Why we killed the like button
Every product reflexively adds a like button in the first week. Here's why txtfeed refused — and what we built instead.
There's a reflex in consumer product design that goes like this: when you launch a feed of any kind, you add a like button within the first week. The button is cheap to build, users expect it, and the metric it produces (likes per post) is legible to investors and product managers alike. Almost no one stops to ask whether the like button is actually doing anything useful for the product.
txtfeed doesn't have a like button. We considered adding one, decided against it, and we've stayed with that decision for good reasons.
The problem with likes is that they only carry positive signal. A like tells you "I saw this and I didn't hate it." That's not a quality signal; it's barely an attention signal. Because likes are unidirectional, they can only rank content by popularity, never by quality. Popular and quality are correlated but not identical, and for a reading product that difference matters a lot — optimizing for popular gives you clickbait, optimizing for quality gives you depth.
We replaced the like with upvote + downvote, the Reddit model. The ratio is the signal, not the raw count. A piece with 100 upvotes and 0 downvotes is genuinely loved. A piece with 100 upvotes and 200 downvotes is polarizing and probably clickbait. The algorithm treats these differently — and the user-visible result is that clickbait dies faster on txtfeed than it would on any like-only platform.
There are second-order effects too. Likes train users to reward low-effort engagement-bait because it's easier to double-tap a hot take than to downvote it. Upvotes + downvotes train users to distinguish between "this made me react" and "this is worth ranking highly." The distinction sounds subtle; in aggregate it shapes the entire content ecosystem.
The lesson for product builders is to resist reflexive feature additions. A feature that every competitor has is not, by itself, a reason to add it. Every feature changes the system, and the features you decide not to ship are just as much of a design choice as the ones you do.
Found this useful? Share it.
See it for yourself. No signup required.
Open txtfeed