roon says AI alignment has made no fundamental advances since RLHF

roon (@tszzl), an AI researcher on X, says there have been no fundamental advances in AI alignment since RLHF, reinforcement learning from human feedback.

He says most alignment work is still folk theory rather than science or engineering, and that capabilities have outpaced interpretability and alignment, as he predicted.

The post is a single claim on X with no supporting data, benchmarks, or named studies attached.