Good Morning! Google just unveiled the Fitbit Air, a screenless, "invisible" wearable designed to take the fight directly to Whoop by trading screens for high-level AI coaching. Plus I’ll show you how to generate 10+ images in a single Gemini response
OpenAI Overhauls the Realtime API with GPT-5 Reasoning
Google Launches the Fitbit Air
Anthropic’s Mythos Rewrites the Rules of Cyber Defense
How to Generate 10+ Images in a Single Gemini Response
4 new AI tools worth trying
AI MODELS
OpenAI just dropped a suite of voice intelligence features that move far beyond simple chatbots. These tools are designed to listen, reason, and act in real-time, bridging the gap between human conversation and autonomous action.
GPT-Realtime-2: A realistic vocal model now powered by GPT-5-class reasoning, allowing it to handle complex, high-stakes requests during live voice interactions.
GPT-Realtime-Translate: Built for "keeping pace" with human speakers, offering conversational translation across 70+ input and 13 output languages.
GPT-Realtime-Whisper: A live transcription engine that converts speech to text instantly as the conversation unfolds.
Voice that Acts: Unlike previous versions, these models are designed to "take action" as a conversation unfolds, moving the technology from simple call-and-response to functional automation.
This isn't just about a smarter voice assistant; it’s about the "Action Layer" of AI finally getting a voice. By embedding GPT-5 reasoning into a low-latency voice API, OpenAI is enabling a new generation of autonomous agents that can manage customer support, conduct live education, and handle complex logistics without a human ever needing to touch a keyboard.
AI NEWS
The screenless wearable Google teased earlier this year is finally here. The Fitbit Air is a tiny, passive tracker that ditches the display and buttons to focus entirely on comfort and 24/7 data collection, powered by a Gemini-infused "Health Coach."
It’s incredibly small, 25% smaller than the Fitbit Luxe and weighing just 5.2 grams without the band, designed to be forgotten whether you're at the gym or in bed.
Features a screenless, button-free design that focuses on passive vitals: heart rate, HRV, AFib detection, and sleep stage tracking with an improved recovery-focused Sleep Score.
The centerpiece is the Gemini-powered Google Health Coach, which offers a conversational interface to give you personalized suggestions, dynamic fitness plans, and insights that adapt to your real-time performance.
Despite its size, it packs a week of battery life and a bi-directional magnetic charger that gets you a full day of juice in just five minutes.
Google is betting that the future of fitness isn't another screen on your wrist, but an "invisible" assistant that lives in your health data. By undercutting Whoop on price and integrating Gemini directly into the coaching experience, Google is trying to turn the wearable from a data-gathering gadget into a proactive, autonomous health advisor. It’s the ultimate "vibe" device for people who want the insights of a smartwatch without the digital distractions.
AI NEWS
Mozilla researchers just revealed that Anthropic’s Mythos model is fundamentally shifting the balance of power in cybersecurity. By using agentic systems to hunt for flaws, Firefox has cleared more "technical debt" in a single month than it usually does in a year.
Mozilla reported that Mythos discovered a wealth of high-severity bugs, including a 15-year-old error in how the browser parses HTML elements.
The impact on productivity is massive: Firefox shipped 423 bug fixes in April 2026, a 1,200% increase over the 31 fixes shipped in April 2025.
Mythos is successfully uncovering "sandbox" vulnerabilities, complex, multi-step flaws that usually command Mozilla’s top $20,000 bug bounty, at a volume human researchers have never matched.
The breakthrough is attributed to the "agentic" nature of the model, which can assess its own work and filter out the false positives that made previous AI security tools a headache for engineers.
We are entering a "Great Patching" era. Anthropic CEO Dario Amodei argues that because there are a finite number of bugs in any given piece of software, frontier models could eventually "finish" the job of securing our most critical tools. If defenders can use models like Mythos to clear out a decade's worth of vulnerabilities before bad actors can find them, we might finally see a world where software is secure by default rather than a constant game of whack-a-mole.
HOW TO AI
🗂️ How to Generate 10+ Images in a Single Gemini Response
In this tutorial, you will learn the updated "Reinforcement Prompting" technique to bypass multi-panel image errors and generate 12 or more individual, high-fidelity images in a single response using the Gemini Nano Banana 2 model.
🧰 Who is This For
People creating social media content at scale
Marketers running ad creatives and A/B tests
Designers exploring multiple visual directions fast
Content creators needing bulk thumbnails or posts
STEP 1: Upload Your Reference
To maintain character or style consistency across all your images, start by uploading a reference photo to the Gemini chat. This is particularly useful for "Vibe Coding" projects or brand-specific content where the subject needs to look identical in every generated scene.
STEP 2: Use the "Reinforcement Prompt"
The key to avoiding "multi-panel" single images is using a prompt that forces individual calls to the generation engine.
Copy and adapt this structure:
"Generate each item in the list below as a separate individual image. Call the image generation model based on the number of items below. Do not stop generating until all requested images are generated. Each image must follow these exact specifications: [Global Settings: Cinematic look, 16:9 aspect ratio, film grain, etc.]"
STEP 3: List Your Specific Scenes
Below your global instructions, provide a numbered list of the specific scenes you want.
While the previous limit was 10, the Nano Banana 2 model can now handle at least 12 images in one go.
For example:
Character walking through a neon-lit cyberpunk city.
Character sitting in a high-tech lab drinking coffee. ...and so on, up to 12.
STEP 4: Manage the Render & Download
Once you send the prompt, the generation might take a few minutes. If the UI seems to hang or images aren't appearing, refresh or duplicate the page; the images are often generated on the backend even if the browser struggles to render them all at once.
Note that while generation is batched, you currently still need to click the download button on each image manually to save them to your device.
Want to keep unlock all the premium content?
Become a paying subscriber to get access to all other subscriber-only content.
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
AirPods with cameras reached an advanced testing stage; the cameras will feed data to Siri to help answer questions, rather than take photos or video.
OpenAI launches Trusted Contact, an optional safety feature for ChatGPT that lets adult users assign an emergency contact for mental health and safety concerns.
OpenAI is out GPT-5.5-Cyber, a security-focused variant of the model, in a limited preview capacity to vetted cybersecurity teams.
Spotify launches Save to Spotify, a command-line tool that allows AI agents to upload AI-generated audio summaries and personal podcasts to a user's account.
🌎 GPT5.5 Instant: OpenAI’s new default model for ChatGPT.
🎆 ChatGPT Images 2.0: OpenAI’s upgraded image generator
⚙️ Rork Max: AI tool to build native iOS apps
💻 Holo 3: Open AI agent that can use computers like a human
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