> We are certainly closer now to being able to prototype and go to market faster with a product.

What are the higher-order effects when anyone can do this, and *aaS becomes a market for Lemons?

I think that just because anyone can do it, doesn't mean they will. Lots of people have really great ideas but very few actually commit to execution. Ultimately ROI will go down, deincentivizing the commercialization of that thing someone wanted to bang out in a weekend.

In the very long term, software will become a commodity, as you mentioned. Process and workflow may move into JIT delivery for the need at hand, in theory the data layer will be comprehensive and clean and the days of clicking around a bunch of stuff to fulfill process needs will move into a lower latency activity like...talking to your agent.

I saw a quote today by Brian Eno(1995) that said: "So the question becomes not whether you can do it or not, because any drudge can do it if they're prepared to sit in front of the computer for a few days; the question then is: of all the things you can now do, which do you choose to do?" and it resonated with me a lot.

If you can gin these things up in a weekend then why would you bother with a monthly subscription model for software? The only valuable part is the specification and possibly the hardware to run it. If I were a CTO trying to save money I might pay for the labor to develop good specs, but I would prioritize getting out from under software companies with a rent seeking models and 80 to 90% margins

It's not a market for lemons. We can share info about the lemons and all choose to use the good ones. There's no information asymmetry.

in the 90-ies anyone could easily prototype with tools like Access (and all the other "4GL" tools which were similarly all the rage back then). That still didn't preclude companies from buying their major software from software vendors instead of doing it themselves.

In some sense having customer able to prototype what they want is a good thing. I did it myself as i was at the time on that side, and having a quick-whip-it tool was a good thing to quickly get some feature that was missing in the major software before that major software would add it (if at all). (And if one remembers for example Crystal Reports - while for "reports", it and the likes were in many senses such quick-whip-it tools for a lot of such customization that was doable by the customer.)

So, after initial aftershock - "Ahhhh, we don't need software companies anymore!" - we'll get to the state with software companies still doing their thing just with a lot of AI as specialization is one of the main thing in modern economy and AI becomes most powerful tools of the trade. (and various AI components themselves will be part of software delivery, like say a very fine-tuned model (hosted or on-premise) specific to the customer and software - Clippy on steroids)

(Of course some companies wouldn't survive the transition just like some companies didn't survive the transitions to client/server, cloud, etc. while some new companies will emerge like Anthropic has today or Borland had at the time)

It means the same: random lottery of mass, with everuone else failing.

American capitalism hides the depressing fact that rarely does the best succees.

AAI momentum is parallel to just buying lottery tickets and doing so with the belief that you know the real odds, so one can overwhelm with quantity of tickets.