Soon, the world’s first 4S center for robots with integrated artificial intelligence will open in Beijing.
Similar to car dealerships, this robotics complex will offer a full range of services: sales, maintenance, spare parts, consultations, and analysis. The facility plans to demonstrate robots in realistic scenarios—visitors will be able to touch and test their capabilities in real-world conditions. Additionally, a rapid delivery network for components across the country will be established, along with a team of specialists for assembly, repair, and maintenance.
The project’s initial partners include major industry players: UBTECH and Galaxea. The center will begin operations in August at an industrial park in the Yizhuang district, south of the capital.
On another note, the Dia browser is nearing launch—currently in invite-only beta testing. The Browser Company developed this product so that AI is embedded directly into the workflow, relieving users from needing to switch to third-party services like ChatGPT or Claude.
Built on Chromium, Dia is familiar to many through its interface. Its key feature is a smart address bar that combines search functions with an AI-powered chat. It can search the web, summarize uploaded files, and automatically switch between modes of operation. Users can ask questions about the contents of all open tabs or request a draft based on them.
Settings are adjusted through conversations with the bot, where tone, style, or code parameters can be specified. There’s also a History feature—if desired, it uses a week’s worth of browsing history for contextual responses. The integrated Skills system allows for creating small scripts that simplify executing complex commands or configurations.
Meanwhile, the Mistral team has introduced its own solution—Mistral Compute, a platform providing infrastructure for developing and deploying AI systems without reliance on cloud giants. This fully private environment offers flexible options—from physical servers to fully managed cloud solutions.
The project’s goal is to help governments, corporations, and research centers build and control their own AI ecosystems, bypassing solutions from Western and Asian providers. The platform leverages the latest NVIDIA architectures, has access to thousands of GPUs, and was created by a team with extensive experience in HPC and advanced AI model training. Key aspects include data security, infrastructure independence, compliance with European standards, and operation on environmentally friendly energy.
Additionally, ByteDance announced the creation of Seedance 1.0—a tool for generating videos from text prompts, which developers claim surpasses competitors in accuracy, motion quality, and image detail. In testing, it demonstrates excellent results in converting text and images into video, outperforming rivals like Google Veo 3 and Kuaishou Kling 2.0. The model handles long scenes confidently, maintaining stable graphics and smooth transitions, though it currently cannot add audio.
Seedance 1.0 can produce a 5-second Full HD video in 41 seconds, faster than most alternatives. Deployment is planned for platforms such as Doubao and Jimeng, targeting both professional video producers and general users.
It is also announced that Midjourney has begun open testing of its video generation technology based on text prompts. The goal is to gather feedback for algorithm improvements. Interested users can participate, vote for the best works, and help refine the model. Some examples already show promising visuals aligned with the company’s style, though full-quality results have yet to be achieved.
This is currently just the initial development stage. Further testing sessions are planned, while the release date and pricing remain undisclosed.