# Nano-Banana vs Pico-Banana to Google going into Space! *Google's Nano-Banana vs Apple's Pico-Banana to Google going into Space and Whatsapp in iWatch.* By [Aniket Rawat](https://scaleengineer.com/authors/aniket-rawat) Published: 2025-11-08 Canonical: https://scaleengineer.com/news/latest-ai-developments-october-2025 --- **OpenAI’s IndQA for India ** **\-****[ ](https://github.blog/security/supply-chain-security/our-plan-for-a-more-secure-npm-supply-chain/)**OpenAI has launched **IndQA**, a new benchmark designed to test how well AI models understand Indian culture, languages, and context across 2,278 questions in 12 languages and 10 domains\. Built in collaboration with local experts, IndQA aims to make AI more inclusive and culturally aware\. [Read the Official Blog Post\.](https://openai.com/index/introducing-indqa/) **Google steps out of Earth \- **[Google’s Project Suncatcher research ](https://blog.google/technology/research/google-project-suncatcher/)moonshot that envisions deploying solar\-powered satellite constellations in space to run AI workloads using Google\-built TPUs\. By harnessing nearly continuous sunlight and high\-bandwidth free\-space optical links, the project aims to eventually scale AI compute far beyond Earth\-bound data centers\. ## October 2025: Biggest AI Updates You Might’ve Missed The AI world has been buzzing with breakthroughs, platform updates, and new agentic capabilities\. From OpenAI’s latest security\-focused agent to Google and Amazon’s enterprise AI ecosystems, here’s a look at the most significant AI developments from October 2025\. ### OpenAI Unveils “Aardvark” – An AI Security Researcher OpenAI announced **[Aardvark](https://openai.com/index/introducing-aardvark/)**, a new AI agent designed to detect and patch software vulnerabilities automatically\. Unlike traditional tools like fuzzing, Aardvark leverages large language model reasoning to assess exploitability and propose fixes at scale\. The goal is to give defenders an edge in a world where software flaws multiply faster than humans can patch them\. ### Multi\-Agent Coding Takes Off with Cursor 2\.0 AI coding editor **Cursor** released [version 2\.0](https://sdtimes.com/ai/cursor-2-0-enables-eight-agents-to-work-in-parallel-without-interfering-with-each-other/), introducing an interface where up to **eight agents can work in parallel** without interfering with each other\. The update also introduced **Composer**, a new model optimized for fast, low\-latency agentic coding\. This signals a shift toward “multi\-agent collaboration” in programming environments\. ### Workato Brings Agents to SaaS with Enterprise MCP Integration company **Workato** launched **[Enterprise MCP](https://www.workato.com/agentic/mcp)**, a platform that turns existing APIs and workflows into reusable “agent skills\.” It enables LLM\-based agents like ChatGPT or Claude to securely access and act on enterprise data, solving one of the biggest bottlenecks in agent adoption: system integration\. ### GitHub, JetBrains, and Microsoft Lead Developer AI Evolution **GitHub** unveiled **[Agent HQ](https://sdtimes.com/ai/github-unveils-agent-hq-the-next-evolution-of-its-platform-that-focuses-on-agent-based-development/)**, a new command center for orchestrating multiple coding agents across GitHub, Copilot, and VS Code\. Meanwhile, **JetBrains** introduced **[DPAI](https://sdtimes.com/ai/jetbrains-launches-open-benchmarking-platform-for-measuring-ai-productivity/)**** Arena**, an open benchmarking platform to measure AI’s real impact on developer productivity\. **Microsoft**, on the other hand, launched planning capabilities for Copilot in Visual Studio, allowing it to break down complex tasks into smaller, manageable steps bringing structured reasoning to coding\. ### Enterprise Platforms Join the Agentic Era **Amazon** rolled out Quick Suite, an agentic platform that connects AWS with 1,000\+ apps, enabling natural language automation and multi\-application workflows\. **Google** introduced Gemini Enterprise, a unified platform combining advanced Gemini models, no\-code tools, and a governance framework for enterprise\-grade AI operations\. ### Industry Innovations Keep Coming Other notable launches include: - **Anthropic’s Claude Sonnet 4\.5**, now leading coding benchmarks globally\. - **Red Hat Developer Lightspeed** integrating AI across its hub and migration toolkit\. - **OutSystems Agent Workbench**, helping enterprises create custom AI agents from their data\. - **VibeSec** by OX Security, embedding security directly into AI coding models\. ### AI Meets Governance and Safety California made headlines by passing the first AI safety law for frontier models, emphasizing transparency, accountability, and responsible innovation\. This marks a new chapter in aligning AI advancement with public trust\. Read the full details in this [Blog Post](https://sdtimes.com/ai/october-2025-ai-updates-from-the-past-month/) **Nano\-Banana vs Pico\-Banana** \- [Apple has released ](https://github.com/apple/pico-banana-400k)**[Pico\-Banana\-400 K](https://github.com/apple/pico-banana-400k)**, a large\-scale dataset of 400,000 image edit pairs built from real photographs and synthetic edits\. It’s designed to accelerate research in text\-guided image editing by providing high\-quality, instruction\-based data across 35 edit types and multiple subsets\. **Whatsapp inside iWatch** **\- **[Meta is trying out a new WhatsApp Apple Watch app that goes beyond simple notifications\. ](https://9to5mac.com/2025/10/30/whatsapp-now-testing-an-apple-watch-companion-app/)With a TestFlight build, users can scroll chats, reply with voice or text, and react all on the watch\. Full details and timing for public release aren’t final yet\. ### Buzz of the Week: **Semantic Kernel** **Semantic Kernel** is an open\-source framework by Microsoft that helps developers connect traditional code with AI models like GPT\. Instead of treating AI as a black box, it lets you combine normal programming logic \(loops, APIs, data\) with AI reasoning and natural language understanding\. This means you can build smarter apps that “think” in context for example, summarizing documents, planning tasks, or generating content based on user input\. It acts like a bridge between your app’s logic and the AI’s capabilities\. Developers can also add memory, planning, and skill modules to make AI responses more consistent and goal\-oriented\. In short, Semantic Kernel makes it easier to build AI\-powered apps that feel both intelligent and controlled\. ### Things that launched\. Things that went viral\. Things you'll pretend to try\. ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/9765551e-0037-4459-b1c5-75b0e223908c/image.png?t=1751469165) **xTiles** **[xTiles](https://adhd.xtiles.app/)** is a visual board/app for developers that lets you map ideas, code snippets, and research in a flexible canvas\. ##### Rich **[Rich](https://github.com/Textualize/rich)** is a Python library to beautify logs, terminals, tables, and CLI output with minimal code\. **MigrationMiner** [MigrationMiner](https://github.com/hussien89aa/MigrationMiner) is a research\-grade tool that detects and tracks library migrations at the method\-level in Java codebases\. ### **Build Braincells, Not Just Features** This weekend’s read: [The Business Challenge: When Legacy Code Limits Growth\.](https://www.developer-tech.com/news/the-business-challenge-when-legacy-code-limits-growth/) This week’s watch: [Most Disturbing Things Found on Google Maps\.](https://youtu.be/SgHEWQDFGmo?si=JEzOUgDvZj7hjh4U) At last, ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d627c115-b093-4f5e-8c31-76bdd6305b90/image.png?t=1762329517)