3 comments

[ 2.7 ms ] story [ 16.4 ms ] thread
> We apparently live in the clown universe, where a simple TUI is driven by React and takes 11ms to lay out a few boxes and monospaced text

im not on twitter much these days, but damn people were not kind to anthropic

yeah Friday on Xitter was a pretty disproportionate response imo!
I'm curious if the 16ms budget is actually driven by the token stream here. Even with the fastest providers right now, you are looking at maybe 100-150 tokens per second. That is orders of magnitude below the 30MB/s terminal parsing limit mentioned in the article.

Unless I am missing something about how they handle the diffs, the bottleneck is surely the inference latency and not the render loop. It seems like a lot of architectural complexity for a data stream that is inherently slow.