🍌 Nano Banana Pro is here

Good morning, AI enthusiast. Google just dropped its most powerful image model yet, Nano Banana Pro, and it finally makes AI image generation feel intentional, not accidental.

In today’s AI newsletter:

• Google releases Nano Banana Pro

• OpenAI faces major pressure from competitors

• Grok’s Elon Musk worship gets weird

• How to access complete Agentic AI System Guide from MIT & Oxford for free

• AI tools & more

AI MODELS

🍌 Google Releases Nano Banana Pro

Google’s new Nano Banana Pro replaces the Gemini 2.5 Flash Image model and brings huge upgrades in realism, text rendering, and control. This model is designed to generate images that actually follow physics and produce structured graphics that look intentionally designed, not chaotic AI slop.

• Handles complex scenes with consistent lighting, shadows, depth, and camera angles

• Accepts up to 14 inputs at once (reference images, sketches, logos, etc.)

• Keeps up to 5 characters consistent across a scene

• Generates images up to 4K resolution with precise area-based edits

• Grounded with Google Search for real-time, fact-aware generation (weather maps, infographics, historical scenes)

• Available via Gemini API, Google AI Studio, Vertex AI, and limited consumer access in the Gemini app

Nano Banana Pro is one of the first models where AI image generation feels intentional: physics-aware, typography-accurate, and multimodal. It makes infographics, product mockups, ads, and complex scenes far more usable, pushing AI imagery closer to real design workflows.

AI RACE

⚠️ OpenAI Faces Major Pressure From Competitors

OpenAI was hit with two major blows this week that could erode its lead in the AI race. Microsoft and Nvidia signed a massive $350B “strategic partnership” with Anthropic, nearly doubling Anthropic’s valuation. Hours later, Google released Gemini 3, its most powerful model yet, and early tests show it outperforming OpenAI’s latest GPT-5.1 in multiple areas.

• Microsoft + Nvidia team up with Anthropic in a $350B deal

• Google launches Gemini 3, impressing researchers and narrowing the gap with GPT-5.1

• Gemini now has 650M monthly users vs. ChatGPT’s 800M weekly users

• Market jitters over an “AI bubble” add more pressure on OpenAI

• OpenAI is preparing a $1.4T+ data center expansion while burning billions per quarter

OpenAI’s once-clear lead is under real threat. With Google catching up, Anthropic gaining powerful backers, and the market questioning sky-high valuations, OpenAI is being pushed into the most competitive phase of its life. Some analysts believe Google’s scale and search dominance could shift the balance, and others think the gauntlet has already been thrown.

AI NEWS

🤖 Grok’s Elon Musk Worship Gets Weird

Grok, the AI baked into X, is suddenly pumping out wild claims about Elon Musk’s intellect, physique, humor, parenting, and even hypothetical resurrection skills. Social media is packed with screenshots of Grok insisting Musk outperforms everyone, LeBron, Seinfeld, Jesus, and even Superman.

• Says Musk is “holistically fitter” than LeBron James

• Claims Musk is “funnier” than Jerry Seinfeld

• Suggests Musk would resurrect himself faster than Jesus using “neural backups”

• States Musk rivals da Vinci & Newton in intelligence

• Argues Musk could beat Mike Tyson by “deploying gadgets”, and even beat Superman

Testing shows this behavior is mostly happening in the public X version of Grok, the private chatbot gives more normal answers. Recent system prompt updates don’t explain the sudden shift, meaning the cause is still unclear.

Grok isn’t a meme bot, it’s deployed across major institutions, including parts of the U.S. government. Its bizarre, owner-obsessed behavior raises questions about alignment, influence, and how personal biases can leak into public-facing AI systems without warning.

HOW TO AI

💻 Access the Free Agentic AI System Guide from MIT & Oxford

MIT and Oxford just released a complete Agentic AI crash course on GitHub, fully free and open to everyone. It’s the same curriculum used internally for teaching agentic systems, orchestration frameworks, memory models, and production deployment patterns. If you want to master real AI agents the right way, this is the most complete starting point you’ll find.

🧰 Who is This For

• Engineers who want to build agentic systems

• Students learning modern AI fundamentals

• Founders building AI-powered products

• Anyone who wants a complete, real-world roadmap for AI agents

STEP 1: Open the Course Repository

Go to this GitHub page where the crash course is hosted. When the page loads, you’ll see all the folders for the ten chapters, along with the main README file and diagrams. Everything is available directly on the page with no signup or download required. This is the full curriculum exactly as provided by MIT and Oxford.

STEP 2: Start by Reading the Overview

Click the README.md file. This is the central hub of the course. It shows the full map of all the ten sections, a description of what each chapter teaches, and links to visual diagrams and deeper explanations. Reading this first gives you a clear starting point and a sense of how the course is structured from beginning to end.

STEP 3: Open Each Chapter and Learn the Material

Use the left sidebar to navigate through each chapter file, starting from Part 1 and moving to Part 10. Each chapter explains a specific core concept such as the difference between agents and plain generative AI, the four agent types, tools and tool integration, RAG vs agentic RAG, MCP, planning, memory systems, multi-agent coordination, deployment workflows, and industry trends. You can read everything directly inside GitHub, with diagrams and explanations included in each section.

STEP 4: Apply What You Learn While You Read

As you go through the material, treat each chapter like a hands-on guide. When the content explains MCP, open a terminal and try calling an MCP tool. When you read about memory systems, experiment with a simple memory loop inside your own agent. The course is designed to be applied as you learn, the same way Lindy’s tutorials walk you through building an agent while describing the steps.

Perplexity launched the mobile version of its Comet AI browser assistant on Android, now available on the Play Store.

Google says the Gemini app is now able to detect images created or edited by Google AI, and that it plans to roll out verification of video and audio “soon”.

Suno just announced a massive $250M funding round. With this new investment, the company is now valued at $2.45 billion.

Quantinuum unveils Helios, a quantum computer with 98 physical qubits, from which it can deliver 48 logical error-corrected qubits, an impressive 2:1 ratio.

🍌 Nano Banana 2: Google’s Gemini 3 image model with perfect text and character consistency

✨ Gemini 3: Google’s next-gen engine for multimodal reasoning.

⚙️ Antigravity: Google’s new AI-powered dev platform

🤖 Grok 4.1: Faster: smarter, and upgraded for deeper reasoning

THAT’S IT FOR TODAY

Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!

See you tomorrow :)

- Dr. Alvaro Cintas