Just Enough Residual Value

Ruby Justice Thelot, an NYU professor of design and media theory, published a piece on a16z's outlet arguing that enshittification is not real — that what people are describing is platform narcissism, which he defines as "the belief that the platforms one uses should continue to respond to one's idiosyncratic user needs." The observation underneath that is sound. Platforms do reorient toward new cohorts. The people who found Instagram in 2012 are not who Instagram is being built for in 2026, and "it stopped being for me" is a genuinely different claim from "it got worse." Those get conflated constantly, and separating them is a real service.

Then he goes to the numbers, and the numbers are where it comes apart.

Start with the weakest, which isn't really evidence at all. Social media users grew from 2.5 billion in 2016 to nearly 6 billion today — mostly a story about people getting online for the first time. New markets, new cohorts, first phones. Lock-in explains why nobody leaves; it does not explain why anyone arrives. Set that figure aside; it was never the disputed part.

The other two are the ones doing work. Instagram's monthly actives are up 140 percent over ten years, daily time-on-app rises year over year, and the American Customer Satisfaction Index has social media climbing from 70 to 75 out of 100 between 2020 and 2025, with the 2026 Entertainment Study putting 30,886 respondents behind that figure.

Both of those are numbers the theory predicts go up.

Read Cory Doctorow's model rather than the meme of it. Three stages: platforms are good to users to attract them; then they degrade the user experience to serve business customers; then they claw the surplus back from everyone for shareholders — leaving, in his phrasing, "just enough residual value in the service to keep end users stuck to each other." Stuck. Not gone. The mechanism is lock-in, and lock-in produces more accounts, not fewer. The extraction is calibrated to stop precisely where leaving starts, which means the theory's own prediction for a satisfaction score is: adequate, stable, mid-seventies.

That is not a refutation of the model — but I should be careful how far I take it, because a theory that reads a rising satisfaction score as confirmation and a falling one as confirmation too has stopped being a theory. So name what would break it. Doctorow's model fails if you find users whose exit is genuinely cheap — portable graph, portable archive, somewhere else that works — reporting rising satisfaction anyway. That test is runnable. Nobody has run it, and Thelot did not run it either. He measured satisfaction among people who cannot leave and reported the number was going up.

And look at what the instrument measures. The ACSI Entertainment Study scores app reliability, loading speed, content freshness, and ad relevance, on subscores running 68 to 84. Read the fourth one twice. Ad relevance is not a neutral vital sign. Better targeting is one of the specific mechanisms by which surplus moves from the user to the advertiser — it is the middle stage, operationalized. The survey is scoring a symptom of the thing as evidence against the thing.

The load-bearing sentence in the piece is the one about price. Platforms are free to consumers, Thelot argues, so if users were trapped the satisfaction scores would show it. That is a category error about where the cost sits. Instagram costs nothing at the door. What it costs is at the exit: your photographs, your message history, a decade of social graph, and every person you know who is not going with you. Enshittification is a theory about switching costs. The rebuttal measured the entry price, found it was zero, and concluded the exit was clear.

The venue is worth a sentence, not a dismissal. There is no disclosure line and a16z holds equity across the platform economy, but "a VC published it" is a bad reason to throw out an argument. The narrower point is that a firm whose returns depend on the durability of lock-in ran a defense that treats lock-in metrics as satisfaction metrics. That is not corruption. It is a blind spot, and it has an address.

Emanuel Maiberg, writing at 404 Media, made the plainest version of the counter: you can hate something with your entire being and still participate in it. He has the polling to back it: 52 percent of Americans say they spend too much time on these platforms, and 60 percent would ban them for kids under sixteen. None of that shows up in a survey about loading speed.

And it can't, because no such survey exists. There is an index for app reliability and one for ad relevance, fielded to thirty thousand people, because platforms and advertisers need those numbers and will pay for them annually. There is no index for what it costs to leave. Nobody with a research budget wants that figure on a chart, so it isn't on one. The whole public argument about platform decay is conducted inside instruments built by one party to it — which means Thelot did not cherry-pick his data. He used the only data there is. That is more generous to him than this piece has been, and considerably worse news than anything he set out to argue against.

Six billion accounts and a 75 out of 100. The number the argument actually turns on has never been collected.