ParentFull threadmachiaweliczny·From my experience these large context are just techinical feasability but there needs to be a lot better internal benchamarks to claim it really works. I've tested on real world task and it all fails so far.View on HN