What the others said about dataset size, but there's no reason to fully "reinvent the wheel".

Train a Lora on your posts and bolt it on to your favourite LLM. It won't quite be you, but you'll be horrified and entertained by how close it can get :D