💭 when learning something technical from an llm, good to always assume the llm is wrong and bullshitting you and treat it like a puzzle to find the gaps and mistakes in its reasoning