Writing
The Coming Two-Tiered Internet
AI makes content cheap, and cheap content changes the texture of the internet. The popular prediction is that everything drowns in slop. I don't think that's quite right. What I expect instead is a split: the open feed gets noisier, and trust becomes the thing people pay for — with money, attention, or membership.
I work on content moderation systems, so I get to watch the supply side of this up close. The volume of plausible-looking machine-generated content is not a future problem. It's a present one, and it grows in proportion to how cheap generation gets.
The split is economic before it is aesthetic
A two-tiered internet emerges when the cost of generating content falls faster than the cost of establishing trust. One tier optimizes for volume, engagement, and zero friction — that's where the generated material pools, because that's where it's profitable. The other tier sells something scarcer: signal, verified identity, taste, the boundaries of a community that vouches for its members.
This isn't a new dynamic so much as an old one at higher intensity. Email went through it with spam, and the lesson there wasn't that classifiers saved the day. What worked was changing the economics — making it expensive or unrewarding to send junk at scale. I expect the same lesson to apply here: you don't beat infinite cheap content with detection alone, you beat it by making trust the scarce input that distribution depends on.
What platforms will have to do
If engagement becomes easy to manufacture, then engagement stops being a useful ranking signal, and platforms will be forced to rank on other things: who made this, what's their track record, can the origin be verified, has anyone with reputation at stake vouched for it. Identity and provenance shift from account features to ranking features. Content understanding systems have to start asking not just "what is this?" but "where did this come from and why does it exist?"
Curation becomes a real product in this world, not a nostalgic preference for the human touch. Where the cost of being wrong is low — entertainment, ambient browsing — the free tier works fine, and honestly a lot of generated content there will be perfectly enjoyable. Where the cost of being wrong is high — health, money, news, anything you'd act on — people will route around the open feed toward sources that stake reputation on being right.
My working heuristic: use the free tier for discovery, but not for final belief formation. The open feed is good at surfacing things and bad at certifying them, and the gap between those two jobs is about to get much wider.
So no, I don't think the internet dies under AI content. I think it stratifies around trust. The question worth asking, if you're building anything in this space, is which side of that line you're building for.