There should be 128,000 token limit for the model I use, before the start of the conversation starts to get cut short to make room for more context. Meaning it shouldn't really be forgetting how to write or place certain characters in that thread. Its not just for writing but rn it's affecting my writing the most.
I used to be able to use a thread in half of a day now I have to use 3 sometimes 4. I suspected chat threads were getting shorter a while back. But if I dont jump to a new thread it really just forgets everything and starts making stuff up. Just to be clear I used to get around 95% of a full chat before it started to struggle now i dont think im even getting past 70% full before having to switch.