← Blog · English · Français · Deutsch · Español · Italiano · Português (Brasil) · 日本語 · 한국어
You open your streaming app. Three thousand titles, a wall of glossy thumbnails, a row that promises it's "Top picks for you." Twenty minutes later you've watched four trailers, added nothing to your list, and you close the app to scroll a different one.
Sound familiar? The strange truth of the streaming age is that the more an algorithm learns about us, the less we seem to know what we actually want to watch. We think the problem is too much choice. It isn't. The problem is who's doing the choosing.
A recommendation engine doesn't ask "what will this person love?" It asks "what will keep this person watching?" Those sound similar; they're not. The first is about your taste. The second is about retention — minutes watched, sessions per week, churn avoided. The feed is tuned, relentlessly, to the metric that keeps the subscription alive, and that metric is rarely "you went to bed satisfied."
So you get the safe bet. The thing that's a bit like the last thing. The crowd-pleaser with the highest completion rate. Algorithms are brilliant at finding the median of what people like you will tolerate — and almost useless at finding the odd, specific, slightly difficult film that becomes your favourite of the year.
Three of them, mostly.
Homogenisation. When everyone is nudged toward the same high-performing titles, taste flattens. The long tail — the small films, the foreign series, the weird documentary — gets buried because it doesn't move the metric.
The bubble. The feed shows you more of what you already clicked, which teaches it to show you more of what you already clicked. It's a mirror, not a window. Real discovery — the kind that surprises you — is exactly what a feedback loop can't give you.
The data. To predict you, the machine has to model you: what you watched, when, for how long, where you stopped. That profile is valuable, and it's rarely kept only to recommend movies. The price of the "free" suggestion is a remarkably detailed map of your attention.
Think about the last great recommendation you actually acted on. It almost certainly didn't come from a feed. It came from a friend who said "you, specifically, need to watch this."
That sentence carries things an algorithm can't fake:
Recommendation, at its best, isn't a prediction. It's a small act of friendship.
Before the feed, this is simply how it worked. You'd be at dinner and someone would lean in: "Have you seen…?" A handful of people whose judgement you trusted, swapping the few things genuinely worth your evening. No ranking, no infinite scroll, no one optimising anything. Just taste, passed hand to hand.
That model didn't scale to a billion users — so the industry replaced it with one that did. But "scales well" and "works well" are not the same promise.
Rekko is, frankly, a reaction. We wanted the dinner table back, on a phone, without the surveillance. So there is no public feed and no recommendation algorithm. Instead you create small private circles — family, friends, the three people whose film taste you actually trust — and you pass recommendations directly, with a personal note. Your friend gets a nudge, opens the title, adds it. That's the whole loop, and it closes the way a conversation does.
And because the point is trust, privacy isn't a feature bolted on — it's the architecture. There are no third-party trackers and no ads. You sign in with Apple, with no email or phone required. Your lists, ratings, circles and recommendations sync through your own iCloud, not a server of ours. We can't mine what we never hold.
The algorithm will keep getting better at keeping you watching. It will never care whether you enjoyed it. The people who know you already do — Rekko just gives them a place to tell you.