"it was a proof of concept, not a commercial product"
Within that distinction, lies dragons.
The issues I have with existing systems are divided in two categories: bugs/issues/not being polished enough, and conceptual inadequacy.
Most of my issues with actual products fall within the first bucket, because the second bucket actually means I'd want another product. This is where innovation is.
That being said, unlike Kay, I don't loathe what we have. I'd be very happy if it was just better. But, like him, I thin the fact it's bad has deep roots in the way it is conceived.
I think another commenter hit the nail on the head when he said that Kay was mostly disappointed with users.
About Steps specifically, I haven't tried it out, but I harbor a lot of doubt. A huge amount of work goes into making stuff merely tolerable.
Something I do have experience with is Pharo Smalltalk (which I believe is the most advanced offering in the market). In theory, having your own small universe based on coherent principle is heaven. In practice, it feels very constricting to work inside that box (the Smalltalk image), because you don't have access to some of your best power tools, or even to some basic modern niceties.
(Pharo Smalltalk and great, and improving steadily, you should try it out. But I don't see myself calling myself a Smalltalk developer/hacker any time soon.)