Go Watch This
A decisive watch picker that returns one title, not another list. Built to remove decision fatigue from streaming.
The problem with streaming is not the content. It is the choosing.
Every platform is a browsing surface that assumes you want to research your next watch. Catalogs, carousels, recommendations, more carousels. By the time you find something, you are too tired to enjoy it.
Go Watch This is built on a different premise: ask for only the necessary context, then return one clear answer.
You pick your country, your streaming services, whether you want a movie or series, and an optional genre. The app returns one title. If it is not right, you pick another — it will not repeat what it has already shown you. That is the whole interaction.
The design principle that drove everything was: one answer beats another list.
Most recommendation tools are still browsing tools in disguise. They give you a ranked list, a grid of posters, a shortlist to scroll. That recreates the same decision fatigue the tool was meant to solve.
Go Watch This refuses the list. The answer is the answer. The next action is obvious from the first screen.
Under the hood, the app pulls live availability data from a streaming catalog API, so the recommendation is always something you can actually watch — not a title that left the platform six months ago. Picks are filtered by your exact streaming subscriptions, genre preferences, and media type.
OpenAI reranks the candidate pool when configured. The filtering and memory logic runs server-side so API keys never touch the browser.
The memory layer is small but meaningful. Titles you have watched, liked, or dismissed are remembered across sessions — likes boost similar genres, dislikes suppress the title and reduce adjacent categories. The app gets slightly better at reading your taste the more you use it.
This is also an experiment in building with AI in the loop from the start — not as a feature bolted on, but as part of the product’s core decision-making. The reranking step is one example: AI is not generating recommendations from scratch, it is selecting from a filtered, trusted pool. That keeps the result grounded and fast while still letting the model apply judgment about fit.
Built with Next.js, App Router, Streaming Availability API, Turso for persistence, and OpenAI for optional reranking.