By skyforbes I love using the Gem feature in Gemini but today after the 3.0 model is out I'm facing a dilemma. When I created my Gems I always put in lots of files as context and before with 2.5 Pro model even though the quality of the responses were consistent I could always rely on the Source Chips to know which part of the response was actually backed with sources and which wasn't. Today when I used the new model the quality was immensely improved and it consistently put out correct answers but the inclusions of Source Chips were 'hit and miss', sometimes they were included and sometimes there were none, does anyone have any idea why this is happening, or if this is a problem on my side (as in instructions, etc..) or it's just a bug during the rollout of a new model.