The morning briefing.
While you slept: SoftBank's massive AI data center plans in France, Anthropic's new interview policy, and a Netflix engineer's open-source tool to cut AI costs.
AI Infrastructure Investment. SoftBank announced a colossal 75 billion euro investment to build AI data centers in France, targeting 5 gigawatts of capacity by 2031. This significant commitment marks SoftBank's largest AI infrastructure push in Europe, underscoring the global race for compute power. Meanwhile, Anthropic's valuation has reportedly surpassed OpenAI's, with its run-rate revenue now defined by a mix of consumption-based sales and monthly subscriptions. These developments highlight the intense capital flow into scaling AI capabilities.
AI Capabilities and Limitations. Research reveals that leading AI search agents, including GPT-5.4 and Kimi K2.6, often rely on pre-trained knowledge rather than active web research for current information. This limitation causes performance drops when models cannot access memory, challenging their real-time utility. Further, EY Canada's cybersecurity report was found to contain numerous hallucinated citations, emphasizing the ongoing struggle with AI accuracy and the critical need for robust verification in professional contexts. These findings highlight the inherent challenges in ensuring AI reliability.
AI's Workforce and Societal Impact. Anthropic has implemented a policy banning AI tools during job interviews to assess candidates' authentic thinking and ethical reasoning. This move reflects industry concerns about evaluating human skills in an AI-augmented hiring landscape. Separately, a Trump-linked startup, Foundation Robotics Labs, plans to deploy humanoid robots in military applications within 12-18 months, raising significant questions about the future of defense technology. Discussions also continue on AI's role in aged care, with experts stressing that technology should complement, not replace, human interaction.
AI Tools and Hardware Trends. A Netflix engineer open-sourced an app to slash AI inference costs, offering a practical solution for optimizing large AI workloads. This comes as GitHub Copilot's new token-based billing model is causing consternation among developers, signaling shifts in AI tool monetization. In hardware, MSI is launching a new monitor capable of switching between three resolutions and refresh rates, catering to diverse user needs. Additionally, Apple's Beats brand is teasing new over-ear headphones, indicating upcoming consumer electronics innovations.
SoftBank plans 75 billion euro AI data center buildout in France
SoftBank plans to build AI data centers with up to 5 gigawatts of capacity in France, representing its largest AI infrastructure investment in Europe. Facilities worth 45 billion euros are set to be established at three sites in northern France by 2031.
Trump-linked startup plans to put humanoid robots in the military
Foundation Robotics Labs, a startup with ties to the Trump family, aims to deploy humanoid robots in military applications. The company anticipates this deployment within the next 12 to 18 months.
AI search agents confirm existing knowledge instead of researching the web
Researchers found that leading AI search agents like GPT-5.4 and Kimi K2.6 primarily use the web to confirm pre-trained knowledge. Their performance significantly declines when asked about recent events not covered in their training data.
Anthropic bans AI tools during job interviews to assess candidate thinking
Anthropic has prohibited the use of AI tools during job interviews, implementing up to five rounds to test candidates' skills, values, and ethical reasoning. This policy aims to evaluate how applicants genuinely think.
Anthropic study finds men use AI coding agents more than twice as often as women
An Anthropic study revealed that researchers with typically male names use AI coding agents over twice as often as those with typically female names in social science research. This gender gap for coding agents is wider than for general AI use.
EY Canada cybersecurity report contained numerous hallucinated citations
A cybersecurity report published by EY Canada was found to have most of its citations hallucinated. This incident highlights the challenges of verifying AI-generated content in professional publications.
Netflix engineer open-sources app to slash AI bills
A Netflix engineer created and open-sourced an application called Project Headroom designed to significantly reduce AI inference costs. The tool aims to help other companies manage their AI-related expenditures.
GitHub Copilot’s new token-based billing spurs consternation among developers
GitHub Copilot's shift to a token-based billing model has caused significant dissatisfaction among developers. Many users are expressing concern over the new pricing structure.
Can AI deliver more humanity in aged care for Australia’s ageing population?
Australia is exploring the use of AI, including companion robots and virtual experiences, to support its aging population in aged care. Experts emphasize that while technology can assist, it should not replace the essential human element.
OpenRouter raises $113M Series B funding round
OpenRouter, a platform for accessing various AI models, has successfully closed a $113 million Series B funding round. This investment will support the company's continued growth and development.
MSI's next-gen monitor can switch between three resolutions and refresh rates
MSI is introducing a new monitor capable of dynamically switching between three different resolutions and refresh rates. This innovation allows users to optimize settings for various applications, from 4K gaming to high-refresh-rate competitive play.
Microsoft Office 2019 and 2021 for Mac view-only conversion
Microsoft Office 2019 and 2021 for Mac are reportedly converting documents to view-only mode for perpetual license holders. This change impacts users who prefer to own software licenses outright rather than subscribe.
Komi-learn offers continuous memory and self-improvement for coding agents
Komi-learn is a new project that provides continuous memory and self-improvement capabilities for AI coding agents. This aims to enhance the agents' ability to learn and adapt over time.
Lamine Yamal shares pictures teasing new Beats headphones
Spanish soccer player Lamine Yamal posted images teasing upcoming Beats headphones, hinting at a new design and a pink colorway. This suggests a new product launch from Apple's audio brand is imminent.