I still have the same question about this that I have every time someone proposes it: If there is some hurdle that AI solves that means someone can create a highly profitable business alone, why should someone pay that business to solve whatever problem it solves instead of also getting the AI to solve it? Like if implementation becomes less and less of a barrier, that implementation is also less of a moat, and thus at least software companies per se would seem to not be of much value compared to just having access to the same AI people use to build them. This would basically require a permanent high barrier to access to AI to work out, which to be fair, may be the future a lot of these silicon valley prognosticator types might be hoping for, but if that doesn't happen, which seems more likely, the value of companies resets to whatever assets they can leverage, and at least software, if not any strategic or technical advantage an AI could help with, no longer is a source of surplus value

I already think valuation is a very gameable metric, so I guess you could trivially get this done if you meet a VC who will just buy you a billion dollar valuation on the belief that this is a real thing. I know some enthusiastic kids in the world I could maybe throw a thousand bucks at to buy a GPU in exchange for a millionth of their company, and technically wouldn't that make them a one-person unicorn already?