Error correction is fascinating because it mirrors challenges in other fields - like how in statistical NLP, you need redundancy and cross-validation of multiple weak signals to get a strong reliable output. The overhead ratio (hundreds of physical qubits per logical qubit) feels similar to ensemble methods where you trade efficiency for reliability.