🎙️ In today’s episode:

• DeepRoute’s RoadAGI Breakthrough: At the 2025 Force conference, DeepRoute.AI partnered with Volcano Engine to build physical-world agents using the Doubao large model.
• Baidu’s All-in-One AI Camera: Baidu Netdisk (think of Google Drive) just got a major AI upgrade. The new AI Camera lets users edit, scan, translate, and solve problems, all in one app.
• Huawei WATCH 5 Unveiled: Huawei launched its first HarmonyOS AI smartwatch. The WATCH 5 features fingertip SpO₂ tracking, blood pressure risk detection, gesture control, and an onboard NPU.

🔗 Links & Resources:

Huawei WATCH 5: https://consumer.huawei.comDeepRoute.AI: https://www.deeproute.aiBaidu Netdisk: https://pan.baidu.com

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• Red Note debuts its first foundation model: Red Note has launched dots.llm1, built by its Humane Intelligence Lab.
• ModelBest updates MiniCPM to version 4.0: One of 2024’s most downloaded open-source models gets faster and more adaptive.

🎙️ In today’s episode:

• SenseTime x Ant Group: SenseTime’s “Little Raccoon” AI assistant joins forces with Ant Toolbox to offer smart, multimodal data analysis, coming soon to Alipay.
• SiliconFlow’s Big Raise: The AI infra startup lands major funding from Alibaba Cloud, expanding its one-stop model hosting platform SiliconCloud.
• DeepDTAGen: Researchers unveil a new multitask AI model that predicts drug-target affinity and generates target-specific compounds, potentially transforming pharmaceutical R&D.

🔗 Links & Resources:

DeepDTAGen: https://www.nature.com/articles/s41467-025-59917-6

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• Kuaishou Launches “Auto Think” Reasoning Model: The Kwaipilot team unveils Auto Think, an open-source model that switches between “thinking” and “non-thinking” modes depending on task complexity.

• Qwen3-Embedding Series Goes Open Source: Alibaba releases new embedding models in 0.6B, 4B, and 8B sizes. The 8B model tops the multilingual MTEB leaderboard, and the smallest version may be optimized for mobile, just as Alibaba confirms a partnership to support Apple Intelligence in China.
• OceanBase Expands Its “Data and AI” Strategy: The Alibaba-backed database platform now integrates with 60+ AI partners like LangChain and LlamaIndex.

🔗 Links & Resources:
Auto Think (Kwaipilot): https://kwaipilot.ai/Qwen3-Embedding: https://huggingface.co/Qwen/Qwen3-Embedding-0.6BOceanBase x AI Ecosystem: https://en.oceanbase.com/

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• VeBrain: China’s First Generalist Robot Brain: Shanghai AI Lab unveils VeBrain - a unified AI system that lets robots see, think, and act.
• DingTalk 7.7.0: Enterprise AI for Everyone: Alibaba’s enterprise tool DingTalk rolls out a major update, making its AI-powered spreadsheet tool free for all users.
• Meituan Launches JiBai, AI for Hotels: China’s super-app Meituan debuts JiBai, the country’s first hotel-side AI management system. It’s part of a wider AI strategy that includes no-code development tools and a new travel assistant.

🔗 Links & Resources:
DingTalk: https://www.dingtalk.com/enMeituan: https://www.meituan.com/en-US/about-us

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• Manus Launches Text-to-Video Feature: AI startup Manus has released a new tool that turns text prompts into video stories, challenging OpenAI’s Sora.
• Tencent Unveils VISVISE for AI Game Creation: Tencent is building a full-stack AI game development platform called VISVISE, featuring tools like Auto LUV and Skirt AI.
• Shido Bets on Affordable AI Eyewear: Suzhou-based startup Shido Intelligence has raised seed funding to bring smart glasses to the masses.

🔗 Links & Resources:
Manus: manus.imTencent VISVISE: https://visvise.com/index

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• Alibaba Launches AI-Native Coding Tool: Alibaba Cloud just introduced Lingma, its own AI-powered developer environment.
• Kling 2.1 Raises the Bar for AI Video: Kuaishou’s video generation tool Kling just dropped version 2.1, offering faster rendering, better motion realism, and a “Master” edition tailored for professional content creators.
• Dunhuang’s Digital Library Cave Goes Global: Tencent’s AI helps bring ancient Buddhist texts and cave art into the modern era with an interactive, multilingual, AI-powered digital archive from the historic Mogao Caves.

🔗 Links & Resources:
Lingma IDE: https://www.alibabacloud.com/help/en/lingma/getting-started/individual-edition-quick-start?spm=a3c0i.29987553.9641670990.i0.454c172dzS0H69

Digital Library Cave: https://cave17.e-dunhuang.com/

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• China’s First Cardiovascular AI Model: China’s first large language model focused entirely on cardiovascular medicine.
• Tongji University Upgrades Its Campus AI: AI isn’t just transforming businesses, it’s making its way into campus life.
• Astribot S1 Makes a Splash at ICRA: At ICRA 2025, a humanoid robot from China stole the spotlight.

🔗 Links & Resources:

Astribot：https://www.astribot.com/product

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• DeepSeek's Surprise Model Update: Chinese AI startup DeepSeek quietly launches R1-0528, an enhanced version of its R1 reasoning model.

• BioProBench: Peking University introduces BioProBench, a massive new benchmark for evaluating how well language models understand biological protocols.

• Lumos Robotics Gets Funded: Lumos Robotics wraps its third funding round in six months, bringing total early investment to nearly 200 million RMB.

🔗 Links & Resources:

• DeepSeek R1-0528: https://huggingface.co/deepseek-ai/DeepSeek-R1-0528
• BioProBench: https://huggingface.co/datasets/BioProBench/BioProBench
• Lumos Robotics LUS1: https://www.youtube.com/watch?v=artIZt1JEDU&t=90s

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.

🎙️ In today’s episode:

• Tencent’s Answer to Supabase: Tencent Cloud unveils a backend service platform designed for the AI coding era. The AI Toolkit lets developers build, deploy, and manage apps using natural language — solving the “last mile” problem in AI-assisted software development.

• Opera Neon: Opera, now majority-owned by China’s Kunlun Tech, introduces Opera Neon — a browser that acts more like an AI agent. With natural language input and offline capabilities, Neon can code, generate content, and automate tasks.

• Maotai Bets on AI with ModelBest: China’s liquor giant Maotai makes a bold investment in ModelBest, an edge AI startup known for its MiniCPM models that run on devices like smartphones and cars.

🔗 Links & Resources:

Opera Neon: https://www.operaneon.com/

⚠️ Disclaimer: This episode was scripted and narrated using AI tools. All software is licensed for commercial use, and content complies with platform guidelines.