Jevons Paradox: Automation Edition

A quickie today. I spent a big chunk of the day building (and refining) an AI workflow that ate 6 slide decks, re-cut them, wrote a 20 minute script, and generated an entirely new presentation. Then it did the whole thing again. And again. Three full passes, and it never produced anything good enough to actually share.

Ultimately though, what I was left with was: nobody needed this presentation - it was a nice to have. Before AI, I never would have made it. The cost was too high to bother, and that would have been completely fine. The work didn’t get done because it didn’t need doing. What changed isn’t that the work got more valuable. It’s that it got cheap enough to do anyway.

Which is just Jevons paradox wearing a different hat. Make a resource cheaper and people don’t use less of it, they use far more. Claude made slide creation nearly free, so I didn’t save the afternoon, I spent it generating decks that really weren’t needed.

I’m reminded, once again, of the XKCD automation theory-vs-reality comic. It holds up perfectly: AI has almost erased the “writing code” line, and on occasion, it quietly balloons everything else.