Introducing Ruliad
An independent research lab looking for what comes after today's models.
We're Ruliad, an independent AI research lab.
Scale has taken today's models a long way, but the way they learn has barely changed. They're trained once and then frozen, they need enormous amounts of data, and it's hard to see why they do what they do. We think the next step in AI will come from better learners, not only bigger models, and we started the lab to work on that.
What we work on
We explore the AI tech tree one branch at a time: new architectures, new learning algorithms and the systems built on them. Some branches will be dead ends. We'd rather find that out quickly and move on.
We also use AI in our own research, to run more experiments than we could by hand.
What to expect
We'll publish as we go: papers, replications, technical notes and models, including the results that didn't work out. Our first release, DeepThought-8B, is out today as open weights.
If you're working on related problems, we'd like to hear from you at team@ruliad.co.