The lunch briefing.
Mid-day check: SoftBank commits 75 billion euros to French AI data centers, Apple's AI glasses face delays, and Anthropic bans AI tools in job interviews.
Capital and Compute. SoftBank is making a monumental commitment to AI infrastructure, with plans to invest up to 75 billion euros in France for new data centers. This significant outlay aims to establish 5 gigawatts of capacity by 2031, marking Europe's largest AI infrastructure investment. Such large-scale projects underscore the escalating global demand for compute resources necessary to power advanced AI models. The sheer financial and physical scale of these endeavors highlights the foundational role of hardware in the ongoing AI revolution.
AI Development Challenges. The practical application of AI continues to present complex engineering hurdles, moving beyond theoretical models to systemic problems. AI retrieval at scale, for instance, is now seen as a systems challenge rather than merely a tooling issue, requiring sophisticated architectural solutions. Furthermore, research indicates that leading AI search agents often prioritize confirming existing knowledge over genuine web research, especially when confronted with recent information. This suggests a potential limitation in how current models engage with novel data.
AI Ethics and Policy. The ethical implications of AI are increasingly under scrutiny, influencing company policies and revealing societal biases. Anthropic recently banned AI tools during job interviews, emphasizing human critical thinking and ethical reasoning in its hiring process. Concurrently, an Anthropic study found a significant gender gap in the use of AI coding agents, with men using them more than twice as often as women in social science research. These developments highlight the ongoing need to address fairness and accountability in AI's integration into professional and research environments.
Hardware and Consumer Tech. The consumer hardware landscape is seeing shifts driven by both regulatory pressures and evolving AI capabilities. Apple's anticipated AI smart glasses have reportedly been delayed until late 2027, with a lighter Vision Air model potentially following by 2029. Meanwhile, new EU legislation is set to bring back user-replaceable batteries in portable tech products, emphasizing repairability and longevity. These trends reflect a dual focus on advanced AI integration and sustainable, user-friendly device design.
SoftBank plans 75 billion euro AI data center buildout in France
SoftBank intends to invest up to 75 billion euros to build AI data centers with 5 gigawatts of capacity in France. This represents the company's largest AI infrastructure investment in Europe, with 45 billion euros planned for three sites by 2031.
AI retrieval at scale is becoming a systems problem, not a tooling problem
AI retrieval architectures are evolving beyond semantic similarity and vector search, now facing challenges as a systems problem at scale. Production environments require more complex solutions than simple tooling.
Making sense of the debate over AI psychosis
A debate is emerging regarding whether tech CEOs are "uniquely prone to AI psychosis," as discussed on a recent episode of TechCrunch's Equity podcast. This conversation explores the psychological impacts and perceptions surrounding AI development.
Apple AI glasses launch pushed back to late 2027, Vision Air to arrive by 2029
Apple's AI smart glasses, initially expected early next year, have reportedly been delayed until late 2027. A cheaper, lighter version of the Vision Pro, dubbed Vision Air, may follow by 2029.
1-Bit Bonsai Image 4B Image Generation for Local Devices
A new development called 1-Bit Bonsai Image 4B focuses on image generation optimized for local devices. This technology aims to enable efficient AI image creation without relying on cloud infrastructure.
This model is not a real person: how AI is shaking up fashion
Generative AI is transforming the fashion industry through digital twins and AI-sculpted models. Retailers are using AI-generated imagery to advertise products, with expectations for clear labeling.
AI search agents often confirm what they already know instead of actually researching the web
A study by the Harbin Institute of Technology found that leading AI search agents like GPT-5.4 and Kimi K2.6 primarily confirm pre-trained knowledge rather than actively researching new information. Their performance declines significantly when tested on recent events.
Anthropic bans AI tools during job interviews to see how candidates actually think
Anthropic has implemented a policy banning AI tools during job interviews, conducting up to five rounds to assess candidates' skills, values, and ethical reasoning. This approach aims to evaluate genuine human capabilities.
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 more than twice as often as those with typically female names. This gender gap is wider for coding agents than for general AI use.
I Put a Datacenter GPU in My Gaming PC for £200
An individual successfully integrated a datacenter GPU into a gaming PC for a reported cost of £200. This project demonstrates a cost-effective approach to enhancing personal computing power.
Odysseus – self-hosted AI workspace
Odysseus is presented as a self-hosted AI workspace, offering users a private environment for AI development and experimentation. This tool caters to those seeking greater control over their AI infrastructure.
User-replaceable batteries are coming back in a big way
New EU legislation, including Commission Regulation (EU) 2023/1670, mandates design changes for portable tech products to improve longevity, repairability, and recyclability. This will lead to the return of user-replaceable batteries.
New Apple TV, HomePod mini set to launch this fall; Siri Remote ‘refresh’ possible
Apple is reportedly planning to launch new Apple TV and HomePod mini devices this fall, primarily featuring chip upgrades. The Apple TV 4K may also receive an updated Siri Remote.
Netflix wiz creates app to slash AI bills, then open sources it
A Netflix engineer developed an application called Project Headroom to significantly reduce AI computing costs. The tool has since been open-sourced, potentially offering savings to other organizations.