Superalignment

The two get conflated constantly, and the conflation is most of why AI projects stall. Capability is a question about the model and it is largely solved. Permission is a question about accountability, and it is answered by whoever will be responsible when it goes wrong. No amount of extra capability answers it.

Why it matters

The bottleneck is almost never capability, and treating it as one wastes a year.