A Path To AI Enshitification

“Enshitification” is a term coined by Cory Doctorow that describes a process where digital platforms decay from beneficial services into user-hostile, rent extracting monopolies. What if AI is in that beneficial services stage? This would assume that AI companies are becoming a digital platform and engaging in user capture, which partially explains the large investment money, because we know platforms are profitable. There are also not as many platform as one would expect, which worries me. I’ve not thought of AI as a platform, but depending on how you use it, it kind of is really, especially if you add a “buy” button, which we know is coming down the pike eventually.

I cannot help but feel an upcoming catastrophe for AI users, specifically those who have the $20 a month subscriptions who use AI somewhat regularly but not nearly as much as others. Maybe we use it as a search engine, maybe for deeper research, or as a tutor, note taker, support for our side coding projects, or work support. Then there are power users who are cranking through huge code bases, running MCP servers, and creating complex agents. They do pay more, but are those fees helping AI companies offset expense? Are we looking at a rising tide of expenses?

Users like me know we subsidize the heavy users, and users like me also know we are subsidized by investment money. From the outside, AI seems to create both massive fixed expense in terms of processing/memory and variable expense in terms of electricity. My fear is that at some point we wake up and we lose these $20 a month subscriptions, and those become $200, or even $2000. Then the lower market is using cheaper, less robust, maybe even free AIs filled with biased responses and advertisements. What if there is a divide between the AI enabled and the AI disenfranchised? Is this the future?