Observed arrival · 2026-10-18
Chris Novick’s local-model performance notes
An engineer’s site collecting dated writing on systems, local models, Rust, and engineering.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Local-model engineers and systems programmers
- Worth noticing
- One listed experiment reports a 2.4× decoding speedup on an AMD Radeon AI PRO R9700.
Field notes
The homepage indexes posts by date and includes short excerpts that give readers a quick sense of each experiment. The excerpts identify hardware, model variants, context length, and—in one case—a fork of vLLM; the speedups are presented as the author's reported results, and the extracted page does not show the benchmark methodology. Other listed writing ranges from an economics argument about inference hardware to a reflection on imposter syndrome.
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