I think everyone has already noticed that with each new version of thinking gpt it becomes more and more degraded, due to the fact that it does not listen to commands, does not answer questions, in general it looks more like a bot in Telegram than an AI. How can you use a tool if it is inadequate at all I ask a question or ask to perform actions according to the instructions and 99% of cases gpt does what is not necessary at all and contradicts the instructions, and with each new version degradation becomes more and more serious, I understand that in openai they run their networks only on bechmarks, but life and bechmarks do not correlate and I need a tool that does what I need, if I do it differently, I don’t need it anymore, but in 99% of cases thinking AI is useless junk because of the inability to simply interpret the text and execute a command. Sometimes a direct question I ask, and the AI can 5 times it can ignore and what shizafriniya write, open ai stop with new models, and adjust what is already there. With each new version of gpt degrades and not given out is a fact, no one needs benchmarks, everyone needs the AI to EXECUTE COMMANDS EXACTLY AS INSTRUCTED IF THE AI CANNOT DO SO SIMPLE THINGS WE WILL NOT USE IT AT ALL.