← Back to Blog
AI News

Unitree Targets Shanghai IPO After Shipping 5,500 Humanoid Robots as Anthropic Pushes Opus 5 Token Efficiency

Unitree humanoid robots Shanghai IPO Reid Hoffman Mark Pincus Prentis Anthropic Opus 5 token efficiency Midjourney Co-Star OpenAI AI keypad Canadian legislator LLM Wang Xingxing

Hangzhou-based Unitree shipped 5,500 humanoid robots in 2025, capturing over 25% of the global market, according to Time. Founder Wang Xingxing launched the company after attending a UFC fight and heading straight to his lab, and Unitree is now preparing for a Shanghai IPO as humanoid robotics competition intensifies across China and the United States.

Prentis, a new AI lab co-founded by LinkedIn co-founder Reid Hoffman and Zynga founder Mark Pincus, is in talks to raise $100 million, per Bloomberg. The startup is betting that automating routine computer tasks will soon outpace coding as AI's biggest use case, positioning itself alongside established players like OpenAI and Anthropic in the race to build general-purpose digital agents.

Anthropic released Opus 5 with a focus on token efficiency rather than a raw capability leap, according to TechCrunch. The model reflects an industry shift where cheaper options are increasingly good enough for most applications, putting pressure on premium pricing as competitors like Google's Gemini and Meta's Llama 3 continue to close the performance gap.

Midjourney acquired the personalized astrology app Co-Star, Bloomberg reported on Thursday. The AI startup, which has evolved from generating cat images to producing full-body ultrasound scans, is now expanding into consumer astrology with Co-Star's free daily horoscope platform that serves millions of users.

OpenAI launched a new AI keypad designed for developers, according to The Verge. The hardware accessory will appeal to some coders who want quick access to AI tools but may mystify the broader public, marking OpenAI's continued push beyond software into physical interfaces.

A Canadian legislator read what appeared to be a direct large language model response during a floor speech, including the phrase "Here's a more natural, flowing version of that section," per multiple reports. The incident highlights growing concerns about AI-generated content in government proceedings as tools like ChatGPT and Claude become more widely used by political staff.

Sources:

Share this post: