# This Week in AI: Chips, Oversight, and Evolving Roles Hardware breakthroughs, government scrutiny, and job role transformations dominated AI news this week. - **Source:** [O'Reilly Radar](https://www.oreilly.com/radar) - **Published:** 2026-07-10 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/5fa2e7dd-0eca-4d4a-a1dc-3b3467b8f59c ## Nanoscale Hardware Leap IBM unveiled 0.7nm chip technology - one-third the width of a DNA strand - using vertical transistor stacking for 50% better performance. OpenAI/Broadcom's Jalapeño chip specializes in LLM inference, while NVIDIA's 45°C liquid cooling reduces data center environmental impact. Physical infrastructure emerges as AI's next frontier beyond algorithms. ## Governments as Gatekeepers The US lifted export controls on Anthropic's Claude models but added cybersecurity classifiers. OpenAI proposed a 5% government equity stake to build public trust. Epoch AI reports vulnerability disclosures spiked 3.5x post-Mythos launch. State oversight is becoming a permanent condition for frontier AI access. ## Specialized Model Families OpenAI's GPT-5.6 tiered preview introduces task-specific Sol, Terra, and Luna models, following GPT-Rosalind for biology. Anthropic's Claude Science integrates research tools for life sciences. The trend moves from generalist models to purpose-built architectures serving niche domains. ## Role Revolution Microsoft and AWS invest billions in AI deployment units. Claude Code's creator identifies 5 functional archetypes: prototyper, builder, sweeper, grower, and maintainer. Traditional titles like 'engineer' fail to capture these hybrid roles bridging customer needs and technical implementation. ## AI-Native Transformation SAP cuts costs to focus on AI talent acquisition. Initiatives like ServiceNow-Accenture's partnership emphasize 'forward-deployed engineers' who embed with clients. Organizations must rapidly restructure internally to operationalize AI ambitions at scale. ## The Inference Multiplier Small efficiency gains in inference - not training - create massive impact when multiplied across millions of daily users. Frontier labs now design custom hardware like IBM's angstrom-era chips. The biggest opportunities shift from software to physical infrastructure optimization. ## Global Access Uncertainties Non-US entities face unclear terms for accessing frontier models. Governments' unilateral control over model availability creates new power dynamics. Proposed public-private equity models aim to redistribute AI's economic benefits to taxpayers while addressing security concerns. ## Sources covering this story - https://www.oreilly.com/radar/this-week-in-ai-chips-checks-and-changing-jobs/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._