MIT, Mass General Brigham, and Harvard Medical School researchers have created a deep-learning tool to anticipate heart failure progression up to one year in advance. Their PULSE-HF model analyzes ECG data to predict the likelihood of a patient’s heart function worsening, helping prioritize care and potentially improving outcomes for those at greatest risk.

MIT Assistant Professor Matthew Jones is leveraging machine learning and experimental tools to better understand how and when tumors become resistant to treatment. His work focuses on decoding molecular mechanisms, such as extrachromosomal DNA amplification, to predict tumor behavior and ultimately improve patient care.

On July 15, 2026, Mira Murati’s startup, Thinking Machines Lab, launched Inkling—a 975B parameter (41B active) open-weights model under Apache 2.0. This multimodal Mixture-of-Experts (MoE) accepts text, audio, and images to output text/code. With a 1M-token context window, it balances speed and depth via adjustable reasoning and is built for local fine-tuning and agentic workflows.