Any AI system can err, and we're direct about how: Leo is engineered so that errors are the recoverable kind — a wrong character or word you can spot against the image displayed alongside the text — rather than the fluent, plausible fabrications general-purpose chatbots produce on historical material. Leo also automatically detects failure patterns (like repetition loops) and retries rather than showing you bad output. Always verify against the original: that's why the image sits beside the transcription.
