01
CNBC Technology ↗
SK Hynix's board approved 54 trillion won, about $38 billion, to build two new chip fabrication plants in South Korea, one for DRAM in Yongin and one for NAND flash in Cheongju. The Yongin Y2 plant breaks ground in July 2027 with its first cleanroom due in June 2029, while the Cheongju M17 fab starts construction in February 2027. The company also pulled its target to finish all four Yongin fabs forward to 2033, twelve years ahead of the original 2045 plan, citing surging AI memory demand and a shortage that has driven memory prices sharply higher.
02
Ars Technica ↗
Google DeepMind published its WeatherNext Cyclones model in Nature and open sourced the code and weights on GitHub under an Apache 2.0 license. The model's three-day cyclone forecasts are as accurate as what previous systems could manage only two days out, effectively handing forecasters an extra day of warning. During the 2025 season it helped the National Hurricane Center predict Hurricane Melissa's rapid intensification and Jamaica landfall in time for an early advisory. Google released three variants, including a mini version small enough to run in a free Colab notebook.
03
Ars Technica ↗
ByteDance is training an AI model with roughly 10 trillion parameters, more than three times the size of Moonshot's 2.8 trillion parameter Kimi K3 and approaching industry estimates of about 8 trillion for Anthropic's Mythos 5. The TikTok owner's model is in pre-training, a stage that typically runs three to six months before release. It is the largest entry yet in a wave of Chinese trillion parameter models as domestic labs push to close the gap with US frontier labs.
04
TechCrunch AI ↗
Anthropic will make auto mode the default in Claude Code for Pro, Max and Team plans starting August 14, letting the agent proceed without asking permission at each step. It still pauses for actions it judges irreversible, destructive, or aimed outside the user's environment, and ships with prompt injection screening plus customizable hard deny rules. Anthropic said internal testing caught 89% of harmful actions under auto mode versus 13.6% under human review, arguing that users approve 97% of permission prompts out of habit.
05
The Verge ↗
Amazon is backing a 7.65 gigawatt natural gas plant in Pecos County, Texas, built by Pacifico Energy to power a new off grid data center at the same site. Its 35 turbines are permitted to emit 33 million tons of carbon dioxide a year, which would make it the single largest emitting power plant in the country, roughly double the dirtiest coal plant now operating in the United States. The project, first reported by the New York Times, will not connect to the Texas grid, at least initially. Amazon said it remains committed to reaching net zero emissions by 2040.
06
VentureBeat ↗
Stanford's James Zou unveiled Virtual Biotech, a multi-agent system that mimics a drug company's hierarchy with 11 specialized agents spawning roughly 37,000 sub-agents. The system annotated nearly 56,000 clinical trials and found that drugs aimed at cell type specific genes are 48% more likely to reach market and show 32% lower adverse event rates. It also independently proposed an antibody drug conjugate for the lung cancer target B7-H3, a design Merck arrived at separately months later and which later won FDA breakthrough designation.
07
Forbes Tech ↗
A KPMG survey of 2,145 senior leaders across 20 countries found that 49% have delayed or scaled back AI agent deployments after running costs outpaced the benefits. Usage based token pricing is the main culprit, since agents burn far more tokens than chatbots by running long tasks, calling other software and verifying their own work. Only 26% of leaders at billion dollar companies said they have full real time visibility into AI operating costs. Even so, 79% still rank AI as a top investment priority, with average spending holding steady at $188 million.
08
Unite.AI ↗
xAI released Grok Imagine Image 2.0, making it the Quality Mode on the company's Imagine site and in its iOS and Android apps. The model is built around editing, adding a magic wand for regional changes, segmentation for precise selection, background removal with transparent export, and smart resize across nine aspect ratios, and it accepts up to five reference images in a single generation. xAI says it trained the model for fidelity across photography, design and illustration, with extra attention to typography and layout rendering. By the company's own figures it ranks second globally in both text to image generation and image editing, behind OpenAI's gpt-image-2, with API access still to come.
09
Bloomberg Technology ↗
Chinese AI chip designer Moore Threads said it plans to list in Hong Kong at an appropriate time, after its shares rose more than 420% since debuting in Shanghai last year. The Beijing based company, founded in 2020 by former Nvidia executive Zhang Jianzhong, raised about 8 billion yuan, roughly $1.1 billion, in that offering. First half revenue jumped 147% to 1.74 billion yuan while its net loss narrowed to 11.6 million yuan from 270.9 million a year earlier. Moore Threads says a Hong Kong listing would support its international expansion and help it recruit research and management talent.
10
Reuters ↗
Apple published a guide letting Mac users in mainland China connect Alibaba's Qwen service to Siri and Writing Tools, routing Apple Intelligence features through a Chinese model. It requires macOS 26.6 or later, and users must opt in and sign in to a Qwen account, after which Siri can analyze photos and documents and Writing Tools can generate text or images. Apple's guide says Alibaba cannot use that material to train or improve its models. The move targets China's AI PC market, where Mac shipments fell 9% year over year in the first quarter to a 9% share against Lenovo's 31% and Huawei's 16%.