No more context rot?

there has been a huge improvement in long context retrieval in less than a year, but why is no one talking about this? i feel like this deserves a lot more attention as context rot was a huge problem with llms and now it's almost solved(?)