AI is becoming better at execution and security, but its actual business output remains debated

Qwen Go! has added a "My Apps" feature, allowing users to build AI tools for daily or professional scenarios just by chatting. The system automatically handles database configuration and deployment. Once authorized, it can access data from platforms like Taobao, Amap, and Damai, and supports sharing with others.

Elon Musk announced that Grok Bot can now place calls to individuals or businesses on behalf of users, handle requests, and record call summaries. The feature supports automated dialing and keypad interactions. The Mac version also added Tailscale private network connectivity and payment account syncing.

Developers released the open-source Skill dimensional-drawing plugin. By pointing it at component source code or screenshots, it generates a technical dimension sheet in an Apple accessory style, clearly displaying spacing, corner radii, and layer architecture instead of just simple text descriptions.

Nvidia has open-sourced its robot video annotation engine, Flora. It uses three fine-tuned Qwen models to handle action segmentation, execution scoring, and result verification. The engine is already used in GR00T training and can annotate 200,000 hours of video per day on 512 GPUs, at 1/125th the cost of GPT-6 Astra.

Nanjing University, SenseTime, and other teams have open-sourced the smartphone agent OpenMobile-2, providing 9B and 27B versions based on the Qwen series. The project includes the complete training environment, data, and reproduction instructions. The 27B version scored 79.9 on the AndroidWorld benchmark.

Torsten Slok, chief economist at Apollo Global Management, found that excluding big tech, the profit margins and revenue per employee of the S&P 493 have not improved due to AI. He believes there is currently no data to support the idea that AI has significantly boosted productivity outside the tech industry.