◉GEMINI LABJP
●NANOLITE — Nano Banana 2 Lite is here: Google's fastest and most cost-efficient Gemini Image model, made for running lightweight image generation cheaply●OMNIFLASH — Gemini Omni Flash is in public preview, a natively multimodal model that lets enterprises and developers build custom, dynamic video workflows●AGENTS — Managed Agents expand with background: true for async server-side runs and polling, remote MCP server integration, and refreshing credentials across interactions●MEMORY — The Memory Bank IngestEvents API is generally available, decoupling event ingestion from memory generation so you can stream content continuously●THROUGHPUT — Provisioned Throughput now lets you submit up to seven pending orders for the same model and region●DEPRECATE — Image generation models shut down on August 17, and the Grok 4.1 family on the Gemini Enterprise Agent Platform on August 20●NANOLITE — Nano Banana 2 Lite is here: Google's fastest and most cost-efficient Gemini Image model, made for running lightweight image generation cheaply●OMNIFLASH — Gemini Omni Flash is in public preview, a natively multimodal model that lets enterprises and developers build custom, dynamic video workflows●AGENTS — Managed Agents expand with background: true for async server-side runs and polling, remote MCP server integration, and refreshing credentials across interactions●MEMORY — The Memory Bank IngestEvents API is generally available, decoupling event ingestion from memory generation so you can stream content continuously●THROUGHPUT — Provisioned Throughput now lets you submit up to seven pending orders for the same model and region●DEPRECATE — Image generation models shut down on August 17, and the Grok 4.1 family on the Gemini Enterprise Agent Platform on August 20
← Back to Blog

Gemini Weekly Highlights (Mar 28 – Apr 3) — Computer Use Takes Center Stage, Live 3.1 Amazes

weekly highlightsGeminiComputer UseGemini LiveGoogle ADKweekly recap

Hello, I'm Masaki Hirokawa from the Gemini Lab editorial team.

April is here, and the first week has been a remarkable one. If I had to sum up what dominated our publishing schedule this week, it would simply be: Gemini taking control of the computer. We published a series on Computer Use, along with hands-on guides for Gemini Live 3.1, Google ADK with TypeScript, and even AI-generated music. It was a packed week.

Let me walk you through what we covered.

This Week's Standout Series: Gemini Computer Use

I'll be honest — the topic I was most excited to write about this week was Gemini Computer Use.

The idea of AI controlling a browser isn't entirely new, but with the Gemini 3.1 Pro Computer Use API becoming genuinely practical, the experience of actually running it felt different. It felt real.

We approached it in three parts. First, Gemini Computer Use Complete Guide: From Setup to Browser Automation covers everything from API key setup to writing Python code that actually controls a browser. It's designed to get you up and running from scratch.

Next, Gemini Computer Use Practical Guide — Form Automation, Data Extraction, and Browser RPA digs into real-world use cases: automating repetitive forms, extracting structured data from the web, and replacing traditional RPA workflows. I chose examples that would give you that "wait, I used to do this manually every day" moment.

For those thinking about enterprise or team-level deployments, Building Business Automation Workflows with Gemini Computer Use shifts focus to multi-step workflow design — how to chain operations, handle errors gracefully, and build something robust enough to trust in production.

Computer Use is still an experimental capability in many respects, but after spending this week with it, I came away thinking: this is the real deal. I'll keep covering it as it matures.

Gemini Live 3.1 — When Conversation AI Starts Feeling Different

The other highlight that genuinely surprised me this week was Gemini Live 3.1 Practical Guide: Speed Boost, 2× Context, and Emotional Awareness.

The faster response times and expanded context window were expected improvements. But the emotional awareness feature caught me off guard. The way Live 3.1 adjusts its tone based on how you're speaking — it sounds simple when you describe it, yet experiencing it feels qualitatively different. There's a sense of being genuinely heard, not just processed.

AI conversation has always had a slightly clinical quality to it. Gemini Live 3.1 has started to change that, at least in my experience. The guide covers everything from setup to advanced use cases, so give it a try and see what you think.

Google ADK × TypeScript — Multi-Agent Development Comes to the JS Ecosystem

Google ADK × Gemini — Building TypeScript Production Multi-Agent Systems is a piece I'd been wanting to write for a while.

With Google's Agent Development Kit now fully supporting TypeScript for production use, developers coming from frontend or Node.js backgrounds can build sophisticated multi-agent systems on top of Gemini without having to write Python. That's a significant shift in accessibility.

The article goes deep — coordinating multiple agents, robust error handling, and deploying to production. If you've been waiting to build Gemini-powered agents in TypeScript, this is the guide for you.

Also This Week: AI Music and Digital Watermarks

On a slightly different note, we published Complete Guide to Generating 3-Minute Songs with Gemini Lyria 3 Pro.

Specifying lyrics, genre, and emotional tone to generate full tracks — it's a reminder of just how far generative AI has come beyond text. For musicians, content creators, or anyone who needs original background music, the cost of audio production is changing dramatically.

We also published Google SynthID Complete Guide — How AI Content Watermarking and Detection Works. As AI-generated content becomes more prevalent, the ability to verify provenance with transparency matters more than ever. SynthID is a fascinating piece of infrastructure for a more trustworthy AI ecosystem.

Looking Ahead to Next Week

With April underway and Google I/O 2026 on the horizon, the pace of announcements is only going to accelerate. We're keeping a close eye on Gemini 3.1 Pro enhancements, deeper Workspace integrations, and expanded language support.

Gemini Lab will be here to cover it all — with practical, hands-on guides that go beyond the headlines.

Thanks for reading. If you have questions or topics you'd like us to cover, we'd love to hear from you.

For those who want to go deeper on Gemini API development and production deployments, the official Google AI documentation is an excellent resource. We also recommend Generative AI with LangChain by Ben Auffarth for a thorough grounding in building and deploying generative AI applications end-to-end.