The author declares that "software composability" is the solution as though that is a given fact. Composability is as much a dead end as the AI he describes.

The author declares that "software composability" is the solution as though that is a given fact. Composability is as much a dead end as the AI he describes. Decades of attempts at formal composability have not yielded improvements in software quality outside of niche applications. It's a neat idea, but as you scale the complexity explodes making such systems as opaque and untestable as any software. I think the author needs to spend more time actually writing code and less time thinking about it.