CHINA'S DEEPSEEK PREPS AI AGENT FOR END-2025 TO RIVAL OPENAI

Summary from the TLDR newsletter

DeepSeek is developing an AI model designed to carry out multi-step actions on users' behalf with minimal direction. The system is meant to learn and improve based on its prior actions. Since its R1 model, the Chinese upstart has only put out minor upgrades. While local media attributes R2's delay to DeepSeek's determination to get it right, others have speculated about various glitches in training or development.