Thursday, 27 August 2026

AI Mode now tracks flight prices in 180+ countries and books hotels through Booking.com, Expedia, and other partners, paid via Google Pay.

Google's video model gains scene extension to 40 seconds, keyframe control, and a cheaper 360p draft mode, already live in Adobe Firefly and Runway.

OpenAI, Anthropic, Google, Microsoft and cybersecurity firms signed an open letter urging coordinated defense as AI-enabled cyberattacks scale up.

ChatGPT's Free and Go tiers in India start showing ads next month, with a 50-brand pilot and agency partnerships with WPP and Omnicom.

AWS is adding 2 million Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs by 2028, tripling an order placed just five months earlier.

Anthropic will rent about $45B in AI compute from Nscale over six years, running on Nvidia Vera Rubin chips from a West Virginia data center.

Nvidia posted $96.2B quarterly revenue, up 117% in data centers, then gave its first-ever forward guidance implying $673B in FY2028 sales.

The Model Hardware Standard lets AI agents drive lab and robotics equipment through common read/write commands, cutting integration from months to hours.

New research shows compressing the KV cache is 1.2 to 2x cheaper than tensor parallelism across configurations for serving large-context models.

Particle's Radar transcribes and indexes podcasts with MCP access, so AI agents can search, quote, and clip audio they otherwise can't parse.

Plaud One earbuds pair with an eSIM-enabled case so you can instruct connected AI agents without a phone or laptop nearby, from $249.

OpenAI published research arguing students who use ChatGPT alongside explicit critical-thinking training show broader, better answers, not just faster ones.

Microduck waddles, picks up objects, and recovers from falls, with its SDK, simulation environment, and RL training stack open on GitHub.

Google DeepMind tested Gemini Flash Lite under encrypted double-blind evaluation, hiding model weights from testers and test prompts from Google.