Alex K

on the atlas — 25

• Building a robotics research setup that lives next to my desk – dfdx labs1 savers
• Durability & Consistency - ZeroFS Documentation1 savers
• Crawling a billion web pages in just over 24 hours1 savers
• Orbit - Ultra-efficient RL Pipeline1 savers
• Mirage Persistent Kernel: A Compiler and Runtime for Mega-Kernelizing Tensor Programs1 savers
• Kimi-K2.5 Inference Benchmark - Luminal1 savers
• Compiling Models to Megakernels - by Luminal and Joe Fioti1 savers
• Real-time LLM Inference on Standard Datacenter GPUs (3,000 tokens/s per request)1 savers
• Fara1.5 - A family of frontier computer use agent models - Microsoft Research1 savers
• Nemotron_Diffusion_Tech_Report_v1.pdf1 savers
• You Only Need Minimal RLVR Training: Extrapolating LLMs via Rank-1 Trajectories1 savers
• SkyNomad: On Using Multi-Region Spot Instances to Minimize AI Batch Job Cost1 savers
• [2605.20873] PlanningBench: Generating Scalable and Verifiable Planning Data for Evaluating and Training Large Language Models1 savers
• PopuLoRA: Co-Evolving LLM Populations for Reasoning Self-⁠Play1 savers
• Learnings from 100K Lines of Rust with AI | Cheng Huang’s corner1 savers
• What Makes Low-Bit Quantization-Aware Training Work for Reasoning LLMs? A Systematic Study1 savers
• Orthrus: Memory-Efficient Parallel Token Generation via Dual-View Diffusion1 savers
• The Matrix Calculus You Need For Deep Learning1 savers
• You don’t need to work on hard problems47 savers
• Building a web search engine from scratch in two months with 3 billion neural embeddings14 savers
• The case against conversational interfaces « julian.digital12 savers
• Precise Manipulation with Efficient Online RL5 savers
• How to engineer luck - by George Mack - High Agency4 savers
• An Interview with MatX CEO Reiner Pope About LLM Chips2 savers
• The Man Behind AlphaGo Thinks AI Is Taking the Wrong Path | WIRED2 savers