xAI — the company behind X, formerly Twitter — just published the full code that powers their recommendation feed. Not a simplified demo. The actual engine. The one that decides, thousands of times per second, which posts to show you and in what order.
They also included a pre-trained model you can download and run on your own computer. No building from scratch. No years of research. It's roughly 3 gigabytes — about the size of a few movies — and it works out of the box.
Until now, if you wanted to build something that intelligently ranks content — say, a niche community platform, a curated editorial feed, or a personalised product catalogue — you'd either pay enormous sums for proprietary tech or spend years building it. That bar just dropped significantly.
Think of it like this: someone handed you the blueprints and the working engine from a Formula 1 car, free of charge, and said "do whatever you like with it."
For founders building anything where what the user sees first actually matters — media, e-commerce, communities, newsletters — this is worth paying attention to.
The algorithm that shaped public opinion for hundreds of millions of people is now inspectable by anyone. That's new. What would you build if you could teach your own platform to surface exactly what your audience cares about?
Recommendation algorithm — the invisible logic that decides what content you see, and in what order. Every feed you scroll is controlled by one.
Open-source — software whose inner workings are made public. Anyone can read it, copy it, or build on top of it.
Pre-trained model — an AI that has already done its learning. You just plug it in, rather than teaching it yourself from zero.
Apache 2.0 license — a common open-source permission slip. It lets you use the code commercially without owing anything back.