I think a lot of folks who are having a bad experience are using the model in harnesses that aren't yet optimized for GPT-5. For every new model release, there's a lag time between release + when the companies that integrate the model have it working well. Folks rush to switch

→ View original post on X — @mattshumer_