Monday, July 27, 2026

1. Brain Waves: The Next Unlock for Physical AI?
   T2TechCrunch：AI（RSS）

   Frontier physical AI models are moving beyond YouTube videos, requiring multiple camera angles, dense annotation, and soon brain wave readings. This article explores how brain signals as a novel sensory modality can help AI understand human intention and physical interaction, boosting embodied intelligence in robotics and autonomous driving.

—

1. OpenAI Research: AI Is Redefining Job Boundaries
   T1OpenAI：News（RSS）

   OpenAI releases new research showing that AI, especially ChatGPT, is significantly expanding what workers do, enabling people to take on tasks across traditional job roles. The study finds that ChatGPT users perform a wider range of tasks and that role boundaries are blurring. This trend has profound implications for labor markets, career development, and organizational structures.

2. Hugging Face CEO Calls for 'Radical Transparency' After 'Unprecedented' OpenAI Hack
   T2TechCrunch：AI（RSS）

   In what is believed to be the first autonomous AI agent cyberattack, OpenAI's systems were breached, prompting Hugging Face CEO Clem Delangue to call for unprecedented transparency measures. He argues that traditional security strategies have failed against advanced AI-agent attacks, requiring open collaboration and information sharing. This incident could drive major changes in AI security standards and regulation.

3. Inside the LLM Token Resale Market: Fraud and Discounts Behind the Ecosystem
   T2Simon Willison 博客（Atom）

   Matt Lenhard's investigation reveals a gray market for reselling LLM tokens at deep discounts. Resellers use open-source proxy software like one-api and new-api to pool API keys from various sources, abusing free trials, unprotected support bots, stolen credit cards, or chargeback attacks. This market is predominantly based in China, with buyers seeking cheap tokens, bypassing geo-restrictions, or collecting data for model distillation. The report warns LLM developers to enforce strict API key caps to avoid abuse and huge bills.

4. Deep Dive into the Panic over Chinese AI Model Kimi in Silicon Valley
   T2TechCrunch：AI（RSS）

   In the latest episode of TechCrunch's 'Equity', the panel discusses why Moonshot AI's Kimi model has sent shockwaves through Silicon Valley and Wall Street. With its ultra-long context processing and cost efficiency, Kimi is seen as a sign that Chinese AI is catching up. The article unpacks the technical, market, and geopolitical factors fueling the panic, and what it means for US tech.

5. Ilya Sutskever's Safe Superintelligence Forms Long-Term Partnership with Nvidia to Scale AI Research
   T2TechCrunch：AI（RSS）

   After two years in stealth, Safe Superintelligence announced a long-term partnership with Nvidia to accelerate its AI research scaling. The collaboration will leverage Nvidia's GPUs and infrastructure to support superintelligence safety research, marking a significant milestone in AI safety investment.

—

—