Google’s new Tensor G6 brings performance, efficiency, and post-quantum security to the Pixel 11 series
THE SO WHAT
Tensor G6 with a 50% larger TPU and Titan M3 adopting NIST post-quantum standards makes Pixel 11 a reference design for on-device AI plus forward-leaning security. Mobile app and AI teams should assume more inference will move client-side and start planning for a world where post-quantum crypto is a user-visible feature, not just a backend concern.
READ THE SOURCE
MORE FROM THE WIRE
Applied AISome Claude users are mad that Anthropic’s new watermarks will catch them cheating at their jobs, classes
Watermarking is turning AI output provenance into an enforcement tool, not just a safety feature. If your workflows rely on undisclosed AI assistance—internal or external—assume detection will get easier and update policies, training, and expectations before HR or customers do it for you.
Applied AIGrok Gets Cursor-Driven Upgrade, Claims to Be Competitive With Top Models
Grok leaning into a cursor-driven interface while carrying reputational baggage around non-consensual content is a reminder that UX and capability are now table stakes—trust posture is the differentiator. Enterprises experimenting with frontier models need a clear red line on abuse histories and content controls baked into vendor selection, not bolted on later.
Twitch confirms Amazon is training its AI models on livestreams
Twitch’s opt-out model means creator content is default fuel for Amazon’s AI stack—data rights are now a negotiation point, not boilerplate. Platforms and brands relying on UGC should expect creators to demand explicit value exchange for training rights, from rev-share to better tools.
David Protein founder's candy launch is full of AI typos. The brand's president says it's intentional 'rage bait.'
Deliberate AI typos as “rage bait” show how fast consumers are learning to read — and react to — AI tells. If your brand leans on AI for copy, assume every artifact is now a potential engagement mechanic or reputational landmine, not a neutral mistake.