John Schulman - Long-Horizon Reinforcement Learning (via searchlore.ai)