The author records themselves on audio 24/7 and uses an AI to process the information, creating a personal digital assistant that can perform tasks and provide insights. The system uses a keyword to indicate when a command is being given, and the audio is processed at the end of the day to extract the relevant information. The author notes that this approach has advantages over traditional virtual assistants, including greater control and privacy. The system is still a proof of concept and has limitations, such as the delay in processing commands and the need for manual transcription.