This feels like a discovery problem. These platforms are optimization engines for content consumption, not for genuine recommendation. Their goal is to keep you on the service, not to help you find the perfect movie for a rainy Tuesday night.

As a builder, this led me to a prototype (https://lumigo.tv/en-US): what if you could describe your mood or intent in plain language and get a tailored, unbiased shortlist? I've been working on lumigo.tv to test this. The core is an AI agent that you query like, "a thought-provoking sci-fi movie from the 90s" or "a cozy British mystery series." It searches a database of titles and returns matches with ratings and where to stream them.

The technical hypothesis is that a conversational, intent-based search can cut through the noise better than collaborative filtering or genre rows. No ads in results, no promoted titles—just a direct query-to-match engine.

My question to HN isn't about the specific tool, but the broader principle:

Is the dominant "infinite scroll of posters" model the end-state for discovery, or is it a legacy UI that we've just accepted?

Can a neutral, conversational interface ever compete with the billion-dollar optimization of platform-native algorithms?

What would a technically ideal discovery layer look like? Would it be a meta-layer across all services (like a better JustWatch), or is deep integration with one platform's catalog necessary?

I'm sharing this not for feedback on the site itself, but to discuss the architecture of discovery. Is solving the "what to watch" problem more about better data, a better interface, or changing the fundamental incentives away from engagement maximization?