Observed arrival · 2026-10-23
Daniel Rushton: Fast kernels, honest numbers
A systems engineer’s personal site about GPU kernels, AI evaluation, and the gap between benchmark scores and trustworthy results.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- GPU systems engineers and AI-kernel evaluators
- Worth noticing
- The planned harness compares model-written CUDA and Triton kernels under standard and hardened checks, then reports the score gap.
Field notes
Rushton says he spent seven years at NVIDIA working on the OptiX ray-tracing SDK before turning his writing toward AI inference kernels and evaluations. The project list gives date ranges for its planned work, including a FlashAttention implementation in Triton and a calibrated LLM judge. Outside computer science, the page describes ongoing audio work, from vacuum tube amplifiers to a small home studio.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
◎NicheUnusually specific use
◉HumanPersonal, local, civic, or handmade
One card from the complete issue