I need your help to create brain-rot dataset (huggingface.co)

Currently the dataset contains around 35 examples of instruction-output pair which is no way close for an effective fine tuning. So I am looking for contributers to the dataset who has got some 'ohio skibidi rizz'