I keep coming back to AI safety because it sits at an uncomfortable intersection: the most important technical problem of the decade, and one where the social, philosophical, and political dimensions matter as much as the math. It’s a thread you can’t really pull on from one direction.
What pulls me specifically is the question of how systems that are smarter than us at narrow things — and increasingly broader things — get aligned with what we actually want, when we ourselves don’t agree on what we want. The problem touches engineering, governance, epistemics, and the older question of how humans cohere around shared meaning at all.
I’m not coming in as a researcher. I’m coming in as someone who reads, who builds, who notices. The thread is here because the noticing has been happening long enough that it’s earned a section.
This is dummy content. The real essay will replace it.