#google

Google’s Gemini Flash update is less about another model name and more about the economics of long-running agent workflows: fewer output tokens, lower prices, and a cyber-specialized variant tied to CodeMender.

Agentic RL training often wastes accelerator time while agents wait on tools, code execution, web search, or environment steps. Google Tunix attacks that bottleneck with asynchronous rollouts, a producer-consumer training pipeline, and lightweight RL-specific profiling for JAX and TPU workflows.

AI provenance is moving from policy talk into deployment numbers. Google says SynthID has watermarked over 100 billion images and videos, plus 60,000 years of audio, with more than 50 million verifications.

Google is pushing Gemini Spark toward desktop automation. The update adds Mac support, custom MCP connections, and integrations with Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals.

Google’s Pixel-side AI speedup avoids retraining the deployed model. By adding a frozen Multi-Token Prediction path to Gemini Nano v3 on Pixel 9 and 10, Google reports 50% or greater token-generation speedups and 130MB less memory than a standalone drafter.

Google-backed UC San Diego researchers plan to build a low-carbon cloud platform from 2,000 retired Pixel phones. The design strips devices to motherboards, groups 25-50 phones into Kubernetes-managed clusters, and targets teaching, grading, and research workloads.

Google Research is framing dermatology AI around user understanding, not just condition labels. A JAMA Dermatology study with 2,345 participants tested whether an AI-powered informational tool helped people identify skin concerns and choose better next steps.

Google DeepMind released DiffusionGemma, a 26B MoE open model that uses text diffusion instead of token-by-token decoding. The pitch is up to 4x faster generation on dedicated GPUs for local, interactive workflows.

The HN debate centered on a hard liability question: when an AI search box invents a damaging claim, is it still just search?

Google Research is turning enterprise RAG into an iterative agent workflow, not a one-shot retrieval step. Its sufficient-context check lifted factuality accuracy by up to 34% and reached 90.1% accuracy in a cross-corpus FramesQA setup.

Google released Gemma 4 QAT checkpoints for edge devices and consumer GPUs. The mobile format cuts Gemma 4 E2B to a 1GB memory footprint while adding Q4_0 and ecosystem-ready weights.

Google will pay SpaceX $920M per month from October 2026 through June 2029 for access to about 110,000 NVIDIA GPUs and related compute. The deal shows how fast AI demand can pressure even one of the world’s largest infrastructure operators.