Curiosity, with a method.
Null Theory is an independent AI research lab studying the foundations of intelligent systems: how they learn, how they reason, and what computation makes possible.
There is a meaningful difference between getting a system to work and understanding why it works. We care about both, but we begin by asking questions that can be turned into experiments.
That usually means building small versions of complicated things. An attention mechanism, a training loop, a quantum circuit. Something simple enough to inspect, test, and eventually explain.
Build to understand.
Implement the mathematics and the mechanisms, rather than treating them as black boxes.
Test the intuition.
Make clear predictions, use controls and baselines, and treat surprises as information.
Show the work.
Publish methods, limitations, and honest distinctions between ideas and evidence.
Team
Aaron Arambula
Founder
Computer science at Arizona State University. Builds the lab’s core tooling, including aaronnet, our from-scratch autograd engine.
GitHubHow we label work
- Plan
- A question we have scoped but not started.
- In progress
- Code or writing exists, but the result is not final or the artifacts are not yet public.
- Result
- Code and data are public, and the claims can be checked.
Contact
For collaboration, questions, or feedback on our work, write to us.
research@nulltheory.org