DeepSeek R1 GAVE ITSELF a 2x Speed Boost - Self-Evolving LLM | Matthew Berman | Podwise