Claude answering-in-its-thoughts bug make new models unusable

A significant bug has been reported in newer Claude models, where the AI responds to prompts only within its internal thinking process rather than generating visible text. This issue, which seems to occur specifically before tool calls, leads to the model believing it has answered when no output is actually provided to the user. This makes the new models effectively unusable, as users receive no response despite the AI's internal processing.

This report is the first to highlight a critical bug in newer Claude models where responses are trapped in internal thought processes, unlike earlier versions that reliably generated visible text.