CoreWeave Launches Self-Improving AI Agents Platform

CoreWeave has unveiled a platform enabling enterprise AI agents to learn and improve autonomously using real-world data, bypassing the traditional slow cycle of testing and fine-tuning. The offering uses serverless reinforcement learning to post-train and fine-tune large language models for reliability, reducing costs by over 40% and accelerating training by 1.4 times. The platform separates training and inference instances so iteration cycles do not compete, compressing hours of training into seconds. The launch comes as enterprise AI agent adoption surges, with 57% of companies now running agents in production, according to LangChain's 2026 State of Agent Engineering report.