Netflix Prize, 20 Years Later: Take on the challenge with our new codebase

Netflix Prize, 20 Years Later
Are you curious about how classic ML benchmarks behave in the age of AI-assisted development?

Event format: Hybrid: in person in Budapest and online
Date: September 23, 2026
Time: 6 PM CEST
Venue: Villányi út 40/B, 1113 Budapest, Hungary
Online link: https://meet.google.com/ami-jwaw-sbt

Preliminary program:
18.00-18.05: Domonkos Tikk: Introduction
18:05-18:25: Gábor Takács: The released framework and algorithmic advances
18:30-18:40: István Pilászy: A case study on the controlled use of AI
18:40-18:50: Bottyán Németh: AI in autoreserch
Short break
From 19:00: hands-on demonstration on how to use the released framework

Twenty years after the launch of the Netflix Prize (NP), we are revisiting one of the most influential machine learning competitions of all time.
Members of the original Gravity team, which also led The Ensemble team (which finished tied at the top of the leaderboard), recently revisited the NP task: not by reopening the old codebase, but by rebuilding the approach using today’s hardware, software infrastructure, deep learning frameworks, published knowledge, and AI-assisted coding tools.
The result was both exciting and humbling: even after two decades of progress, the Netflix Prize remains a remarkably hard benchmark. Our invited paper, “Netflix Prize revisited: a hard problem even after 20 years,” has been accepted to the RecSysChallenge workshop at RecSys 2026, and we have released the revival code base publicly:
https://github.com/gtakacs/netflix_prize
About one week before the workshop in Minnesota on October 2, we are organizing a hybrid meetup in Budapest and online to bring together people interested in recommender systems, machine learning competitions, deep learning, and AI-assisted coding.
Our goal is to invite the community to work on this famous task again.
We will discuss what made the original Netflix Prize special, what we learned from revisiting it with modern tools, why the task is still difficult, where deep learning helped or did not help, and how AI coding assistants performed in a real, metric-driven machine learning project.
You can join us in person or online.
Whether you’re an AI expert, data scientist, Kaggle and ML competition enthusiasts, recommender-systems practitioners, developer, or simply curious about what today’s technology can do, come along and find out with us:
Netflix Prize: Reloaded.

And of course, there’ll be food and drinks too! Stick around after the talks for some food, drinks, and informal conversations with fellow AI enthusiasts.
Whether you want to contribute code, try a new model, revisit an old idea, discuss recommender-system methodology, or simply join the conversation around a landmark ML challenge, you are warmly invited.

Let’s see what the community can do with the Netflix Prize, 20 years later.