The user tells the device what they have accomplished during the day. Speech is captured through a microphone and transcribed using a distilled Whisper model. The transcription is then processed by a locally running Qwen2.5 language model that has been fine-tuned to behave like a sarcastic and argumentative upper-class British character. The language model evaluates the user’s claims and generates a response. Kokoro Text-to-Speech converts the response into spoken audio. If the system believes the user has been productive, it can open the revolving door and release the drink. If it detects an unconvincing or dishonest claim, the device may activate and swing an electronic fly swatter as a humorous physical response. The system runs on a Raspberry Pi 5 and uses an ESP32 to control the servos. Because the language model and speech-processing components use a large share of the Raspberry Pi’s capacity, the first response may take around one minute, while later responses typically take approximately 15–30 seconds. The system runs on a Raspberry Pi 5 and uses an ESP32 to control the servos. Because the language model and speech-processing components use a large share of the Raspberry Pi’s capacity, the first response may take around one minute, while later responses typically take approximately 15–30 seconds.