Sometimes it strikes me that we probably give these AI agents a little too much credit for their supposed ingenuity.

As I write this the internet is abuzz with this term called AGI, or artificial general intelligence, yet another “look what our agents can do!” moment in the AI news cycle which is almost certainly (and mostly) a marketing push by the big AI bros. General intelligence is the current gold standard for which they are aiming, suggesting that their agent is just plain clever enough to do basically anything.

But are they doing anything or just getting more robust at doing what we ask them?

I was thinking about my first forays into building webpages, and it struck me that the first time I built a site I was coding raw HTML into a text editor. 

A few years later I used a CMS, but still nearly everything inside the system was complex and hand-coded.

Even more years later the CMS handled a lot of the content, but the design was still finicky and needed constant turning of CSS and div containers, and it took time and thought and effort.

This very afternoon I was tweaking the design of this site and most of it was done with toggles and menus and sliders and select boxes.

A couple years from now will I be just nudging an agent baked into WordPress and with access to my admin panel to brainstorm the look and feel of my current project? Probably.

But none of that makes them creative, or clever, or generally intelligent. Rather, at least in this context, it simply turns them into ever better tools to accomplish the tasks many of us have been doing by hand for decades.

Some people will, of course, hand over the whole workflow to algorithms and agents to build and write and publish and clutter up the internet with slop. That is inevitable, but then that too is just con men using ever better tools to run their con.

The hype tells us that we’ve unleashed something bigger. But the proof often seems as more a magic trick and sleight of hand as it is credit for some class in which we’ve all been schooled.