Researchers at Stanford have introduced SleepFM, a new AI model capable of predicting over 130 health conditions, including dementia, heart attacks, and Parkinson’s disease, by analyzing a single night of sleep. The model was trained on 600,000 hours of sleep data from 65,000 participants, tracking brain waves, heart activity, breathing, and muscle signals.
When body signals were misaligned, such as deep sleep paired with elevated heart rate, the model flagged it as a warning for future illnesses. The team linked 25 years of medical records from the Stanford Sleep Clinic with sleep data, testing predictions across over 1,000 disease categories. SleepFM predicted Parkinson’s disease with 89% accuracy, dementia with 85%, heart attacks with 81%, and overall mortality risk at 84%.
This model demonstrates how sleep data can be valuable for early disease detection. As wearable devices become more sophisticated, predictive health monitoring could move directly from the lab to users’ wrists.
In Brief: Tech World Highlights
- OpenAI Chief Product Officer Fidji Simo outlined the 2026 product roadmap in a new blog, aiming to transform ChatGPT into a proactive “personal super-assistant.”
- Abu Dhabi’s TII released Falcon H1R 7B, a small hybrid reasoning model that outperforms competitors up to seven times its size in math and programming, with double the inference speed.
- Microsoft rebranded its Office 365 productivity suite as the “Microsoft 365 Copilot app,” using the same name as its AI assistant.
- Nvidia unveiled the Rubin platform at CES 2026, combining six new chips into a single AI supercomputer and offering five times the training compute compared to the Blackwell line.
- Liquid AI released LFM 2.5, a new family of SOTA open-weight models for on-device AI, covering text, image, and audio, outperforming competitors of similar size in benchmarks.
AI Trending Tools:
- Alexa.com: Amazon’s AI experience across voice, mobile, and web platforms.
- Falcon H1R: TII’s hybrid reasoning model with strong capabilities in math and programming.
- Granola: Enhances meeting notes with transcription directly from Mac audio recordings.
