Mobile Ecosystems - Going head-to-head with Doubao, Lenovo aims to be a cross-device AI assistant
[Research]
- AI-Generated Faces Can Be Detected with 5 Minutes of Training - Teeth, hairline, and ear symmetry are the giveaways
- MeViSv2 Multimodal Video Understanding Dataset Released - Adding 150,000 seconds of audio data, challenging the limits of AI logical reasoning
[Open Source]
- LEANN: RAG for Everything, Saving 97% Storage Space - Run fast, accurate, and private RAG on personal devices
- MediaCrawler: Xiaohongshu/Douyin/Bilibili Scraper Tool - 40k+ Stars, a must-have for content analysis
[Other]
- Epic is Giving Away “Disco Elysium - The Final Cut” for Free Today - A masterpiece, grab it!
❓ Related Questions
How to experience ChatGPT’s latest features (e.g., upcoming Skills)?
ChatGPT’s advanced features (like GPT-4o and the upcoming Skills) usually require a ChatGPT Plus subscription. For users in mainland China, this can often mean payment difficulties or account registration restrictions.
Solution:
- Aivora offers ready-to-use ChatGPT Plus accounts
- Super fast delivery, use immediately after ordering, no payment or registration hassles
- Provides stable, exclusive accounts with worry-free after-sales support
Visit aivora.cn to see the full list of AI account services.