OpenAI Expands GPT-Rosalind with Life Sciences Capabilities

Researchers from UIUC, UC Berkeley, and Chroma released Harness-1, a 20-billion parameter open-source search agent that scores 73% on information recall benchmarks, outperforming GPT-5.4 (70.9%) and other proprietary models. The model is available under Apache 2.0 license on Hugging Face. Harness-1 achieves its performance by offloading search session management to a structured software environment rather than relying on expanded context windows, suggesting that model efficiency matters more than raw parameter size for autonomous retrieval tasks.

OpenAI has launched the Economic Research Exchange, a platform designed to study artificial intelligence's effects on employment, productivity, and broader economic outcomes. The initiative opens applications for selected research projects that will examine AI's economic impact. The program represents a structured effort to generate empirical evidence on how AI deployment affects labor markets and economic performance.

Andy Konwinski, billionaire co-founder of Databricks and Perplexity AI, is advocating for AI researchers to remain in academia and publish openly rather than joining Big Tech companies. His pitch comes as frontier AI firms including OpenAI, Anthropic, and Google have reduced public disclosure of training details, model architecture, and computational resources. Konwinski argues that open research is essential for democratic and societal reasons, citing a 2017 Google paper that became foundational to today's most popular AI models.

NVIDIA announced physical AI agent skills at CVPR designed to streamline workflows for autonomous vehicle, robotics, and vision AI research. The tools address fragmentation across separate development stages, from scene reconstruction to policy training and evaluation. NVIDIA also released Cosmos 3, an open foundation model for physical AI, and Alpamayo 2 Super, a 32-billion-parameter driving model.