AI News for July 6, 2026 — 7min.ai

[NEWS] Nvidia's next-gen Kyber AI rack slips to 2028 on manufacturing snags: Manufacturing difficulties have pushed Nvidia's next-generation Kyber rack system to 2028, a delay of more than a year, according to SemiAnalysis.
[RESEARCH] Baidu's 'Unlimited OCR' reads dozens of pages in one pass by mimicking human forgetting: Baidu researchers built an OCR model that processes dozens of document pages in a single inference pass while keeping memory use and speed constant regardless of text length.
[NEWS] ByteDance and Alibaba pull AI companion features as Beijing tightens rules: ByteDance's Doubao and Alibaba's Qwen are disabling features that let users build and chat with customized humanlike AI companions, ahead of new Chinese regulations governing human-AI interactions.
[NEWS] UK's Cooper warns AI could pose a 'Hiroshima'-style threat without global rules: British Foreign Secretary Yvette Cooper called AI the biggest security challenge of the decade and urged the US, China, and others to agree international rules before it is too late.
[NEWS] SK Hynix targets $29B Nasdaq listing to tap AI investors: SK Hynix will list on the Nasdaq around July 10, aiming to raise about $29 billion in what could be the biggest-ever first-time share sale by a foreign company.
[NEWS] China's Biren raises $892M to scale GPUs and challenge Nvidia: Shanghai Biren Technology is raising HK$7 billion ($892.5 million) to boost production of its GPUs, joining a fierce domestic battle for Nvidia's share of the Chinese market.
[NEWS] Boston Dynamics' Atlas robot delivers the game ball at a World Cup match: A six-foot Boston Dynamics Atlas humanoid walked pitchside at halftime of Brazil vs.
[RESEARCH] AI tutor posts 0.71-1.30 SD learning gains in Dartmouth course study: A new AI tutor achieved effect sizes of 0.71-1.30 standard deviations in a Dartmouth course, according to a study circulating among researchers.
[RESEARCH] Harvard study: AI-native startups build smaller, flatter teams with fewer juniors: Researchers at Harvard Business School and INSEAD found AI-native startups are 25% smaller, employ about 13% more engineers, and have roughly 15% fewer entry-level workers and managers than non-AI peers.
[RESEARCH] DiscoBench finds AI search agents fail by not asking, not searching: A new benchmark from Tencent Hunyuan and Tsinghua University finds AI search agents rarely fail at multi-step research because of the search itself, but because they don't ask users for clarification when queries are ambiguous.

About 7min.ai · Subscribe to newsletter