What TikTok got right about discovery
Algorithmic discovery without subscriptions is the most important UX innovation of the last decade. Here's why it works — and how txtfeed applies it to text.
Before TikTok, every social platform was built on subscriptions. You followed accounts, the platform showed you their posts, and discovery happened in the margins — through hashtags, search, the explore tab. Your feed was a function of who you'd already chosen to listen to.
TikTok inverted that. The For You page doesn't care who you follow. It shows you whatever the algorithm thinks will hold your attention next, drawn from the entire library. If you scroll past, the algorithm learns. If you watch to the end, it learns. If you share, it learns harder. Within an hour, your feed is uncannily personal — and you never had to follow a single account.
This is the most important UX innovation of the last decade, and most platforms still don't understand it. Twitter's "For You" tab is half-hearted. Reddit still requires subreddit subscriptions. LinkedIn keeps asking who you want to connect with. They're all bolting algorithmic discovery onto a subscription model, and it shows.
The reason TikTok's approach works isn't the algorithm — it's the friction. Subscriptions are friction. Choosing what to follow is friction. Pruning your feed when it gets noisy is friction. Every step a user has to take to shape their experience is a step where they might leave.
Algorithmic discovery removes all of it. The user does nothing except consume and react. The platform does the work of figuring out what they like. The cost is that the platform has to be very good at reading signals — but signals are easier to extract than commitments.
txtfeed applies this to text. There are no accounts to follow. There are no subreddits to subscribe to. You vote on what you like, you skip what you don't, you save what you want to keep — and within thirty seconds the feed starts adapting. By session three, the algorithm is calibrated. You never had to make a single "who do I want to read" decision.
The hard part isn't the algorithm itself. The hard part is having enough content for the algorithm to choose from. TikTok solved this by making content creation effortless. txtfeed solves it by ingesting from 500+ sources — Reddit, Hacker News, Wikipedia, RSS feeds — and letting cross-source semantic dedup catch the duplicates. The algorithm gets a buffet to choose from, the user never sees the same story twice.
If you're building a discovery product in 2026 and you're still asking users to follow things, you're already behind. The future is feeds that figure you out without you having to tell them anything.
See it for yourself. No signup required.
Open txtfeed