LW - Scaling of AI training runs will slow down after GPT-5 by Maxime Riché | The Nonlinear Library | Podwise