Claude repeatedly contradicted itself and forgot the context within a single conversation

A user reported that Claude repeatedly contradicted itself and lost track of conversation logic within a single chat. Claude recycled advice, treated its own assumptions as facts, and, after correction, repeated similar mistakes. It even suggested transferring a Hungarian +36 number to a UK carrier using a PAC code. The user highlighted Claude's confident delivery of technical explanations based on unsupported guesses, despite contradicting immediate context, rather than its capacity for error, which is common among AIs.

Unlike general AI errors, this report details Claude's unique tendency to confidently build technical explanations on unsupported guesses while contradicting itself within a single conversation.