Researchers at UCSD have successfully implemented DFlash, a block-diffusion speculative decoding method, on Google TPUs to bypass the sequential bottlenecks of traditional autoregressive drafting. By "painting" entire blocks of candidate tokens in a single forward pass rather than predicting them one-by-one, the system achieved average speedups of 3.13x, with peak performance nearly doubling that of existing methods like EAGLE-3. This open-source integration into the vLLM ecosystem optimizes TPU hardware by leveraging "free" parallel verification and high-quality draft predictions for complex reasoning tasks.

LiteRT is a production-ready framework designed to help mobile developers unlock the power of Neural Processing Units (NPUs), overcoming the performance and battery limitations of traditional CPU or GPU processing. By providing a unified API that abstracts away hardware complexities, it allows industry leaders like Google Meet and Epic Games to deploy sophisticated AI models for real-time video, animation, and speech recognition with significantly higher efficiency. The platform further supports developers through benchmarking tools and cross-platform compatibility, enabling seamless AI deployment across mobile devices, AI PCs, and industrial IoT hardware.

A2UI v0.9 introduces a framework-agnostic standard designed to help AI agents generate real-time, tailored UI widgets using a company’s existing design system. This update simplifies the developer experience with a new Agent SDK for Python, a shared web-core library, and official support for renderers like React, Flutter, and Angular. By decoupling UI intent from specific platforms, the release enables seamless, low-latency streaming of generative interfaces across web and mobile applications. Integrating with broader ecosystems like AG2 and Vercel, A2UI v0.9 aims to move generative UI from experimental demos to production-ready digital products.

This week in search I covered, yep, heated Google search ranking volatility kicking in the middle of this week. Google updated its spam policies to say it also applies to Google's AI responses in Search. Google Discover...

Google's John Mueller said, "The indexing API is inundated by bloggers trying to act like legitimate sites." This means that Google needs to be more careful about who and what they accept through the Google indexing API.

Google is testing showing a new icon in the autocomplete search suggestions, as you type your search. The icon has a magnifying glass with the Gemini logo on it. It suggests a longer query, a prompt, and when you click it, it takes you to Google Search but with the AI Overview response expanded and fully open.

Google updated the leading paragraph in the search spam policies to clarify that the policies apply to the Google Search AI responses, such as AI Overviews and AI Mode (or whatever else is AI-generated). Google said, "the Google Search spam policies also apply to generative AI responses in Google Search."

Google Ads is testing a new "Create video" beta feature that has the Gemini logo next to it. So this feature uses Gemini, Google's AI, to create the video for you for your Demand Gen campaigns.

Google is now showing directly hotel booking links directly inside the AI-generated responses within AI Mode. This can lead to sending hotels direct traffic in the AI response, which seems like a win to me.

The blog post outlines the transition of a brittle sales research prototype into a robust production agent using Google’s Agent Development Kit (ADK). By replacing monolithic scripts with orchestrated sub-agents and structured Pydantic outputs, the developers eliminated silent failures and fragile parsing. Additionally, the post highlights the necessity of dynamic RAG pipelines and OpenTelemetry observability to ensure AI agents are scalable, cost-effective, and transparent in real-world applications.

To enable small businesses with AI-powered customer support directly into the WhatsApp Business app, we’ve launched Business AI in India. Available in all native languages in India, this feature enables eligible businesses to respond to customer queries 24/7, capture leads, book appointments, and drive sales — without needing any additional tools or platforms.

Making Every Customer Conversation Count

Business AI on WhatsApp can be customized based on the business’s own information, allowing them to automate responses to frequently asked questions and assist customers with queries related to products and services, pricing, discounts, shipping and more. Starting soon, we will also introduce the ability for Business AI to facilitate payments directly within a WhatsApp chat using UPI.  When there is a more complex query or a specific need that needs to be addressed, a business owner can take over the conversation from the AI agent. Over the coming weeks, Business AI will be available for all eligible businesses to use on the WhatsApp Business app. 

According to a Kantar study (2025), 91% of online adults in India chat with a business on a weekly basis, making messaging the preferred way for Indians to interact with businesses, and WhatsApp is central to that connection between businesses and their customers. Business AI on WhatsApp gives small business owners the tools to be available, responsive, and competitive — at any hour of the day.

“Small businesses are the backbone of India’s economy, and we deeply understand the value of every customer conversation for them. Over the years, we’ve consistently heard that managing high volumes of customer queries with limited resources remains one of the biggest challenges for small businesses. This is where we believe that AI can be a game-changer for them. With the introduction of Business AI on WhatsApp, we’re now putting that power directly into the hands of small businesses — ensuring they never miss a customer query outside business hours or struggle to keep up during peak demand.” – Ravi Garg, Director Business Messaging, Meta India

Small Businesses Already Seeing Early Results

Small businesses across India who have integrated Business AI on WhatsApp Business app are already seeing the impact of Business AI in transforming how they connect with customers and drive growth.

Soil Concept, a 100% plant-based personal care brand, was losing potential customers when queries came in outside business hours. Whereas, The Purple Sunset, a customized gifting business specializing in personalized hampers, was struggling to manage 60-70 daily customer queries alongside other business operations. Both these businesses were part of the select business

“We were losing leads that came in late at night, and now with 24/7 support, our conversion rate has skyrocketed to 80-90%. Setting up Business AI on WhatsApp was incredibly simple — no coding, no complex third-party software. I just uploaded our product catalog and documents, and the AI learned the details and tone of my business. It’s been a game-changer in helping us manage and grow our customer base to over 15,000 customers.” – Tuba Siddiqui, Co-founder, Soil Concept said

“To my surprise, it was an easy process — within a few hours, Business AI learned everything about my business and was able to reply to customers on my behalf, exactly how I would. It’s helped me close 6-7 orders daily directly through queries handled by the AI, and I’ve seen a 40% increase in sales with the potential to grow even further.” – Gunveen Kaur, Founder, The Purple Sunset said

At this time, Indian small and medium-sized businesses must meet certain criteria to be eligible for Business AI and use the WhatsApp Business app. To learn more about business AI on WhatsApp and get started, businesses can visit https://www.facebook.com/business/ai/business-ai/whatsapp.

How to use and set up Business AI

Eligible businesses using the WhatsApp Business app can get started by going to the Tools tab and selecting ‘Your Business AI’, where they can follow a few guided steps to set up the feature.

Once enabled, Business AI on WhatsApp can immediately begin assisting with customer conversations by answering questions, recommending products, sharing key business information based on what the business has already added to their profile and catalog, and help close the sale.

Businesses remain in control at all times — they can step in to respond to customers directly whenever needed, adjust how Business AI on WhatsApp works, or turn the feature off entirely. 

For more information on how to set up your Business AI, visit the Help Center.

The post Introducing Business AI on WhatsApp for Small Businesses in India appeared first on Meta Newsroom.

Genkit is an open-source framework designed to help developers build production-ready, agentic AI applications using TypeScript, Go, Dart, and Python. The framework utilizes a powerful middleware system that intercepts generation calls to inject custom behaviors like retries, model fallbacks, and human-in-the-loop tool approvals. By attaching hooks at the generate, model, and tool layers, developers can ensure high reliability and deterministic control over model outputs. Furthermore, Genkit allows for the creation and stacking of custom middleware, all of which can be inspected and debugged through a dedicated Developer UI.

How to transition from stateless chatbots to production-grade agents capable of managing long-running enterprise workflows, such as HR onboarding, that span days or weeks. It introduces the Agent Development Kit (ADK) and its architectural shifts, specifically using durable state machines and persistent session storage to ensure an agent never loses context during "idle time" or server restarts. By leveraging event-driven webhooks and multi-agent delegation, the tutorial demonstrates how to build resilient systems that "sleep" during pauses and wake up to resume complex tasks with high reasoning accuracy.

Trump’s China summit brought Nvidia, Apple, and Tesla leaders into talks shaped by AI chips, trade pressure, and market-access demands.

The post Trump’s China Summit Turns Into a Big Tech Power Play appeared first on TechRepublic.

Integration of Arm Scalable Matrix Extension 2 (SME2) and the Google AI Edge software stack enables high-performance, on-device generative AI by turning the CPU into a powerful matrix-compute accelerator. Using Stability AI’s "stable-audio-open-small" model as a case study, it outlines a streamlined "Convert, Optimize, and Deploy" pipeline that utilizes LiteRT, XNNPACK, and KleidiAI to automate hardware acceleration. The resulting implementation achieves over a 2x speedup in audio generation and a 4x reduction in memory usage while maintaining high audio quality on Arm-powered mobile devices and laptops.

Google Cloud has introduced a high-performance integration that connects Rapid Storage directly to PyTorch via the fsspec interface to eliminate AI training bottlenecks. By utilizing Google’s Colossus architecture and bidirectional gRPC streaming, the solution offers up to 15 TiB/s aggregate throughput and significant reductions in latency. These improvements allow developers to speed up total training time by 23% with zero code changes required beyond updating the storage bucket type.

The Google Cloud AI Agent Bake-Off highlights a shift from simple prompt engineering to rigorous agentic engineering, emphasizing that production-ready AI requires a modular, multi-agent architecture. The post outlines five key developer tips, including decomposing complex tasks into specialized sub-agents and using deterministic code for execution to prevent probabilistic errors. Furthermore, it advises developers to prioritize multimodality and open-source protocols like MCP to ensure agents are scalable, integrated, and future-proof against rapidly evolving model capabilities.

Google previewed Android 17 with Gemini AI tools, AirDrop-style sharing, privacy upgrades, multitasking changes, and stronger security controls.

The post Top New Features in Android 17 You’ll Notice This Year appeared first on TechRepublic.

Microsoft is retiring “Copilot Mode” in Edge as it builds AI browsing tools directly into Edge on desktop and mobile.

The post Microsoft Retires ‘Copilot Mode’ as Edge Gets Built-In AI Tools appeared first on TechRepublic.

Kevin O’Leary’s Wonder Valley data center project faces scrutiny as job estimates shift and Utah residents raise environmental concerns.

The post Kevin O’Leary’s ‘Wonder Valley’ Data Center Advances as Job Estimates Shift appeared first on TechRepublic.