I am using deepseek v4 flash from opencode go in vs code,
and the chat shows its process of mulling and thinking to itself, and this is what it is doing.
First: Time spent in the viewer with screenshots, read_page and so on was long → 30 seconds+ → cache expired. Second (just now): After opening it, read_page a few times → this time it may have been under 30 seconds!! (closed it quickly) → cache hit → normal!!!
Oh!!! Then!!!!! on rescan!!!!! is!!!!! the problem!!!!! But!!! the 31-second test!!!!! normal!!!!!!!!!
Ah!!!!!!!!!!!!!!! wait!!!!!!!!!!!!!!!
Could it be!!!!!!!!! that the rescan!!!!!!!!! on an exception!!!!!!!!! saved an empty list!!!!!!!!! into the cache!!!!!!!!! and that cache!!!!!!!!! after 30 seconds!!!!!!!!! was released!!!!!!!!! normal!!!!!!!!!
That is!!!!!!!!! the rescan!!!!!!!!! temporarily!!!!!!!!! failed!!!!!!!!!
What!!!!!!!!! would temporarily!!!!!!!!! fail!!!!!!!!!
What is up with this thing?
Where does it hurt?
When it came out in English I just took it as normal, but looking closely at the messages that occasionally come out in Korean when it loses it, this is what it is doing.
Is it trying to eat up even a few more tokens, or did it learn something wrong from Korean developers..
I was watching it and just burst out, so I am posting it.. ㅋㅋㅋㅋ
