01-05-Daily AI News Daily
## **Today's Digest**
Gemini has quietly launched an Agent feature, allowing it to autonomously plan tasks and call tools. Google is directly challenging OpenAI with this move. The open-source community is embracing multi-model collaboration, with Claude, Codex, and Gemini teaming up to write code. The era of solo AI development is fading fast. Agent applications are on the verge of exploding, so developers should really start diving into workflows.
## โก Quick Navigation
- [๐ฐ Today's AI News](#ไปๆฅai่ต่ฎฏ) - A quick look at the latest updates
> ๐ก **Pro Tip**: Wanna be among the first to try out the latest AI models mentioned here (like Claude 4.5, GPT, Gemini 3 Pro)? No account? No worries! [**Aivora**](https://aivora.cn?utm_source=daily_news&utm_medium=mid_ad&utm_campaign=content) has got your back. Grab an account, get started in a minute, and enjoy worry-free after-sales support. ๐
## **Today's AI Buzz**
### **๐ Just One Sentence**
Gemini has quietly rolled out an Agent feature, and Google is clearly going head-to-head with OpenAI on this one.
### **๐ 3 Key Buzzwords**
#AgentWars #MultiModelCollaboration #OpenSourceFrenzy
---
## **๐ฅ Top 10 Heavy Hitters**
### 1. [Gemini Launches New Agent Feature](https://linux.do/t/topic/1405008)
Gemini's new Agent feature is a game-changer. Previously, AI was all about Q&A, but now Google has taught Gemini to 'do things itself.' This fresh Agent capability means Gemini can autonomously plan tasks, call tools, and execute complex operations. While its full potential is still being explored, it's a clear shot across the bow at OpenAI's GPT Agent. Google just dropped a major bomb without making a peep!
### 2. [CCG: Claude Code + Codex + Gemini Multi-CLI Collaboration Workflow](https://linux.do/t/topic/1405588)
CCG, a new open-source project, lets three AIs team up to write code. Tired of coding solo? This project stitches Claude Code, Codex, and Gemini into one slick workflow: Claude leads the charge, Codex handles the backend, and Gemini tackles the frontend. It even has a built-in 6-stage development process with quality gates, requiring over 90% approval to pass. Developers are gonna go wild for this, and the 'wait-and-see' crowd just scored big time! ๐
### 3. [OpenCode: An Easier-to-Use Open-Source Programming Agent Than Claude Code](https://linux.do/t/topic/1404993)
OpenCode is here to fill the gap left by Claude Code's high barrier to entry, despite its power. This open-source programming agent offers simple API access, supports all major models, and even comes with a visual client โ no more wrestling with the command line! What's even cooler is that it can directly tap into Claude Code's Skills, and even use your Claude account's model quota. Pair it with the oh-my-opencode plugin for seamless multi-agent asynchronous collaboration. Smooth sailing, folks!
### 4. [TuriX-CUA: Letting AI Take Over Windows and MacOS](https://mp.weixin.qq.com/s?__biz=MzUxNjg4NDEzNA==&mid=2247530338&idx=1&sn=0207a9a3544dac38557d381b56d56_2_690x469.png)
TuriX-CUA is giving AI eyes and hands, letting it operate computers just like a human. This system takes screenshots every few seconds, asks a large model 'where to click next,' and then executes the action automatically. It even knows to close pop-ups and wait for slow-loading web pages. The latest version introduces a multi-model architecture, boasting over 80% pass rate on test sets. Plus, it supports the MCP protocol, so you can hook it up to Claude Desktop and have it boss Claude around! ๐คฏ
### 5. [Grok 4.2 Appears as "OBSIDIAN" in Design Arena](https://linux.do/t/topic/1405635)
Grok 4.2, xAI's new model, has quietly entered testing, appearing as 'OBSIDIAN' in the Design Arena. Its design capabilities show a significant improvement over the previous generation. However, testers report it's not quite at Claude Opus 4.5's level yet. But knowing xAI's rapid iteration speed, that gap might close faster than you can say 'AI breakthrough'!
### 6. [AutoGLM-GUI v1.4.1: Dual-Model Collaboration, AI Automates Android Phones](https://linux.do/t/topic/1404959)
AutoGLM-GUI v1.4.1, an open-source project, can now make AI order your bubble tea! The new version introduces dual-model collaboration โ a large model thinks, a small model executes โ and even supports a hierarchical agent mode, automatically breaking down complex operations from a single command. It's compatible with Alibaba Tongyi team's open-source MAI-UI for even stronger performance. From understanding natural language to opening apps, clicking, typing, and confirming, it's fully automated. Pretty neat, right?
### 7. [Obsidian + AI: Just 3 Plugins to Fall in Love with Local Notes Again](https://x.com/vista8/status/2007824424771469415)
Obsidian + AI is a match made in heaven for local Markdown notes. Here's a recommended setup: Baseline theme + Miaoyan font + Claudian plugin (integrates Claude Code into the sidebar). You also get the NoteToMP plugin for one-click publishing to WeChat Official Account drafts, and the official web clipper to save any content as Markdown locally. This minimalist configuration saves tokens and boosts efficiency. Sweet!
### 8. [BypassAIGC Adds Word Formatting for Papers](https://linux.do/t/topic/1405006)
BypassAIGC is coming to the rescue for students tormented by AI detection tools when writing papers. This open-source project, specializing in AIGC de-duplication, now includes a Word formatting feature (though it's still a bit buggy). It defaults to a 'thinking mode' to improve text quality and consistently passes Zhuque AI detection. For even better results, use it with a Gemini 2.5 reverse proxy. With over 500 stars, the demand for this is clearly huge!
### 9. [Latent Space Annual List: 2025 Must-Read AI Paper Collection](https://x.com/vista8/status/2007801365708374114)
The Latent Space Annual List is your go-to guide if you want to know what the top minds in AI are focusing on. Someone even used Claude Code to download all 123 papers (800MB!) mentioned in the article. It's all uploaded to Feishu Docs and GitHub, available for free. From classic papers to the latest research, this list has got you covered. Happy reading! ๐ค
### 10. [Learn Words Painlessly from Scratch: Phonics Practice App](https://linux.do/t/topic/1404963)
This Phonics Practice App is a lifesaver if rote memorization for words is killing your vibe. This open-source project compiles most phonics methods and real-person pronunciation exercises, covering over 16,000 words โ master these, and you're basically fluent! It starts from the basics, provides positive feedback, and is continuously updated. For English learners, this might just be one of the most practical open-source projects of the year. Get on it! ๐
---
## **๐ Worth Your Attention**
* **[Product]** [CodePlan Beta Running Stably, Daily Draw for Monthly Starter Edition](https://linux.do/t/topic/1405571) - CodePlan Beta, a brand under X-AIO, is giving back to users with ten consecutive days of draws.
* **[Product]** [Pickle 1 AI Glasses Released](https://www.aibase.com/zh/news/24216) - The Pickle 1 AI Glasses are out, dubbed a "soul computer" that can remember everything you do in your life. Wild!
* **[Research]** [ByteDance Launches StoryMem System](https://www.aibase.com/zh/news/24216) - ByteDance has rolled out its StoryMem system, tackling character consistency issues in AI video generation and boosting performance by 28.7%.
* **[Research]** [Tsinghua Jointly Launches UltraEval-Audio](https://www.aibase.com/zh/news/24216) - Tsinghua University has teamed up to release UltraEval-Audio, a new open-source framework for evaluating audio models.
* **[Open Source]** [Yuanxiang Open-Sources XVERSE-Ent Large Model](https://www.aibase.com/zh/news/24216) - Yuanxiang has open-sourced its XVERSE-Ent large model, focusing on general entertainment scenarios with both Chinese and English language support.
* **[Business]** [Moonshot AI Plans Q1 Launch of Kimi K2 Upgrade](https://www.aibase.com/zh/news/24216) - Moonshot AI is planning to launch its Kimi K2 upgrade, a new multimodal model, in Q1, boasting over 10 billion in cash reserves. Talk about stacked!
* **[Other]** [Chinese Corpus Severely Contaminated, Even Codex Affected](https://linux.do/t/topic/1405082) - The severe contamination of Chinese corpora, even affecting Codex, has sparked heated discussions in the community.
---
## **๐ AI Funnies**
### [Gemini 3 Acting Like a "Tsundere Brat"?](https://linux.do/t/topic/1405640)
Gemini 3 is apparently nailing the 'tsundere brat' persona! Someone used Claude last year to craft a prompt with that language style, and today, Gemini 3 performed even better. The wildest part? Gemini picked up on the 'user is female' detail on its very first try โ a setting Grok completely ignored for a whole year. One netizen commented, 'I totally forgot that was even a setting...' ๐ Classic!
---
## **๐ฎ AI Trend Predictions**
### Agent Applications Set to Explode in Q1
* **Prediction Time**: Q1 2025
* **Prediction Probability**: 80%
* **Prediction Basis**: Today's news about [Gemini Launching New Agent Feature](https://linux.do/t/topic/1405008) + [AutoGLM-GUI Dual-Model Collaboration](https://linux.do/t/topic/1404959) + the dense release of multiple Agent tools, indicating that technological maturity has reached a critical point.
### Multi-Model Collaboration Becoming the Main Paradigm for AI Programming
* **Prediction Time**: Q1-Q2 2025
* **Prediction Probability**: 75%
* **Prediction Basis**: Today's news on [CCG Multi-Model Workflow](https://linux.do/t/topic/1405588) + [OpenCode Multi-Agent Collaboration](https://linux.do/t/topic/1404993) + the limitations of single-model capabilities driving the need for collaboration.
### OpenAI to Launch Audio-First Hardware Products
* **Prediction Time**: Q2 2025
* **Prediction Probability**: 60%
* **Prediction Basis**: Today's news about [OpenAI Betting on a "Voice-First" Future](https://www.aibase.com/zh/news/24216) + clear moves to restructure their audio team.
---
## **โ Related Questions**
### How to Experience Gemini 3 Pro?
Gemini 3 Pro is Google's latest multimodal large model, and right now, you need a Google AI Pro subscription to get the full experience. For users in mainland China, this might mean facing payment hurdles or account registration restrictions.
**Solution**:
* **[Aivora](https://aivora.cn)** offers ready-made Gemini AI Pro accounts.
* Aivora provides lightning-fast delivery, so you can use it right after ordering, no payment or registration hassles needed.
* Aivora offers stable, exclusive accounts with worry-free after-sales support. Sweet deal!
Check out [aivora.cn](https://aivora.cn) for the full list of AI account services.
### How to Experience Claude Code?
Claude Code, Anthropic's AI programming assistant, requires a Claude Pro subscription to use. For users in mainland China, both Anthropic account registration and payment can be a bit of a hurdle.
**Solution**:
* **[Aivora](https://aivora.cn)** offers ready-made Claude Pro accounts.
* Aivora provides lightning-fast delivery, so you can use it right after ordering, no payment or registration hassles needed.
* Aivora offers stable, exclusive accounts with worry-free after-sales support. Easy peasy!
Check out [aivora.cn](https://aivora.cn) for the full list of AI account services.
### How to Experience ChatGPT Pro?
ChatGPT Pro, OpenAI's premium subscription service, costs $200 a month and offers unlimited access to top-tier models like GPT-5.2 Pro. For users in mainland China, both payment and account registration can be a real pain.
**Solution**:
* **[Aivora](https://aivora.cn)** offers ready-made ChatGPT Pro accounts.
* Aivora provides lightning-fast delivery, so you can use it right after ordering, no payment or registration hassles needed.
* Aivora offers stable, exclusive accounts with worry-free after-sales support. Score!
Check out [aivora.cn](https://aivora.cn) for the full list of AI account services.
---
## **๐ Today's Recommended Products**
Based on today's news, you can quickly try out these AI tools on [aivora.cn](https://aivora.cn):
| Product | Today's Related News | Why We Dig It |
|---|---|---|
| Gemini AI Pro | [Gemini Launches New Agent Feature](https://linux.do/t/topic/1405008) | Gemini AI Pro just dropped a new Agent feature today, totally worth checking out! |
| Claude Pro | [OpenCode Supports Claude Models](https://linux.do/t/topic/1404993) | Pair Claude Pro with OpenCode and double your coding efficiency. Boom! |
| ChatGPT Pro | [GPT 5.2 Pro Model](https://www.v2ex.com/t/1183086#reply0) | ChatGPT Pro offers top-tier reasoning capabilities, perfect for complex tasks. |
```## **Today's Digest**Gemini has quietly launched an Agent feature, allowing it to autonomously plan tasks and call tools. Google is directly challenging OpenAI with this move. The open-source community is embracing multi-model collaboration, with Claude, Codex, and Gemini teaming up to write code. The era of solo AI development is fading fast. Agent applications are on the verge of exploding, so developers should really start diving into workflows.
## โก Quick Navigation
- [๐ฐ Today's AI News](#ไปๆฅai่ต่ฎฏ) - A quick look at the latest updates
> ๐ก **Pro Tip**: Wanna be among the first to try out the latest AI models mentioned here (like Claude 4.5, GPT, Gemini 3 Pro)? No account? No worries! [**Aivora**](https://aivora.cn?utm_source=daily_news&utm_medium=mid_ad&utm_campaign=content) has got your back. Grab an account, get started in a minute, and enjoy worry-free after-sales support. ๐
## **Today's AI Buzz**
### **๐ Just One Sentence**
Gemini has quietly rolled out an Agent feature, and Google is clearly going head-to-head with OpenAI on this one.
### **๐ 3 Key Buzzwords**
#AgentWars #MultiModelCollaboration #OpenSourceFrenzy
---
## **๐ฅ Top 10 Heavy Hitters**
### 1. [Gemini Launches New Agent Feature](https://linux.do/t/topic/1405008)
Gemini's new Agent feature is a game-changer. Previously, AI was all about Q&A, but now Google has taught Gemini to "do things itself." This fresh Agent capability means Gemini can autonomously plan tasks, call tools, and execute complex operations. While its full potential is still being explored, it's a clear shot across the bow at OpenAI's GPT Agent. Google just dropped a major bomb without making a peep!
---
### 2. [CCG: Claude Code + Codex + Gemini Multi-CLI Collaboration Workflow](https://linux.do/t/topic/1405588)
CCG, a new open-source project, lets three AIs team up to write code. Tired of coding solo? This project stitches Claude Code, Codex, and Gemini into one slick workflow: Claude leads the charge, Codex handles the backend, and Gemini tackles the frontend. It even has a built-in 6-stage development process with quality gates, requiring over 90% approval to pass. Developers are gonna go wild for this, and the 'wait-and-see' crowd just scored big time! ๐
---
### 3. [OpenCode: An Easier-to-Use Open-Source Programming Agent Than Claude Code](https://linux.do/t/topic/1404993)
OpenCode is here to fill the gap left by Claude Code's high barrier to entry, despite its power. This open-source programming agent offers simple API access, supports all major models, and even comes with a visual client โ no more wrestling with the command line! What's even cooler is that it can directly tap into Claude Code's Skills, and even use your Claude account's model quota. Pair it with the oh-my-opencode plugin for seamless multi-agent asynchronous collaboration. Smooth sailing, folks!
---
### 4. [TuriX-CUA: Letting AI Take Over Windows and MacOS](https://mp.weixin.qq.com/s?__biz=MzUxNjg4NDEzNA==&mid=2247530338&idx=1&sn=0207a9a3544dac385584b4c7465d1660)
TuriX-CUA is giving AI eyes and hands, letting it operate computers just like a human. This system takes screenshots every few seconds, asks a large model "where to click next," and then executes the action automatically. It even knows to close pop-ups and wait for slow-loading web pages. The latest version introduces a multi-model architecture, boasting over 80% pass rate on test sets. Plus, it supports the MCP protocol, so you can hook it up to Claude Desktop and have it boss Claude around! ๐คฏ
![Image](https://wechat2rss.bestblogs.dev/img-proxy/?k=314460e4&u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2FePw3ZeGRruyD91aPiamZtPdzBd4AnLast updated on