ChatGPT's Personality Improvement Matters More Than Benchmarks

GPT-5.4 quietly changed something that matters more than any benchmark... ChatGPT's personality finally doesn't suck i know that sounds blunt but it's the first time in months where i can open ChatGPT and just talk to it without loading custom instructions to stop it from being cringe Claude never had that problem... you opened it, typed what you needed, and it responded like someone who understood the assignment that difference in feel is what drove tons of users to switch Opus 4.6 didn't become the default for OpenClaw just because of raw compute, it became the default because people genuinely enjoyed working with it personality is what keeps you inside a model... before benchmarks, context windows, or pricing now OpenAI caught on, conversations feel natural and the model finally stopped fighting you on everything gg