Blog

Technical insights on CUDA, GPU optimization, and AI-powered coding

LLMs Can Now Write GPU Kernels That Beat torch.compile

LATEST JAN 20, 2026
We're now seeing multi-agent systems that take your PyTorch code and produce CUDA or Triton kernels with 2x to 14x speedups over torch.compile(mode='max-autotune-no-cudagraphs'). Not on toy benchmarks. On real models like Llama-3.1-8B, Whisper, and Stable Diffusion.
Jaber Jaber | 14 min read

The CUDA Development Workflow Is Broken

NOV 13, 2025 | 11 min read

The Complete Guide to Fixing CUDA Installation Issues

OCT 13, 2025 | 12 min read

How to Break the Scaling Wall

AUG 21, 2025 | 8 min read