By skyforbes Has anyone noticed that Deep Research's output report starts off with reasonably long sections, but that the sections get thinner and thinner as the report continues? It's as if it starts to realize that it is running out of its max output tokens limit and starts to compress in order to complete the entire document.
I'm just wondering if others see this too. I didn't see this in 2.5, or at least, it wasn't bad enough for me to notice.