Using the AI Assistant (DeepSeek R1 in this case), sometimes the output is put behind two "Thinking" layers, with the incorrect one hiding the entire response. This makes it hard to follow.
- Use AI model with reasoning
- After some time, output is hidden behind another "Thinking" layer
I am unable to force the error, but it happens often enough for me, especially on longer threads.
The pictures will show the problem in detail:
Hidden: Here, the output was completely hidden behind a "Thinking" layer:

Upon manually opening this "Thinking" layer, we can see the actual "Thinking" layer (here, hidden) and the final response:

We can also open the actual "Thinking" layer to reveal the AI's reasoning:

No additional "Thinking" layer.
Here we see normal behavior. Reasoning hidden behind "Thinking".
