Is ChatGPT getting dumber?
retained text
I notice that Terra itself gets dumber on High all the time, but the token consumption increases massively. The 5-hour limit is reached after half an hour, when I had 2 hours up to a few weeks ago. Is the intention to make the old models dumber by using the new Astra and having the tokens gone after 10 minutes? I’m getting less and less into ChatGPT because I can’t work anymore. For example, in an md file, there’s a clear instruction about what the chat read, but it still did the task differently and incorrectly, and that happened with Terra High. What is OpenAI supposed to do? It’s not fun anymore. Working on a manual with Terra High cost over 20%, including pushing to Git, which is no longer normal.
document fetched 2026-09-13 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1wehdue/is_chatgpt_getting_dumber