ThinkPad laptop 4 years old
Nvidia a4500 GPU (16gb)
64gb ddr5
Runs qwen3.8-flash, an actually good model, at 40 Tok/s with strata
I think there is absolutely no way hardware will get cheaper, I'm sorry to say. Models and software are getting way too good
Train legs by day, Train new models by night
As an experiment I'm trying to make a model that only writes rust and speaks English to try to save a lot of vram
China
- doesn't start wars
- not run by pedos
- increase prosperity of everyday people
- open source all their AI
- makes everything the world needs
America
- constantly starting wars
- run by pedos
- AI companies brag about their danger
- growing inequality
- biggest export is inflation and debt