🔮 AI godfather’s bold prediction

Good Morning! Geoffrey Hinton, the ‘godfather of AI,’ just made a bold call: Google will win the AI race. Coming from the same person who left Google to warn about AI’s dangers, that’s worth paying attention to. Plus, I’ve got a simple tutorial on the new Nano Banana Pro prompt library

In today’s AI newsletter:

A new Adweek report alleged that Google had privately told agencies it would introduce ads inside the Gemini app in 2026. Within hours, Google’s top ads executives publicly called the story inaccurate, insisting the chatbot will remain ad-free for now.

AI chatbots cost massive compute to run, and tech giants are under pressure to monetize them. Google’s pushback shows there’s still no clear business model for AI assistants, but that may not last long as competitors like OpenAI explore ad frameworks of their own.

Anthropic just rolled out a new Slack integration that lets Claude automatically detect coding requests and hand them off to Claude Code, without you pasting context or switching tabs.

Anthropic is turning Slack into a lightweight coding workspace where debugging, repo lookups, and task hand-offs can happen inside team conversations, tightening the loop between discussion and execution.

In a new interview, Geoffrey Hinton praised Google’s recent AI breakthroughs, arguing the company is finally reclaiming the lead it held years ago. With the launch of Gemini 3, the successful Nano Banana Pro image model, and Google’s growing push into custom AI chips, Hinton believes Google may now outrun OpenAI.

If Google truly regains the edge, the balance of power in AI could shift again, from model quality, to hardware control, to the broader ecosystem. With Google now advancing on all three fronts, even OpenAI is reportedly sounding the alarm.

In this tutorial, you’ll learn exactly how to use the massive new Nano Banana Pro prompt library, a collection of over a hundred expertly crafted prompts sourced from top AI engineers across X, Replicate, and the open-source community.

To get started, open the official prompt library on GitHub. Once you’re on the page, scroll through the README until you reach the Table of Contents. This section organizes everything into categories like photoreal portraits, stylized looks, aesthetic experiments, and more advanced creative setups.

As you scroll, you’ll see example images for each prompt along with the full prompt text directly below them. Choose the prompt that matches the look you want. For example, if you want a clean Silicon Valley-style founder photo, you can open “One-Click Business Photo,” and the page will show you both the preview image and the full JSON or text prompt used to generate it.

After choosing a prompt, scroll down to the area labeled “Prompt:”, this is where the actual JSON or raw text block begins. You need to copy this entire block exactly as it appears. Some prompts are long aesthetic text prompts, while others are structured JSON that starts with a curly bracket and ends with one. Make sure you copy everything completely because these prompts include lighting details, camera metadata, texture rules, consistency instructions, and other technical settings that guarantee high-quality output.

If the prompt is JSON-based, it will contain sections such as “subject,” “description,” or camera styling parameters. Your job here is simply to copy the full block with no changes yet. The idea is to bring the entire structure into Nano Banana Pro exactly as the engineers designed it.

Now open Nano Banana Pro and paste the copied prompt into the main text input box. If you’re using the “Thinking” model, this is where the prompt belongs. Once the full prompt is pasted, you can start editing only the parts that describe the subject. Everything else, the lighting, composition, technical metadata, color science, background rules, realism parameters, should remain untouched.

If the prompt is a normal text prompt, locate the one sentence that describes the main subject. Replace that with your subject while keeping the rest of the structure intact. If the prompt is a JSON prompt, look for sections such as “subject” or “description.” Inside those quotes is where you rewrite your details. Make sure you don’t break the JSON formatting, keep all commas, quotes, and brackets in place.

If the chosen prompt is one that requires a reference image, such as a business portrait or photoreal transformation, Nano Banana Pro will allow you to upload your subject’s photo. The prompt will usually include lines like “Keep the facial features exactly consistent with the reference image.” In those cases, upload the photo before generating so the model knows who it’s working with.

Once the subject details are updated, you’re ready to generate the image. Simply run the prompt through Nano Banana Pro. Because these prompts are engineered with extremely refined composition and lighting rules, the model will output a polished, professional-grade result on the first try.

You don’t need trial and error because the library already contains carefully tuned instructions that handle cinematic lighting, skin texture realism, environmental reflections, camera depth, color grading, and more. Your only job is to bring your subject into the template, the prompt handles everything else behind the scenes.

Trump says the US will allow Nvidia to ship H200 chips to “approved customers” in China and elsewhere, and 25% of the chip sales will be paid to the US.

Google says it's working to create two different categories of AI-powered smart glasses: one with screens launching in 2026 and another that's audio-focused.

OpenAI and Instacart launch a grocery shopping experience inside of ChatGPT, letting customers brainstorm meal ideas and check out via OpenAI Instant Checkout.

CEO Sam Altman said OpenAI plans to end “code red” after releasing a model in January 2026 with improved image generation, speed, and personality.

3️⃣ Mistral 3: Mistral’s next-generation of open

🌱 Seedream 4.5: ByteDance’s improved AI for creating and editing images

🎥 Kling 2.6: New AI video model that can handle audio natively

🤖 Claude Opus 4.5: Anthropic’s new flagship model for coding, agents, and real computer use

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