tweetindex

Alex Corrino @AlexCorrino · 11 May 2026

@AlexCorrino Great post One thing I'd highlight is as KV cache scales its getting offloaded to SSDs NVIDIA has guided this path with CMX. All the context can't be held in HBM alone, especially as we scale from 1 to many agents
5 755Views
24Likes
1Reposts
1Replies
0Quotes
7Bookmarks

Is that a lot?

1.61×vs under 10K median3 570 views is typical
62.01%Reachviews ÷ followers
0.45%Engagement rateof viewers reacted

Compare with the benchmark table →

Open on X →