Retter Reboot #47

Google Releases Gemini 3 Which Aims for Generative UI and Agentic Workflows

Google’s Gemini 3 Pro positions itself as the company’s smartest model yet, using its Canvas environment to generate not just text but full interfaces, code, and even simple 3D/voxel-style scenes from natural language — though in practice, output quality often falls short of Google’s polished demos. The new Gemini Agent can plug into Gmail, Tasks, and the browser to triage inboxes, create to-dos, and walk through bookings or payments, but early testing shows it can be slow, overcautious with confirmations, and not always reliable, highlighting both the promise and current limits of Google’s push into true assistant-style automation.

Source: The Verge

Google Launches Antigravity, an Agentic IDE Built on Gemini 3 Pro

Google has introduced Antigravity, a free “agent-first” integrated development environment that lets developers delegate complex coding work to autonomous AI agents powered mainly by Gemini 3 Pro. Built as a fork of Visual Studio Code, Antigravity offers a familiar editor view plus a manager view where multiple agents can be orchestrated in parallel across workspaces, with direct access to the editor, terminal, and an integrated browser. Instead of opaque tool calls, agents produce verifiable “artifacts” such as task lists, implementation plans, screenshots, and browser recordings, aiming to make long-running, asynchronous coding workflows more trustworthy and auditable for developers.

Source: The New Stack

Google’s Nano Banana Pro Image Model Is Now Free to Try

Google has upgraded its viral Nano Banana image model into Nano Banana Pro (Gemini 3 Pro Image), promising higher-quality, 4K-capable image generation and editing with better text rendering, richer world knowledge, and support for blending up to 14 images and five people into a single scene. The model targets use cases like infographics, posters, and multi-language designs, while offering fine-grained editing controls (local edits, lighting, focus, color grading) and embedding C2PA metadata to help platforms label AI-generated content

Source: The Verge 

ChatGPT Rolls Out Group Chats to All Users Worldwide

OpenAI is launching group chats in ChatGPT globally across Free, Go, Plus and Pro plans, letting multiple people and the AI collaborate in a single conversation to plan trips, co-work on documents or brainstorm ideas together. Users can create shared threads where everyone sees the same context and history, positioning ChatGPT not just as a personal assistant but also as a real-time collaboration hub that can summarize, suggest next steps or pull in external knowledge for the whole group.

Source: TechCrunch

Cloudflare Explains Outage That Took Down ChatGPT, X, and More

Cloudflare has released a postmortem on its November 18 outage, confirming that a misconfigured database query in its Bot Management system generated a bloated configuration file that crashed core proxy services and caused widespread 500 errors across major sites including ChatGPT, X, Canva, and others. The company stresses the incident was not a cyberattack but a self-inflicted configuration and tooling failure, and says it’s rolling out safeguards such as better query validation, stricter config size limits, and slower rollout mechanisms to prevent a single buggy file from disrupting a large chunk of the internet again.

Source: The Verge 

EU Proposes to Weaken Parts of GDPR and the AI Act to Boost Growth

The European Commission has proposed a “Digital Omnibus” package that would ease key parts of the EU’s landmark tech regulations, making it easier for companies to share anonymized and pseudonymized personal data and explicitly allowing personal data to be used for AI training as long as other GDPR rules are respected. The plan also delays and softens some high-risk requirements in the AI Act until supporting standards and tools are ready, reduces cookie pop-ups via browser-level controls, and simplifies compliance for smaller companies 

 Source: The Verge

AI Agents Forecast to Drive 22% of Global Cyber Week Online Sales

Salesforce forecasts that this year’s Cyber Week will hit $334 billion in global online sales, with AI agents influencing 22% of that total — about $73 billion — as shoppers increasingly rely on AI to research products, manage budgets, and plan purchases. Mobile will dominate the action, with 70% of orders and 80% of traffic expected to come from phones, one-quarter of all purchases completed via mobile wallets, and deep discounts in categories like apparel, health and beauty, and home driving both online and in-store demand.

Source: Retail TouchPoints

TikTok Lets Users Tune How Much AI-Generated Content They See

TikTok is rolling out a new control in its “Manage Topics” settings that lets users choose how much AI-generated content appears in their For You feed, using a slider to dial AI videos up or down rather than turning them off completely. Alongside the control, TikTok is deploying invisible watermarking and C2PA Content Credentials to more reliably label AI-made clips—even when they’re edited or reuploaded—and launching a $2 million AI literacy fund to support education around AI safety and transparency.

 Source: TechCrunch

Perplexity Brings Its AI Browser Comet to Android

Perplexity has launched its AI-first Comet browser on Android, bringing most of the desktop version’s capabilities to mobile, including AI-powered search, the ability to “mention” open tabs in questions, voice queries across all tabs, and automatic multi-tab summaries — plus a built-in ad blocker and upcoming features like a cross-site conversational agent, quick-action shortcuts, and a password manager. The company says iOS support is coming next.

Source: TechCrunch 

Perplexity Teams Up with PayPal for Free Agentic Shopping Tool

Perplexity AI is launching a free “agentic shopping” feature in the U.S. that lets users buy products directly from search results, with PayPal integrated to handle payments, buyer protection, and merchant-of-record duties across more than 5,000 merchants. The tool builds on Perplexity’s earlier paid “Buy with Pro” option but will be available to all users ahead of the holiday season, using past search history to better detect shopping intent and personalize recommendations as it joins OpenAI and Google in the race to make AI-native, one-stop shopping experiences.

 Source: CNBC

Google’s WeatherNext 2 Aims to Make Forecasts the Most Useful Everyday AI

Google is rolling out a new AI-powered forecasting model called WeatherNext 2 across Search, Gemini, Pixel Weather, and Maps, using a “Functional Generative Network” to simulate hundreds of possible futures from basic variables like temperature, wind, and humidity in under a minute, delivering faster, higher-resolution, and more accurate predictions up to 15 days out while exposing uncertainty more clearly so people, utilities, and emergency services can better plan around storms, heat waves, and renewable energy output.
Source: TechRadar