Here’s a napkin-sketch proof: for many decades we have had hardware that is capable of dextrously automating specific tasks (eg car manufacture) but the limitation is the control loop; you have to hire a specialist to write g-code or whatever, it’s difficult to adapt to hardware variance (slop, wear, etc) let alone adjust the task to new requirements.

If you look at the current “robot butler” hardware startups they are working on: 1) making hardware affordable, 2) inventing the required software.

Nothing in my post suggested costs go to zero. In the AGI scenario you assume software costs halve every N years, which means more software is written, and timelines for valuable projects get dramatically compressed.