Observations on Google Discover's Cold Start
gaganghotra_ · x · 2026-07-10
This article discusses an easily underestimated issue in Google Discover: the cold start problem in recommendation systems, specifically how "very recent articles" are distributed when they have almost no interaction history.
While monitoring the Discover feed, the author noticed a briefly appearing internal tag, coldstartcard.f:
- This tag only appears for a very short time and is linked to brand-new, trending articles.
- The sample size was tiny: about 6 impressions and 4 unique articles over 48 hours.
- Therefore, it cannot be considered a universal rule for Discover, but it illustrates how recommendation systems handle content that is "potentially hot for the user, but completely cold for the model."
The article raises a key question: when an article has almost no historical signal, how does Google combine its title, entities, images, source, and topic to decide whether to push it to users.
More from Research
- New paper defines self-state attacks, showing OS defenses leave four agent-memory cases indistinguishable — Justgototheeffinmoon · 2026-07-22
- Krea 2 users recommend a two-pass Clownshark sampler setup for sharper image details — listopalafoto · 2026-07-22
- Animation shows how an MLP’s first-layer weights change while learning MNIST — CatAstro_Piyush · 2026-07-22
- Project APE finds verifier reliability drops when papers contain multiple errors — soumitrashukla9 · 2026-07-22
- Project APE says verifier costs fell about 90x in a year as Chinese open models lead — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22