AI

TL;DR: Today’s AI news spans on-device model deployment, agentic compute and tooling, investor/policy scrutiny, and broader social impacts—alongside concerns about failures and misuse.

Product, models, and AI infrastructure

• Google’s Gemma 4 is pitched as running natively on iPhone with full offline AI inference (see Gemma 4 Runs Natively on iPhone).
• AgentFM describes turning idle CPUs/GPUs into a P2P AI compute grid for agent execution and artifact streaming (see AgentFM – agentfm-core).
• A separate conversation links quantum computing to AI GPU competition, citing D-Wave’s stance and Nvidia-related quantum-AI efforts (see D-Wave CEO on quantum vs AI GPUs).

Policy, markets, and social fallout

• Gartner warns that “AI-assisted mainframe exit” projects are likely to underdeliver, predicting market shrinkage (see Gartner on AI mainframe exits).
• Reuters reports investor scrutiny over OpenAI’s $852B valuation amid strategy shifts (see Reuters: $852B valuation scrutiny).
• The Nation highlights the death of AI whistleblower Suchir Balaji and disputes official conclusions (see The Death of an AI Whistleblower).
• Deepfake and surveillance concerns appear in reports about Apple threatening to remove Grok over deepfakes (see NBC: Grok App Store threat) and criticism of Flock Safety’s AI policing approach (see Stop Flock).
• The Guardian describes Iran’s use of AI-generated meme-style video in social media “communications wars” (see Guardian: Iran vs US social media wars).

Model releases

• Gemma 4 on-device/offline inference for iPhones (per Gemma 4 Runs Natively on iPhone).