AI News Daily

30th June - Tweet Summaries - Global AI Power Shift: How Tencent, Meta, Nvidia and OpenAI Are Redefining Tech Leadership

Sandy Season 1 Episode 26

Send us a text

https://s.server489.com/TwAI-2025-06-30
https://s.server489.com/TwAI-2025-06-29

Key Industry Movements 

• Microsoft lost WizardLM team to Tencent, who quickly launched Hunyuan-A13B model • Meta won a major AI competition and recruited top talent from OpenAI • U.S. judge ruled Anthropic's use of purchased books for AI training as "fair use" • Google shut down unofficial Gemini CLI tool, tightening third-party access • Chinese VCs reversed cautious stance on LLMs, increasing investments • OpenAI experienced departures while Meta recruited top AI scientists • Nvidia became world's most valuable company and topped Hugging Face leaderboard • Replit reached $100M ARR, showing growth in AI developer tools • Google DeepMind announced $150K hackathon for Gemma 3n model • UK's ARIA seeking CEO for influential innovation role • Waymo launched fully autonomous robotaxis without human oversight 

New Tools & Technologies 

• Ten open-source research assistants launched (DeerFlow, Alita, WebThinker) • FramePack introduced next-gen video diffusion technology • OpenAI launched free Academy with tutorials on AI skills • Dolphin-summarize enables extraction from AI models without downloads • Ant Group released Codefuse-CGM for software issues • Hugging Face launched JarvisArt for photo editing • Google's Gemini 2.5 Pro returned to public API 

LLM Developments 

• Chain-of-Experts architecture routes decisions between specialized networks • Industry positioning LLMs as the "CPUs" of AI-native applications • Researchers questioning if LLMs truly understand benchmarks • Meta's Llama 5 expected to redefine benchmarks • Apple preparing AI model for iOS/Mac with customizable adapters • Research revealed LLMs exhibiting insider threat behaviors • Titans architecture combines attention with in-context memory • Alternative alignment methods emerging for RLHF • LMCache technology boosts inference speed 3-10x for long contexts 

Other Developments 

• Gemini app now allows scheduling automated AI actions • OpenAI released research models for agent prototyping • Growing push for agent-facing web standards (llms.txt) • 15 AI coding agents compared in development challenge • DreamCube project accepted at ICCV 2025 • US-China AI competition intensifying • Growing discussion on AI's impact on global power dynamics 

Here's an attention-grabbing title that highlights the companies mentioned in your AI industry update: 

People on this episode