Y Hacker News new | ask | show | jobs

user: smaddrellmander created: 2026-04-21 karma: 75 idlemachines.co.uk submissions: The annotated PyTorch training loop 39 points | 9 comments Mae vs. MSE: more than just the mean vs. median debate 1 points | 0 comments DiffusionGemma: Discrete diffusion in a large language model 3 points | 1 comments Heaven knows I'm perplexed now 2 points | 0 comments Reading MAI's efficiency gain. How to pick architectures like serious people 9 points | 0 comments MAI-Thinking-1: Building a Hill-Climbing Machine [pdf] 3 points | 0 comments Are contrastive losses just cross entropy all along? 2 points | 0 comments Every token, everywhere, all at once 2 points | 0 comments 0 points | 0 comments The cut in the Mixture of Experts compute graph 1 points | 0 comments DeepSeek V4 from the Inside 2 points | 0 comments Softmax, can you derive the Jacobian? And should you care? 134 points | 47 comments Gemma 4 is not your standard transformer 2 points | 0 comments