Google goes wild, again... 11 things you missed at I/O

Key Concepts

Gemini 2.5, Gemini Flash, Stitch (UI generator), Jules (async agentic coding agent), Flow (cinematic adventure generator), Google AI Ultra subscription, Gemma 3N, Project Astra (live video feed AI), Android XR glasses, Project Beam (2D to 3D video), VO and Imagen upgrades, CSS carousel primitives.

Google IO 2025: 11 Game-Changing Announcements

1. Google AI Ultra Payment Plan

• Details: A new subscription model required for accessing the full capabilities of Gemini.
• Pricing: $124.99 per month for the first 3 months, then $249 per month. Includes 1 month free and 30 TB of storage for generated content.
• Significance: Highlights the increasing cost of accessing top-tier AI models.

2. Gemini's Dominance

• Performance: Gemini 2.5 surpasses all other reasoning models, including Claude 3. Gemini Flash outperforms Claude 3 in certain benchmarks.
• LM Arena: Gemini is now the best across all LM Arena categories.
• "Trust Me Bro" Benchmarks: Gemini completely crushed its own benchmarks.

3. Gemma 3N Release

• Nature: An open-source model, allowing developers to use it for commercial purposes.
• Performance: Approaching the performance of Claude Sonnet 3.7.
• Significance: Provides developers with a powerful and accessible AI model.

4. Gemini in the Browser (Agent Mode)

• Functionality: Gemini will be integrated into Chrome, enabling it to interact with websites, click buttons, and fill out forms. Formerly Project Mariner.
• Example: Automatically generating opinions and posting comments.
• Impact: Automates web-based tasks and interactions.

5. The Rise of AI Agents

• Context: 2025 is highlighted as the year of AI agents.
• Examples: OpenAI's Codex agent, GitHub Copilot's agent mode, and Gemini's agent mode.
• Integration: Agents will be used in tools like Stitch and Jules.

6. Model Context Protocol Support

• Significance: Google supports the Model Context Protocol, facilitating easier integration for developers.
• Contrast: This is presented as unusual for Google, which typically creates proprietary architectures.

7. Project Astra: Live Video Feed AI

• Functionality: Uses low-latency AI to describe the environment from a live video feed.
• Example: Identifying edible mushrooms in the wild.
• Potential: Integration into Android XR glasses.

8. Android XR Glasses

• Reality: Confirmed to be in development.
• Purpose: To compete with products like Meta Smart Ray-Bans.
• Future Integration: Will incorporate Project Astra technology.

9. Project Beam: 2D to 3D Video Conversion

• Functionality: Transforms 2D video into a 3D experience using AI.
• Application: Enhancing realism in virtual meetings.

10. VO and Imagen Upgrades & Flow

• VO and Imagen: Google's video and image generators receive significant upgrades, including the ability to generate video with sound.
• Flow: A new tool that combines these capabilities, targeted at Hollywood studios.
• Potential: Could be used to generate content for platforms like Netflix.

11. CSS Carousel Primitives

• Innovation: New CSS primitives simplify the creation of carousels.
• Benefit: Reduces the amount of code required (from 10,000 lines of jQuery to a few hundred lines of CSS).

Conclusion

Google IO 2025 showcased a significant leap in AI capabilities, with Gemini leading the charge. The announcements ranged from powerful new models and agent integrations to innovative tools for content creation and user interface design. While the cost of accessing top-tier AI is increasing, the potential applications across various industries are vast. The introduction of new CSS primitives also highlights ongoing improvements in web development technologies.