Big tech moves and startups
KPMG survey finds 29% of executives struggle to control AI costs as vendors shift to usage-based billing. Gartner warns AI coding agents could cost more than developer salaries by 2028....
A Pangram study finds 41% of LinkedIn long-form posts are AI-generated, making it the top platform for AI slop. LinkedIn is now cracking down after losing control of its feed....
Meta's Muse Spark 1.1 outperforms Zhipu AI's GLM-5.2 on coding benchmarks while costing $0.26 per task vs GLM's $0.37, signaling a new phase in the AI price war where American companies compete with China on cost....
DeepSeek is designing its own AI chip while Beijing weighs export controls on advanced models. China is building a fully independent AI stack — and the era of open frontier AI is ending....
Tencent released Hy3, a 295B parameter open-source MoE model that allegedly matches GPT-4 and Claude Opus with only 21B active parameters. Available under Apache 2.0, it could disrupt the closed-source AI market....
Mistral CEO Arthur Mensch accuses proprietary AI labs of storing customer data and using it to compete against them. This sparks the open-source vs closed-source debate and raises questions about AI platform risk....
Tech companies are blaming AI for massive price hikes on devices. But the numbers don't add up — margins are rising while component costs fall. The "AI tax" is just a premium pricing strategy dressed as innovation....
Alibaba banned employees from using Claude Code, citing security concerns. The real story is that Alibaba knows its own AI tools can't compete, and the ban is an admission of defeat dressed as cybersecurity....
The AI bubble isn't popping in stock prices. It's popping in agent expectations. Meta admits agents are failing. The industry is retreating to 'vibe coding' and infrastructure sales. The real crash is a narrative collapse....
Kuaishou-backed Kling AI is closing a $3 billion funding round at an $18 billion valuation, backed by Tencent. The Chinese AI video generation company plans a Hong Kong IPO within 12 months and is positioned to dominate global AI video creation....
Meituan trained a 1.6 trillion parameter AI model called LongCat-2.0 entirely on Chinese chips, proving Beijing can compete without Nvidia GPUs. The model beats Gemini and GPT on coding benchmarks....
South Korea's $550 billion 'RAMageddon' memory chip investment coincides with a federal class-action lawsuit accusing Samsung, SK Hynix, and Micron of price-fixing. The lawsuit alleges the Big Three used AI demand as cover to restrict supply and inflate prices by 40-50%, echoing the 2000s DRAM price-fixing scandal....
Amazon engineers are reportedly distilling Anthropic's Claude models to build cheaper internal versions, despite Amazon's $25 billion investment in Anthropic. The practice is apparently allowed under their partnership agreement, highlighting the growing 'distillation arms race' in AI where companies use competitors' models to build replacements....
South Korea announced a $576 billion semiconductor and AI investment plan led by Samsung and SK Hynix, the largest chip commitment in history. The move is driven by existential threat from China's rising chip industry, but faces massive execution risks and depends on AI demand staying exponential forever....
AI startup Lindy ditched Anthropic's Claude for DeepSeek, saving millions in costs. CEO Flo Crivello calls it 'a matter of survival' as American AI labs price themselves out of the market they created....
South Korea's KOSPI index doubled in six months on AI chip hype, but with two companies controlling half the market and retail investors gambling on leveraged ETFs, this mania is heading for a crash....
ByteDance's Seedance 2.5 generates 30-second AI video clips with 4K output and multi-character scenes, leapfrogging Western competitors and threatening Hollywood's creative pipeline....
Zhipu AI surpasses HK$1 trillion market cap as its open-source GLM-5.2 model ranks second globally in coding benchmarks, challenging American AI dominance....
Norway will ban generative AI tools in elementary schools starting late August 2026, with Prime Minister Støre arguing children must first develop basic reading, writing, and math skills before using AI assistants....
A growing number of American developers and researchers are choosing Chinese AI models like DeepSeek and Zhipu over Western alternatives, driven by performance, openness, and practical advantages....
