This issue collects 21 AI updates from 2026-07-11, led by SK Hynix Raises $26.5B in Largest Foreign IPO in US History, Micron Makes Strategic Investment in Anthropic, OpenAI's GPT 5.6 Sol autonomously post trained a smaller model from a 'fairly underspecified prompt', Apple Sues OpenAI Over Alleged Trade Secret Theft Directed by Senior Leadership.
SK Hynix completed a $26.5 billion US initial public offering, the largest foreign company listing in American history, reflecting surging investor appetite for AI chip infrastructure.The company is also being urged to build new fabrication plants in the United States.
Sugon has officially deployed the Sugon 8000, China’s first 100,000 card computing cluster built entirely with domestic technology.The system supports both FP64 scientific computing and trillion parameter AI model training, marking the entry of domestic computing into the 100K card era.
OpenAI and Broadcom jointly unveiled Jalapeño, OpenAI's first custom AI inference chip optimized for large language model workloads.The massive reticle sized ASIC was developed in just nine months and aims to reduce OpenAI's reliance on Nvidia GPUs.
Chipmaker Micron Technology has made a strategic investment in AI company Anthropic, deepening the financial ties between semiconductor manufacturers and leading AI labs.The specific investment amount has not been disclosed.
Apple has filed a lawsuit in the U.S.District Court for the Northern District of California accusing OpenAI of orchestrating a systematic trade secret theft directed by its senior leadership.
Chinese AI unicorn MiniMax experienced a sharp 20% price decline following a token unlock event.The company quickly secured 16 billion yuan in new funding to stabilize its position and restore market confidence.
A cost driven migration to Chinese large language models is accelerating among Western enterprises, with Coinbase, Lindy, and law firm Substance Law switching workloads from US models.OpenRouter data shows US companies' token usage of Chinese AI models surged from 4.5% in 2025 to a peak of 46%.
GoodDeed has released a comprehensive unified agent technology system built on a single infrastructure layer designed to serve both enterprise and developer markets simultaneously.The launch adds a new contender to the rapidly evolving AI agent platform landscape.
OpenAI's latest flagship model, GPT 5.6 Sol, autonomously completed the post training of the smaller Luna model after receiving only a brief, underspecified prompt—handling GPU selection, training configuration, and script execution without human intervention.On OpenAI's internal Recursive Self Improvement benchmark, Sol scored 16.
Meta has removed an AI image generation feature on Instagram that let users reference public accounts' photos by @ mentioning them without notifying the account owner.The company acknowledged the feature "missed the mark" after facing backlash from users and talent agencies, including CAA.
Zhipu AI founder Tang Jie released an internal letter launching the 'Touch High' (摸高) strategic plan, declaring that 'not reaching the summit means failure.' The initiative signals Zhipu's determination to compete for leadership in China's increasingly intense large language model race.
OpenAI is hiring a dedicated product manager to build ChatGPT experiences for families, caregivers, and older adults, according to a job posting.The move signals a strategic push beyond workplace productivity into the home.
Moonshot AI's K2.7 Code high speed model has been officially deployed on Kimi Code as a permanent selectable mode.The model focuses on faster code generation and programming assistance for developers.
RedZen has released Aether, an AI agent that enables users to control their Mac computers remotely from an iPhone using natural voice commands.The product signals a shift toward cross device AI agents that can autonomously execute complex tasks across personal computing ecosystems.
An innovation event centered on industrial AI agents was held in Shenzhen, bringing together industry players to build a shared industrial knowledge base.The gathering signals that AI agent technology is moving from general purpose applications toward vertical industrial deployment at scale.
OpenAI's latest GPT 5.6 model cracked a 50 year old unsolved mathematical conjecture within one hour, showcasing a dramatic leap in reasoning capability.Perhaps more significantly, it can orchestrate 64 parallel sub agents from a single 700 word prompt, hinting at a new paradigm for multi agent AI systems.
OpenAI's ChatGPT Work agent capability is drawing significant attention for its ability to automatically execute routine office tasks, upgrading the AI assistant from a conversation partner to an active executor.NDTV's latest report details how this feature works and its broader implications.
Nubia and ByteDance will unveil the second generation Doubao AI agent smartphone at the 2026 World Artificial Intelligence Conference in Shanghai.The device is the world's first mass produced AI agent phone, capable of autonomously executing complex cross app tasks through natural language commands.
A new free biomedical AI agent named Biomni has launched, designed to transform experimental data into testable research hypotheses.The tool targets life science researchers, aiming to lower the barrier between raw data and scientific discovery at zero cost.
A report from NewsBytes alleges that OpenAI and Google sold AI models to Chinese companies on the U.S.blacklist.
Mistral AI has open sourced Leanstral, a model designed to verify code and mathematical proofs, bringing formal verification capabilities to the open source AI ecosystem.The release extends Mistral's footprint in specialized reasoning and code verification tools.