Skip to content
Better HN
[broken] Cursor trains a 70B model that achieves >1000 token/s using... | Better HN