Comment by hmottestad

I’m curious about if the model has gotten more consistent throughout the full context window? It’s something that OpenAI touted in the release, and I’m curious if it will make a difference for long running tasks or big code reviews.