Logan Riggs comments on Finding Sparse Linear Connections between Features in LLMs