😴 AI predicts diseases from sleep

Good Morning! A new Nature-published model just turned a single night of sleep into a powerful health predictor, far beyond sleep stages or apnea. Plus, I'll show you how to generate cinematic AI videos with no limits.

In today’s AI newsletter:

• How to Generate High Quality Cinematic AI Videos With no Limits

• AI Predicts Disease From One Night of Sleep

• Amazon Brings Alexa+ to the Browser

• Liquid AI Unveils LFM2.5 for Edge Devices

• 4 new AI tools worth trying

AI MODEL

🛌 AI Predicts Disease From One Night of Sleep

Researchers just introduced SleepFM, a foundation model trained on massive sleep data that can predict long-term health risks from just one night of recordings.

• Trained on 585,000 hours of polysomnography data from 65,000 people (brain, heart, muscle, breathing signals)

• A single night produces a general sleep embedding that predicts 130 diseases with strong accuracy

• High-risk prediction scores include mortality (0.84) and dementia (0.85) on Harrell’s C-index

• Designed to be robust to missing signals, unlike older models that break across clinics

Sleep is turning into a biometric goldmine. Instead of narrowly detecting apnea or sleep stages, foundation models like SleepFM treat sleep as a holistic health signal. If validated clinically, one night of data could become an early-warning system for chronic disease, shifting medicine from reactive care to long-term risk prediction.

AI NEWS

🗣️ Amazon Brings Alexa+ to the Browser

Amazon has launched Alexa.com, letting early-access users chat with Alexa+ directly from a web browser for the first time.

• Alexa+ now works on the web, not just Echo devices or mobile apps

• Users can ask questions, plan trips, create content, and control smart home devices in-chat

• Access is limited to Alexa+ early users, via waitlist or newer hardware

• The move aligns Alexa with ChatGPT, Gemini, Claude, and Perplexity as browser-first AI assistants

This is Amazon admitting that voice-only isn’t enough anymore. By bringing Alexa into the browser, Amazon turns its assistant into a full conversational AI, not just a smart speaker feature. The real battle now isn’t devices, it’s who owns the default AI tab in your browser.

AI MODELS

📱 Liquid AI Unveils LFM2.5 for Edge Devices

Liquid AI introduced LFM2.5, an open-weight multimodal model family built to run efficiently on phones, laptops, and IoT hardware, no cloud required.

• 1.2B–1.6B parameter models optimized for low-memory, CPU/NPU inference with strong text, vision, audio, and agentic performance

• Audio model stands out: native waveform TTS + ASR, ultra-low latency, and real-time speech-to-speech that beats much larger models

• Specialized variants include a top-tier Japanese model and a vision-language model supporting multi-image + multilingual reasoning

• Edge-first deployment via llama.cpp, MLX, ONNX, vLLM, and mobile SDKs, with AMD/Qualcomm NPU optimizations

LFM2.5 shows where AI is headed, smaller, faster, private, and living directly on your device. If models like this keep improving, “always-on AI” won’t mean servers in the cloud, it’ll mean intelligence baked into everything you own.

HOW TO AI

🗂️ How to Generate Cinematic AI Videos With No Limits

In this tutorial, you’ll learn how to create smooth, high-quality cinematic videos using AI, completely free, with no usage caps.

🧰 Who is This For

• Creators who want cinematic videos without expensive tools

• Social media builders making reels, shorts, and visual stories

• Filmmakers and designers experimenting with AI visuals

• Anyone frustrated with video tools that hit limits too fast

STEP 1: Access the Tool

Open your browser and go to the Skyreels platform. Once the homepage loads, you’ll land on a clean dashboard showing multiple AI tools.

Look for the “AI Video” section. Inside that card, click the black “Create” button. This opens the main video generation workspace where all videos are created.

STEP 2: Upload an Image or Use a Text Prompt

Inside the workspace, you can generate videos from either text or images.

At the bottom menu, make sure the “Image or Text” tab is selected.

If you want to create a video from an image, click the image (paperclip) icon and upload your file. After uploading, a crop window will appear. Choose your aspect ratio:

• 16:9 for cinematic scenes

• 9:16 for Reels, Shorts, or TikToks

Click the blue “Crop” button to continue.

You can also skip images entirely and just describe the scene using text.

STEP 3: Configure Motion, Mood, and Style

In the text box labeled “Describe the content you want to create,” explain how you want the video to move and feel. For example, you can describe camera motion, atmosphere, or emotional tone.

You can turn on Auto Mode to let the AI handle all settings, or manually adjust motion intensity if you want more control.

For advanced control, explore additional tabs:

• Effect for visual styles

• Subject Reference to stay consistent

• Extend to create infinite-length videos

These options are what allow the tool to go beyond short clips.

STEP 4: Generate, Review, and Extend Your Video

When everything is ready, click the blue “+ Create” button at the bottom right.

Your video will first show as “In Queue,” then switch to “Being Generated.” Once finished, click the thumbnail to open the video player and review the result.

You’ll notice smooth motion, cinematic pacing, and emotionally synced sound effects baked directly into the video.

If you want changes, use the Edit button. If you want the scene to continue, click Extend and keep the video going, with no limits on length or generations.

xAI raised a $20B Series E, exceeding its $15B targeted round size, with participation from Valor, Nvidia, and others, and says Grok 5 is currently in training.

Meta unveils teleprompter and EMG handwriting features for Meta Ray-Ban Display and expands Pedestrian Navigation for the glasses to four new US cities

Brands complain Amazon's “Buy for Me” AI-powered shopping tool features their products on Amazon without permission and shows descriptions with errors.

Chinese officials are reviewing Meta's purchase of Manus for possible technology export control violations.

🌌 Grok Imagine: xAI’s tool for generating images and videos

⚙️ Claude Code: Anthropic’s AI coding tool that understands deep context

📑 Qwen Image Layered: An image AI that generates images in editable layers

🎨GPT Image 1.5: OpenAI’s latest, faster and more controllable image model

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