story

SubQ: a sub-quadratic LLM with 12M-token context (opens in new tab)

subq.ai

4 pointsmitchwainer6d ago2 comments

2 comments

This is pretty remarkable. We've spent a lot of time finding workarounds for LLMs reading long docs. Now that's gone.

optimizing AI in general. How cool is that?

j / k navigate · click thread line to collapse

This is pretty remarkable. We've spent a lot of time finding workarounds for LLMs reading long docs. Now that's gone.

optimizing AI in general. How cool is that?

j / k navigate · click thread line to collapse