Is AI Progress Stalling? Or is it me?

Arnold Kling

Jul 30, 2026

I am a little concerned about how this little essay will land. A lot of people have deeply-ingrained opinions about AI. If AIDS were not already taken as an acronym, I would say that many have AI Derangement Syndrome. Love AI or hate it, none of us, including me, has enough experience to be able to really be sure where it’s going or what the impact will be. So I would appreciate if if you can keep your comments focused on the limited contents of this essay and not use it as a platform to say things like “Ha! What I’ve always said about AI is…”

The narrow point of this post is that I find that I am not doing anything with AI that I could not have done two years ago. I have seen the stories that AI has gotten much better at math, but I don’t have math problems to solve. Software developers report that it can spend a lot of time working on a coding project, but I do not have a pressing coding project these days. I read that AI agents can use a virtual computer or use your computer, but my attempts have not yielded results that are worth the effort.

I contrast this with my experience as an early creator of a commercial web site, from 1994-1999. Netscape was the great innovator in those days, and I tried all of their tricks. These included cookies, Javascript, frames, and more. Eventually, a Web standards consortium had to get together to sort out all of the user-interface design elements into what is called Cascading Style Sheets, which I also tried. We adopted the JavaWebServer before anyone was sure that it would work. We relied on a database backend for every page on our site.

In those days, every few months web sites could do new things with these tools. Web-based applications became increasingly sophisticated and compelling for users. Many people became webmasters and web developers.

If something similar is happening in AI, I am not seeing it. Maybe I am just too old. When the Web took off, I was in my forties. Now that I am in my seventies, I may have less imagination. I definitely have less stamina. Perhaps if I had been this old when the Web hit, I would have not had what it takes to maintain a web site.

A Missing Layer?

With the Web, those of us who were webmasters and web developers were an important part of the growth of the ecosystem. Similarly, with smart phones, app developers were an important part of the growth of that ecosystem.

With AI, the major labs are developing the core technology. End users are trying to experiment with it. But I am not seeing an in-between layer, which would be people developing applications on top of AI that draw users into the ecosystem.

The labs have gestured in the direction of creating such a middle layer. Remember when OpenAI introduced “GPTs”? Claude “skills” could serve a similar function. But I do not see any ferment coming out of those efforts. Perhaps the lack of a business model for a developer of a “GPT” or a “skill” is a problem. Perhaps the fear that the labs will cannibalize any successful applications is another.

I am starting to worry that AI is falling behind relative to the trajectory the Web took in the late 1990s.

Share