Autoregressive Language Model on the 6502 Processor | Hacker News

A Hacker News discussion centered on an autoregressive language model running on a 6502 CPU.

Readers highlighted memory-saving tricks such as banked memory and streaming weights during inference.

Others debated whether hand-written assembly could make the demo smaller and more efficient on systems like the NES with cc65.

The demo shows how modern AI methods can be adapted to extremely limited retro hardware.