12-31-Daily AI News Daily

今日摘要

Meta just dropped billions to acquire Chinese AI Agent company Manus. Its founder is now a VP, and they hit $125M ARR in just three months. Tencent has open-sourced a billion-parameter text-to-3D motion model, and the open-source coding tool OpenCode has racked up a whopping 44K Stars. Chinese AI is making waves globally, setting a benchmark for going overseas, and the open-source ecosystem is absolutely exploding. Today’s news is definitely worth a deep dive.

⚡ Quick Nav

Want to be among the first to try out the latest AI models mentioned here (Claude 4.5, GPT, Gemini 3 Pro)? No account? No worries! Aivora has got your back. Grab an account, get started in a minute, and enjoy worry-free after-sales support.

今日AI资讯

🎉 Meta dropped billions on Manus, and with that, a Chinese AI Agent team has officially crashed the Silicon Valley party.

#ManusAcquired #ThousandSmallThingsPhilosophy #AIProgrammingSurge

✨ Top 10

1. Meta Acquires Manus for Billions, Founder Becomes VP]( https://x.com/dotey/status/2006074924868440120 )

The absolute mind-blower today? Meta just announced its acquisition of Chinese AI Agent company Manus for billions of dollars. This deal is Meta’s third-largest ever, right after WhatsApp and Instagram. Xiao Hong, the founder, is stepping in as a Meta VP. Manus only launched in March this year, but in just a few months, they hit a whopping $125 million ARR, and now Zuckerberg has scooped them right up. This move by a Chinese team going global with AI Agents is seriously setting a new benchmark.

2. In-depth Interview with Manus Founder: Doing a Thousand Small Things Right is More Important Than Three Big Ones]( https://x.com/dotey/status/2006065780253655070 )

Baoyu put together an in-depth podcast interview with Manus founder Peak, and it’s packed with insights. Peak’s core philosophy? Manus chose an ‘intelligence-driven’ rather than ‘rule-driven’ Agent approach. This means no pre-set workflows; the model figures out each step on its own. No MCP? It reads API docs. No API? It simulates human actions in a browser. This ‘doing a thousand small things right’ engineering philosophy might just be why Meta was so impressed.

AI News Image

3. Tencent Hunyuan Open-Sources HY-Motion 1.0: Billion-Parameter Text-to-3D Motion Model]( https://x.com/shao__meng/status/2006020321787134357 )

Tencent just dropped a bombshell: they’ve open-sourced HY-Motion 1.0, a billion-parameter text-to-3D motion model. You can describe actions in natural language and directly generate high-quality 3D character animations. Built on a Diffusion Transformer architecture, it covers over 200 action categories, and its output can plug right into existing 3D animation pipelines. Game developers and animators, you’ll definitely want to check this out!

4. Study Reveals: Pro Programmers Don’t Vibe, They Control]( https://x.com/omarsar0/status/2006063755449504154 )

A recent study on 99 seasoned developers (with 3-41 years of experience) dropped some truth bombs: pro programmers use AI Agents totally differently from the ‘Vibe Coding’ hype you see online. Every single respondent actively controls software design and implementation, and on average, half of the AI-generated code gets tweaked. They ’tame’ AI with detailed prompts, 70+ step planning documents, and project rules. The takeaway is crystal clear: AI is a collaborator, not an autopilot.

AI News Image

5. OpenCode Open-Source Programming Agent: Stars Soar Past 44K]( https://github.com/sst/opencode )

OpenCode, the open-source programming agent from the SST team, is absolutely blowing up! Its GitHub Stars have already rocketed past 44,000. This fully open-source AI coding assistant can run locally and supports multiple models. If you’re a developer looking to set up your own AI coding environment without relying on closed-source services, this is a solid choice.

6. Anthropic Open-Sources Agent Skills Repository: A 30K Star Skill Library]( https://github.com/anthropics/skills )

Anthropic has officially open-sourced an Agent skills repository, and it’s already racked up 30,000 Stars. This repo is a treasure trove of skill modules for AI Agents, making it super easy for developers to quickly build their own Agent applications. If you’re cooking up an Agent product, you can just grab these and go.

7. Google AI 2025 Annual Review: 60+ Product Updates]( https://x.com/GoogleAI/status/2006066855090192384 )

Google AI just dropped its 2025 Annual Review, highlighting over 60 AI product and feature updates this year. From AlphaFold to Veo 3.1, FireSat to Nano Banana, the core Gemini and Nano Banana models are driving the AI transformation across their entire product line. DeepMind has truly been the central engine for Google AI this year.

8. Chatterbox: Most Advanced Open-Source Text-to-Speech, Nearly 20K Stars]( https://github.com/resemble-ai/chatterbox )

Resemble AI’s open-source Chatterbox has already snagged nearly 20,000 Stars, claiming the title of the most advanced open-source TTS (Text-to-Speech) solution out there. If you’re building voice-related products and want to steer clear of closed-source APIs, this one’s definitely worth a shot.

9. Yume-5B: Generate Interactive 3D Worlds from a Single Image]( https://x.com/Gorden_Sun/status/2005922809244201431 )

The open-source model Yume-5B can generate interactive, manipulable 3D worlds from just a single image, similar to Li Feifei’s team’s world model. The model is already up on Hugging Face, so if you’re curious, go have some fun with it!

10. Orange AI Mobile Drawing App Passes Apple Review]( https://x.com/oran_ge/status/2006122179390238869 )

Orange AI’s mobile AI drawing app finally cleared Apple’s review on the very last day of 2025! It was stuck for four days, waiting since Christmas. Now, you can whip up AI-generated images right from your phone, anytime, anywhere.

AI News Image

💡 Worth Noting

[Products]

  • Z Code integrates Claude Code, Codex, and Gemini CLI into a single AI IDE – perfect for those who aren’t super dev-friendly with desktop interactions.
  • Handy is a free, open-source offline speech-to-text app with over 9400 Stars – a godsend for privacy advocates, running completely offline.

[Open Source]

  • vibe-kanban is a Kanban tool designed to make your coding agents 10x more efficient, boasting nearly 9000 Stars.
  • system-prompts is a massive collection of system prompts for mainstream AI tools, with a whopping 100,000 Stars – it’s got everything from Cursor to Windsurf to Devin.

[Research]

  • A new SSM architecture is outperforming Transformer models with O(n) complexity – it’s already surpassed Transformer baselines for shorter sequence lengths.
  • Meta just released a training method for AI co-scientists, using rubric rewards – their 30B model now matches Grok-4-Thinking.

[Business]

  • Tencent Hunyuan just dropped version 1.5 of its open-source translation model – this 1.8B model only needs 1GB of RAM to run offline.

🤔 FAQs

So, how do you actually get your hands on Claude Code?

Claude Code, Anthropic’s AI programming agent tool, currently requires a Claude Pro or Max subscription. For users in China, this might mean dealing with payment hurdles or account registration restrictions.

The fix? Aivora offers Claude-related account services. We’re talking lightning-fast delivery, ready-to-use accounts right after you order, no payment or registration headaches. Plus, you get stable, exclusive accounts with worry-free after-sales support.

Hit up aivora.cn to check out the full list of AI account services.

Today’s Rundown 🚀

Meta just dropped billions to acquire Chinese AI Agent company Manus. Its founder is now a VP, and they hit $125M ARR in just three months. Tencent has open-sourced a billion-parameter text-to-3D motion model, and the open-source coding tool OpenCode has racked up a whopping 44K Stars. Chinese AI is making waves globally, setting a benchmark for going overseas, and the open-source ecosystem is absolutely exploding. Today’s news is definitely worth a deep dive.

⚡ Quick Nav

✨ Heads up! Want to be among the first to try out the latest AI models mentioned here (Claude 4.5, GPT, Gemini 3 Pro)? No account? No worries! Aivora has got your back. Grab an account, get started in a minute, and enjoy worry-free after-sales support.

Today’s AI News

🤯 Just One Sentence

Meta just dropped billions on Manus, and with that, a Chinese AI Agent team has officially crashed the Silicon Valley party.

🔑 3 Keywords

#ManusAcquired #ThousandSmallThingsPhilosophy #AIProgrammingSurge

🔥 Top 10 Heavy Hitters

1. Meta Acquires Manus for Billions, Founder Becomes VP

The absolute mind-blower today? Meta just announced its acquisition of Chinese AI Agent company Manus for billions of dollars. This deal is Meta’s third-largest ever, right after WhatsApp and Instagram. Xiao Hong, the founder, is stepping in as a Meta VP. Manus only launched in March this year, but in just a few months, they hit a whopping $125 million ARR, and now Zuckerberg has scooped them right up. This move by a Chinese team going global with AI Agents is seriously setting a new benchmark.

2. In-depth Interview with Manus Founder: Doing a Thousand Small Things Right is More Important Than Three Big Ones

Baoyu put together an in-depth podcast interview with Manus founder Peak, and it’s packed with insights. Peak’s core philosophy? Manus chose an ‘intelligence-driven’ rather than ‘rule-driven’ Agent approach. This means no pre-set workflows; the model figures out each step on its own. No MCP? It reads API docs. No API? It simulates human actions in a browser. This ‘doing a thousand small things right’ engineering philosophy might just be why Meta was so impressed.

AI News Image

3. Tencent Hunyuan Open-Sources HY-Motion 1.0: Billion-Parameter Text-to-3D Motion Model

Tencent just dropped a bombshell: they’ve open-sourced HY-Motion 1.0, a billion-parameter text-to-3D motion model. You can describe actions in natural language and directly generate high-quality 3D character animations. Built on a Diffusion Transformer architecture, it covers over 200 action categories, and its output can plug right into existing 3D animation pipelines. Game developers and animators, you’ll definitely want to check this out!

4. Study Reveals: Pro Programmers Don’t Vibe, They Control

A recent study on 99 seasoned developers (with 3-41 years of experience) dropped some truth bombs: pro programmers use AI Agents totally differently from the ‘Vibe Coding’ hype you see online. Every single respondent actively controls software design and implementation, and on average, half of the AI-generated code gets tweaked. They ’tame’ AI with detailed prompts, 70+ step planning documents, and project rules. The takeaway is crystal clear: AI is a collaborator, not an autopilot.

AI News Image

5. OpenCode Open-Source Programming Agent: Stars Soar Past 44K

OpenCode, the open-source programming agent from the SST team, is absolutely blowing up! Its GitHub Stars have already rocketed past 44,000. This fully open-source AI coding assistant can run locally and supports multiple models. If you’re a developer looking to set up your own AI coding environment without relying on closed-source services, this is a solid choice.

6. Anthropic Open-Sources Agent Skills Repository: A 30K Star Skill Library

Anthropic has officially open-sourced an Agent skills repository, and it’s already racked up 30,000 Stars. This repo is a treasure trove of skill modules for AI Agents, making it super easy for developers to quickly build their own Agent applications. If you’re cooking up an Agent product, you can just grab these and go.

7. Google AI 2025 Annual Review: 60+ Product Updates

Google AI just dropped its 2025 Annual Review, highlighting over 60 AI product and feature updates this year. From AlphaFold to Veo 3.1, FireSat to Nano Banana, the core Gemini and Nano Banana models are driving the AI transformation across their entire product line. DeepMind has truly been the central engine for Google AI this year.

8. Chatterbox: Most Advanced Open-Source Text-to-Speech, Nearly 20K Stars

Resemble AI’s open-source Chatterbox has already snagged nearly 20,000 Stars, claiming the title of the most advanced open-source TTS (Text-to-Speech) solution out there. If you’re building voice-related products and want to steer clear of closed-source APIs, this one’s definitely worth a shot.

9. Yume-5B: Generate Interactive 3D Worlds from a Single Image

The open-source model Yume-5B can generate interactive, manipulable 3D worlds from just a single image, similar to Li Feifei’s team’s world model. The model is already up on Hugging Face, so if you’re curious, go have some fun with it!

10. Orange AI Mobile Drawing App Passes Apple Review

Orange AI’s mobile AI drawing app finally cleared Apple’s review on the very last day of 2025! It was stuck for four days, waiting since Christmas. Now, you can whip up AI-generated images right from your phone, anytime, anywhere.

AI News Image

💡 Worth Checking Out

[Products]

[Open Source]

[Research]

[Business]

🤔 Got Questions?

So, how do you actually get your hands on Claude Code?

Claude Code, Anthropic’s AI programming agent tool, currently requires a Claude Pro or Max subscription. For users in China, this might mean dealing with payment hurdles or account registration restrictions.

The fix?

  • Aivora offers Claude-related account services.
  • We’re talking lightning-fast delivery, ready-to-use accounts right after you order, no payment or registration headaches.
  • Plus, you get stable, exclusive accounts with worry-free after-sales support.

Hit up aivora.cn to check out the full list of AI account services.

Last updated on
12-30-Daily