Kung-Fu Robot, Gemini 3.0 Leak: AI Never Stops!

👋 Dear Dancing Queens and Super Troupers,

Just as we were recovering from Spongebob on acid (thanks, Sora 2), the world of AI is accelerating once again. This week, the pace is set by the giants: Google updated Veo/Flow so your videos now have pro-level audio and edits. The icing on the cake is that not only did a leak hint at the date, but CEO Sundar Pichai officially confirmed that Gemini 3.0 is coming by the end of the year.

Meanwhile, Microsoft is making every Windows 11 PC an "AI PC" with the widespread arrival of Copilot Vision and Voice. Our PC has officially become a partner that talks to us and watches what we do. A little spooky, but tremendously practical.

On the hardware front, Apple responded quietly but powerfully by unveiling the M5 chip, tailored for AI, ensuring that even their machines will have their own cutting-edge brain.

Finally, on the competition side, Anthropic is hitting hard with Claude Haiku 4.5, an ultra-fast and three times cheaper model that rivals the best. The race for power is also a race for economic performance.

And if you need a reality check, watch the video of the Unitree G1 robot mastering Kung-Fu better than you can walk.

Here’s this week’s lineup :

👉 Google unveils Veo 3.1 and adds audio to your clips 🎬

👉 Gemini 3.0 officially confirmed: Google accelerates its timeline 🚀

👉 Microsoft transforms every PC into a vocal and visual partner 🗣️

👉 Anthropic launches Haiku 4.5: cheaper, faster, and better than GPT-5? 💸

👉 Apple's M5 chip puts AI on steroids 🍎

If we've forwarded this letter to you, subscribe by clicking on this link !

🔥 If you have 15 minutes

1️⃣ Veo 3.1: Google Puts its AI Camera into Director Mode

The summary :  Google introduces Veo 3.1, an improved version of its video generation engine, powering the Flow video creator. The package promises native audio, smoother transitions, enhanced narrative control, and features like Ingredients to Video and Frames to Video. The whole process becomes more precise for creators.

Details :

• Full-Fledged Audio : With Veo 3.1, generated videos now integrate dialogues, soundscapes, and sound effects naturally.

• Guided Transitions : Thanks to the Frames to Video feature, the user provides a starting and an ending image. Flow creates the transition between the two with visual and sonic coherence.

• Extended Scene : Generating a continuous segment, the Extend/Scene extension function allows a video to be prolonged from its last moment.

• Element Insertion and Removal : The Flow tool allows users to add or remove objects/characters in a scene, while reconstructing the background to make it look natural.

• Available for Developers and Businesses : Veo 3.1 is now accessible via the Gemini API, Google AI Studio, and Vertex AI.

Why it's important :  Because this evolution marks a key turning point: AI video generation is no longer content with animated still images. It expands to audio, narration, and advanced personalization.

For creators, it's a powerful toolkit. For the general public and the industry, it's a step toward "ready-to-use" AI videos that are potentially indiscernible from traditional formats, raising both opportunities and challenges.

2️⃣ Gemini 3.0: An Imminent Announcement? Google Under Pressure After Leak

The summary : CEO Sundar Pichai officially confirmed that Gemini 3.0 will arrive later this year. This announcement highlights an internal leak that indicated Google could unveil its AI on Wednesday, October 22, 2025. Published on X by @chatgpt21, it is already generating significant discussion. Although this timeline remains unconfirmed, expert Mishaal Rahman deems the document "entirely credible." Google, for its part, remains silent.

Details :

• The Date Leak : A recently leaked internal document mentioned the date of October 22, 2025, with the line, "Looking to align with Gemini 3.0 launch moment", suggesting the announcement was imminent.

• Official Confirmation : CEO Sundar Pichai officially tempered the urgency by only confirming that Gemini 3.0 will indeed be launched "later this year."

• Info Credibility : Despite being published by an unverified account (@chatgpt21), expert Mishaal Rahman, a reference in the Android ecosystem, judged the document "entirely credible."

• Forced Acceleration : While Gemini 2.5 has just been deployed, Google seems already keen to avoid being distanced by the competition, according to signals from this leak.

• Strategy Under Pressure : This rumor comes as Google attempts to catch up with the rapid advancements of OpenAI (ChatGPT) and Perplexity.

Why it's important : If Gemini 3.0 is actually unveiled this week, it will be a strong signal from Google showing that AI is no longer a lab luxury, but a war for innovation and leadership. For users as well as professionals, this could redefine how we interact with artificial intelligence in our daily lives rather than just in business.

3️⃣ Claude Haiku 4.5: Anthropic's New Arrow in the Race for Speed

The summary : Just two weeks after the launch of Claude Sonnet 4.5, Anthropic is back. Here is Claude Haiku 4.5, a model that is faster, lighter, and above all, three times cheaper.

Dario Amodei, head of the startup, promises an AI tailored for efficiency: as skilled as Sonnet 4 for coding, but twice as fast and unbeatable on costs. Enough to appeal to busy businesses and developers.

Details :

• Speed Above All : According to Anthropic, Haiku 4.5 surpasses Sonnet 4 in executing on-computer tasks and accelerates apps like Claude for Chrome, which are now smoother than ever.

• Well-Calculated Economy : With a cost reduced by a third compared to Sonnet 4, this model aims for massive adoption in business.

• Sub-Agents to the Rescue : Sonnet plans and Haiku executes: The dependency on a single AI is over. Haiku 4.5 works as a team, with multiple instances able to collaborate to manage a complex project.

• Haiku Becomes the New Standard : This model becomes the default for all free plans of Claude.ai. A strategic choice to popularize the platform while maintaining the same security standards.

• Rave First Reviews : Beta testers, including Gamma and Augment Code, confirm 65% accuracy in text generation and 90% of Sonnet 4.5's capabilities, plus promising integration on GitHub Copilot.

Why it's important :  Anthropic is changing the high-end logic: faster, cheaper, just as strong. By making its premium model accessible to everyone, the startup directly challenges OpenAI and Google. With Haiku 4.5, productivity takes on a Zen accent, but at supersonic speed.

4️⃣ Windows 11: Every PC Becomes an Ultra-Connected AI Companion

The summary: Microsoft propels Windows 11 into a new dimension with Copilot and Agentic. Every PC transforms into an intelligent assistant capable of understanding your instructions, analyzing what you see on the screen, and acting on your behalf. Voice, text, vision: everything is designed for AI to become natural, fast, and secure, directly integrated into your digital daily life.

Details :

• Fluid and Natural Interaction : Users can now converse with their PC via voice or text. "Hey Copilot" instantly triggers actions, and voice usage doubles that of classic text.

• Global Copilot Vision : Available wherever Windows Copilot is offered, Vision analyzes your screen content, guides your projects, and provides personalized recommendations, from troubleshooting to PowerPoint presentations.

• Redesigned Taskbar : The Ask Copilot feature integrates AI directly into your workflow. Ultra-fast results and a dynamic interface for optimized productivity without extra effort.

• Gaming Copilot : In partnership with ASUS, Xbox and Windows unveil the ROG Xbox Ally and Ally X. A simple press allows you to interact with your game assistant for live recommendations and support without leaving the game.

• Security at the Core : Copilot Actions remain under the user's total control, with notifications, approvals, and responsible deployment. Windows 11 maintains its reputation as the most secure system with multiple layers of protection from startup.

• Copilot+ Hardware : Surface, Acer, ASUS, Dell, HP, Lenovo, and Samsung offer AI-optimized PCs with neural processors and extended battery life, starting from $679.99 up to $1,499.99.

Why it's important : Windows 11 is no longer limited to a tool. It becomes an active partner, accessible to everyone, transforming input and navigation into intuitive experiences. Windows 11 PCs with Copilot offer speed, security, and ready-to-use AI, making the computer not only smarter but also profoundly human in its daily use.

5️⃣ Apple's M5 Chip: Supercharged AI Arrives in Your Devices

The summary : Apple unveils the M5 chip, etched in 3 nm and designed for artificial intelligence. Destined for the new 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro, it promises AI performance multiplied by four compared to the M4.

Redesigned GPU and CPU, boosted Neural Engine, extended unified memory: Apple puts its users into digital orbit.

Details :

• 10-Core GPU and Doped-Up Neural Engine : Each GPU core integrates a dedicated Neural Accelerator, quadrupling AI power and improving graphics by up to 45% thanks to 3rd generation ray tracing and dynamic caching.

• Sharpened CPU and Fluid Multitasking : Up to 10 cores, with 4 performance and 6 efficiency, propel the MacBook Pro and iPad Pro with a 15% multitasking gain compared to the M4. Designed to juggle AI applications and pro creations.

• Neural Engine and Local AI : The 16 cores of the Neural Engine accelerate Apple Intelligence functions and image processing on Vision Pro, for AI usage integrated directly on the device.

• Extended Unified Memory : Bandwidth of 153 GB/s (+30%), up to 32 GB configurable, to manage large projects and AI models without going through the cloud.

• Availability: The first devices equipped with the M5 are available for pre-order starting today, launching Apple into a new AI era on MacBook Pro, iPad Pro, and Vision Pro.

Why it's important :  With the M5, Apple transforms its devices into true AI partners. Multitasking, graphics, and neural computations are amplified to offer creatives, gamers, and professionals an ultra-fast tool, ready to support every idea and project. The brand simultaneously maintains its commitment to control and ease of use.

❤️ Tool of the Week: Nano Banana, the AI that styles your photos

Nano Banana is the name of Google's AI image editor that is being integrated into several of their existing products, adding stylization and editing capabilities in a flash.

What is it for?

• Easy Stylization : Transform your images or AI-generated illustrations into various styles (watercolor, anime) via a simple command.

• Boost NotebookLM : Add style options to "Video Overviews" and generate automatic contextual illustrations based on your documents.

• Real-Time Editing : Edit photos taken directly via the Google Search Lens tool or style existing images via prompt.

• Organization : The tool will soon be integrated into Google Photos for editing your stored archives.

How to use it?

👉 To create images from scratch, go to the Gemini app. To edit images in Search and NotebookLM (currently rolling out), use a voice or text command, including via the new ‘Create’ mode in Google Lens.

💙 Video of the Week : The Unitree Robot that Does Kung-Fu