The Unreasonable Effectiveness of Reasoning Distillation: using DeepSeek R1 to beat OpenAI o1 | Latent Space | Podwise