Yandex Music updates its recommendation algorithm to keep listeners' tastes from going stale

Yandex Music has updated its recommendation algorithm to help users keep their musical tastes from calcifying. The new version of My Wave now pushes new artists and tracks from other genres more actively, so listeners' preferences can keep evolving as they get older.

Why tastes calcify

A 2018 study turned up an amusing but sad finding: on average, people stop listening to new music at around age 33. After that, their musical preferences freeze, and new genres and artists stay on the edge of their attention. It happens to almost everyone. With age, people settle into a taste, pick up favorite artists, and stepping outside those boundaries starts to feel strange or pointless.

The statistic is easy to understand: you have found music you love, and choosing from millions of tracks gets exhausting. But it also means missing a lot of great music. The problem is especially acute in the streaming era. There is too much choice, and without a good algorithm, people get lost in the catalog and play the same songs on repeat.

The 2026 algorithm update

In early 2026, Yandex Music rolled out a major update to its My Wave algorithm. Technically, the system is now more powerful: it does a better job of analyzing listening context, accounts for how preferences shift over time, and predicts when a user is ready to try something completely new.

The results are impressive. Over four years of development, the algorithm has learned to generate roughly half of all songs in users' favorite playlists. That does not mean people stopped adding tracks themselves. The recommendations simply became so relevant that users save them as if they were their own finds.

For comparison, most users on YouTube Music or Apple Music curate their playlists themselves. On Yandex Music, the split is nearly even: fifty-fifty between tracks users add on their own and tracks the system recommends.

How the algorithm broadens horizons

The updated system works on several fronts at once:

• It suggests new artists that resemble a user's favorites without duplicating them
• It finds tracks that fit the current mood while stretching genre boundaries
• It balances stability (familiar genres) against novelty (unexpected finds)
• It accounts for context: music for work differs from music for exercise or relaxing
• It adapts to seasonal changes (a summer playlist differs from a winter one)

Systems like this work because they do not try to guess a user's taste once and stop there. They keep learning, tracking which recommendations users save, which they skip, and how their mood and behavior change.

What it means

The update shows that recommendation algorithms are becoming a full-fledged tool for developing musical taste. Instead of freezing on the same artists at 33, people can now discover new music at 40, 50 and beyond. That is thanks to AI that has learned to surface not what is popular and predictable, but what is genuinely interesting to a specific person.