The Most INSANE AI News This Week! 🤯

Key Concepts

• Gemini Canvas: Google's new feature for coding and design.
• Notebook LM: Google's research tool with mind mapping and audio overview features.
• Ernie 4.5 & X1: Chinese AI models comparable to GPT 4.5 at a fraction of the cost.
• Claude with Web Search: Enhanced Claude model with real-time web access.
• Mistral Small 3.1: Open-source, efficient AI model outperforming Google's Gemma 3.
• mCP Agents: Simplified connection to applications via Claude 3 Sonnet in Visual Studio Code.
• Roblox Cube: AI system generating 3D models from text/voice prompts.
• Peer's Object Manipulation: Precise video editing feature.
• Gemini Drawing: AI-powered sketch enhancement.
• AI-powered Game Creation: Generating functional games with simple prompts.
• Podcast Generation: Converting documents into conversational audio.

Gemini Canvas: Superior Coding Capabilities

• Google launched Gemini Canvas, a feature similar to ChatGPT and Claude but with code preview.
• Gemini Canvas excels in coding, particularly front-end development.
• Example: When prompted to create a landing page for an AI news website, Gemini Canvas produced a functional, modern design, surpassing ChatGPT, Claude, and Grok.
• Gemini Canvas is currently free to use.

Notebook LM Updates: Mind Maps and Audio Overviews

• Notebook LM now features mind maps for visualizing connections between ideas from multiple sources.
• This feature is unique among major LLM platforms (GPT, Gemini, Claude).
• Application: Useful for research and organizing complex information.
• Notebook LM also offers an audio overview feature, turning research queries into podcast-like conversations.
• Benefit: Allows for convenient, on-the-go consumption of research summaries.

China's AI Revolution: Low-Cost, High-Performance Models

• Baidu released Ernie 4.5 and Ernie X1 models.
• Ernie 4.5 is comparable to GPT 4.5 in performance.
• Ernie X1 is a reasoning model similar to DeepSeek or OpenAI 01 Mini.
• Key Statistic: Ernie 4.5 costs approximately 55 cents per million input tokens, compared to OpenAI's $40.
• X1 is half the price of DeepSeek's R1.
• These models are difficult to access outside of China due to phone number requirements.

Claude with Web Search: Enhanced Functionality

• Claude now has web search capabilities, overcoming its previous limitation to its training cut-off date.
• Claude performs multiple searches for comprehensive results.
• Example: When tasked with creating a landing page based on search results, Claude created a functional website, while ChatGPT provided only text.
• Claude with web search is now arguably better than ChatGPT for most tasks, especially writing and coding.

Mistral Small 3.1: Powerful, Efficient, and Open Source

• Mistral, a French AI company, released Mistral Small 3.1.
• It outperforms Google's Gemma 3 in benchmarks.
• Features a 128k token context window, multimodal capabilities (text and images), and delivers 150 tokens per second.
• It is multilingual and excels in programming, reasoning, and visual understanding.
• Application: Creating HTML landing pages and cloning websites from screenshots.
• It is open source and can be run locally using Olama.

Simplified AI Agent Building with mCP Agents

• mCP servers allow connection to applications without separate installations.
• Claude 3 Sonnet in Visual Studio Code simplifies mCP agent setup.
• Process: Install Visual Studio Code, add the client extension, plug in the Anthropic API key, and install servers from the mCP Marketplace.
• Example: Creating a Tesla stock analysis website using data from Perplexity's mCP server.
• Claude can design interfaces, browse pages in real-time, and troubleshoot issues.

Roblox Cube: Generative 3D Models

• Roblox introduced Cube, an AI system that generates 3D models from text or voice prompts.
• This is integrated into a platform with 85 million daily active users, including 2.5 million developers.
• Example: Generating a "red buggy with knobby tires" or a "green crystal fantasy sword with gold accents."
• This tool lowers the barrier to entry for 3D content creation.

Peer's Object Manipulation: Precise Video Editing

• Peer is rolling out a feature to manipulate specific objects in a video while preserving the rest of the scene.
• This feature is currently available for creative partners.

Gemini Drawing: Sketch Enhancement

• Gemini added a drawing feature that transforms basic sketches into detailed artwork.
• Example: Turning a simple stick figure into a fully detailed character.
• This is based on Gemini 2.0's native image generation update.

AI-Powered Game Creation: No Coding Required

• Gemini Canvas can create interactive games from simple prompts.
• Example: Creating a functional Pong game with physics, sounds, and scoring from a single prompt.
• This demonstrates the decreasing barriers to creation.

Podcast Generation: Content Repurposing

• Gemini can turn any document into a podcast-like conversation.
• Process: Upload a document, and Gemini generates a conversational audio overview.
• Example: Uploading a book on SEO link building and generating a 6-minute audio summary.
• This allows for easy content repurposing and distribution on platforms like SoundCloud.

Conclusion: A New Era of Accessible Creation

• The pace of AI innovation is accelerating, with groundbreaking features released weekly.
• These tools are becoming more accessible, often available for free or at low cost.
• The limiting factor is now imagination, not technical ability or resources.
• To stay ahead, test new features, compare AI models, and integrate AI into workflows.
• The AI Profit Boardroom and free SEO strategy sessions are available for business growth.