Google released Gemini 3 Flash

Welcome to AI Toast!

Here’s what’s on the menu today:

• Gemini Flash is out ⚡

• Claude is now in your Chrome

• Superhero Tools for Image Editing

• 10 steps to build your AI Foundation with dbt.

• Quick Toasts: Why does every company want the Indian market?

Total read time: About 5 minutes, perfect for a quick coffee break.

Google's Flash Model Just Beat Its Own Pro

Google released Gemini 3 Flash and the model outperforms its predecessor across 18 out of 20 benchmarks while delivering responses three times faster.

• Gemini 3 Flash combines Pro-grade reasoning with Flash-level speed and costs less than a quarter of Gemini 3 Pro's price.

• The model achieves 90.4% on GPQA Diamond, a PhD-level reasoning benchmark, and scores 78% on SWE-bench Verified for coding agent capabilities.

• It delivers 218 tokens per second compared to the 70-80 tokens per second from Gemini 2.5 Pro, making it 175% faster in output generation.

• The model uses 30% fewer tokens on average while maintaining accuracy, which reduces both latency and cost.

• Gemini 3 Flash now serves as the default model in the Gemini app and is available through Google AI Studio, Vertex AI, and Antigravity.

Top Three Use Cases

• Agentic coding workflows allow developers to iterate rapidly with low-latency code generation and real-time feedback loops.

• Multimodal content analysis processes videos, images, and documents simultaneously, turning visual content into actionable plans within seconds.

• Real-time gaming assistance provides guidance by analyzing video and hand-tracking inputs with complex geometric calculations during live gameplay.

Personal Take: The future looks like a compression race. When a Flash model beats a Pro model on benchmarks while costing 75% less, the industry moves toward efficiency over raw parameter count. Expect more models that prioritize speed and cost without sacrificing intelligence, pushing AI into real-time applications that were previously too expensive to scale.

10 steps to build an AI-ready data foundation

AI runs on data. Data runs on dbt.
Here are 10 steps to build an AI-ready data foundation 👇

AI is reshaping enterprise data usage, but without structure, it's just guesswork. The O’Reilly report "Structured for Intelligence" highlights the importance of structured context for conversational analytics and data development.

Download here: Click here

It details how technologies like dbt and the Model Context Protocol (MCP) enable governed and scalable AI analytics. This report offers a practical guide for integrating natural language interfaces, copilots, and AI agents into production safely.
-----------------------
10 steps to build your AI foundation with dbt.

1. Define AI Goals: Identify the capabilities or ML projects you're building. Structure data to support these workflows.

2. Standardize Context: AI needs structured, trusted standards. Use dbt to define and test core business logic, ensuring consistent data for AI training.

3. Implement Governance: Prevent chaos by managing lineage, definitions, and access in your code with dbt, ensuring governed data for AI applications.

4. Ensure Data Quality: Inconsistent data isn't an option. Use dbt's tests to maintain data integrity and model consistency.

5. Provision AI-Ready Data: Transform and clean data with dbt for immediate ML use, ensuring discoverable, aggregated datasets.

6. Centralize Metadata: AI models need context. dbt auto-generates documentation and lineage, making data assets easily discoverable and understandable across your organization.

7. Modern Data Stack: dbt fits seamlessly into the modern data stack, offering the scale and reliability needed for high-velocity data crucial for complex AI systems.

8. Agile Iteration: AI data needs change quickly. dbt's modular, version-controlled setup lets you swiftly adjust, test, and deploy data models to keep up with evolving ML workflows.

9. Foster Data Literacy: With dbt's single source of truth and clear documentation, teams can collaborate on reliable data.

10. Provision Safely & Explainably: dbt provides structured context for AI, ensuring safe, understandable workflows.

----------------------------
Download report here for more on how to use dbt: Try here

Claude Takes Control of Your Browser

Anthropic rolled out Claude in Chrome to all paid subscribers after months of testing browser automation capabilities through a research preview that started with 1,000 Max plan users.

• Claude can now navigate websites, click buttons, fill forms, and complete multi-step workflows directly in Chrome.

• The extension works alongside Claude Code, allowing developers to build in their terminal while Claude tests and debugs in the browser.

• Users can schedule automated tasks like pulling metrics from dashboards, organizing Google Drive files, or logging sales calls to CRM systems.

• The extension runs in a side panel and creates new tabs for tasks while maintaining access to your existing browser sessions and login states.

• Anthropic tested 123 adversarial prompt injection scenarios where malicious websites could trick Claude into harmful actions. Initial tests showed a 23.6% attack success rate in autonomous mode.

• Users control which websites Claude can access and must approve high-risk actions like purchases or data sharing.

The extension requires Google Chrome and doesn't support Arc, Dia browsers, or WSL. Claude blocks certain high-risk categories including financial services sites, and the company advises against using it for sensitive transactions or password management.

Personal Take: Browser automation becomes useful when it handles the tedious parts without requiring perfect accuracy. Claude in Chrome fits tasks where you need speed over precision and where you can review outputs before they matter.

Superhero Tools for AI Image

Nano Banana Pro: An AI image generator and editor that turns text prompts and reference photos into fast, high‑quality visuals using Google’s latest Gemini image models.

GPT Image 1.5: OpenAI’s newest image model that creates detailed, prompt-accurate visuals and edits for use cases like marketing, product shots, and design mockups.

Adobe Firefly: Adobe’s generative AI suite that lets you create and edit images, video, and audio directly inside Creative Cloud tools with commercially safe outputs.

Leonardo: A creative AI platform for generating production-ready images, game assets, and concept art with fine control over styles and workflows.

Imagine Art: A fast AI image generator that supports multiple advanced models (like Flux.1 and Nano Banana) for high-quality text-to-image and image-to-image creation.

Quick AI News Bites

1. OpenAI has officially launched an app store inside ChatGPT where developers can now submit their apps for review and publication. The new App Directory features major integrations including Spotify, DoorDash, Apple Music, Expedia, Zillow, and Canva, allowing users to access these services directly from chat conversations.
   

2. Swedish AI startup Lovable has raised $330 million at a $6.6 billion valuation, more than tripling its worth since July when it was valued at $1.8 billion. The vibe-coding platform, backed by U.S. venture capital firm Accel, claims more than 100,000 new projects are built daily on its platform.
   

3. Google is bringing its Opal tool to the Gemini web app, allowing users to build AI-powered mini apps using simple natural language and visual editing. The integration lets users create custom versions of Gemini called "Super Gems" that can chain together prompts, models, and tools for specific tasks.
   

4. President Trump's administration has launched an interagency review of advanced Nvidia AI chip sales to China that could allow the first shipments of Nvidia's H200 chips, the company's second-most powerful AI processors.
   

5. OpenAI, Google, and Perplexity are rolling out free AI services in India as part of a strategy to attract users and harvest multilingual training data from the world's most populous nation. Daily active users of ChatGPT in India surged 607% year-on-year to 73,000 following the free plan rollouts.

Boost revenue and gain new customers by partnering with us

Reach over 35K AI enthusiasts with your product.

Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!

Got feedback, a story worth toasting, or a wild tech question? I’d love to hear from you; just reply, or find me on X.

Cheers,

— Poonam Soni

X: @CodeByPoonam