> So AI will largely hit a wall, frozen at the current level and work will all shift to cheaper inference.
Looks like an ideal outcome to me until (if) we are able to solve the alignment issue.