culture
Why Netflix's recommendations feel worse every year
They probably are not worse. They are optimising something that used to overlap with your interests and now overlaps less. Here is the mechanism.
5 min
Writing
On taste, algorithms and the twenty-minute scroll.
Notes from building a recommendation engine, and what the data says about how people actually choose a movie.
guide
Why dialogue is so quiet when you stream movies, and how to fix it
Quiet voices, loud explosions. How studios mix movies, how streaming services measure loudness, how TVs fold sound to stereo, and what fixes it.
10 min
guide
Why Netflix thumbnails are different for everyone
Netflix picks the artwork for each title per viewer, using an algorithm called a contextual bandit. How artwork personalisation works and where it goes wrong.
12 min
guide
How movie recommendation algorithms work: the four main types
Collaborative filtering, matrix factorisation, content-based filtering and hybrids. How each movie recommendation algorithm works and where it fails.
13 min
guide
Why streaming apps forget where you stopped
You stopped at the reveal and it resumes ten minutes earlier, or at the start. Not a fault in your account. It is how your position is stored and synced.
8 min
guide
Why streaming picture quality drops in the middle of a movie
The image goes soft in a dark scene and sharpens again a minute later. Your connection did not change. What changed was the version of the movie you were sent.
8 min
guide
Why streaming search cannot find the movie you just typed
You know the movie is there and the search box disagrees. The cause is mechanical, and it comes down to tokens, edit distance and the names one movie carries.
8 min
guide
Why every “movies like this” list shows the same movies
Similar-movie lists converge on the same handful of titles whatever you type. The cause is a measurable property of high-dimensional space, not laziness.
8 min
guide
Why streaming availability data is wrong, and how to check it
Every "where can I watch this" answer comes out of a licensing pipeline with several places to go stale. What breaks, which sources hold up, how to check.
11 min
guide
Free, legal movie streaming: where the movies are, and how to choose
Ad-supported streaming is now a licensed market of its own. Which services carry what, why their catalogues rotate, and why choosing is harder there.
8 min
guide
Why recommendation apps ask you to rate things first
An app that knows nothing about you can only show you what is popular. Getting past that costs your attention, and most apps spend it badly.
7 min
product
How Filmatic picks one movie for two people
Averaging two tastes gives you a movie neither person wants. Two-person mode scores every candidate against both profiles and ranks by the weaker of the two.
6 min
product
How Double Feature finds the movie to watch after yours
Pairing is not similarity. The nearest movie in the model is usually the worst thing to watch next, and the reason why explains how the whole feature is built.
5 min