What is Two-By-Two Tokens?

What is Two By Two Tokens, and why my AI research is important.

Welcome. My name is Noah Weinberger, and Two by Two Tokens is my personal corner of the internet, separate from my work, where I think out loud. I'm an American-Canadian AI policy researcher and a neurodivergent advocate, and I'm autistic. That last part isn't a disclosure I tuck at the bottom of a bio. It's the lens I see this whole field through. AI systems are increasingly built to assist, monitor, and shape people like me, and I'd rather write from inside that experience than pretend to a view from nowhere.

My professional work lives at Lono Collective. Two by Two Tokens is the place for the questions that don't fit a clear deliverable: What can theology and recursion teach us about how a model decides to refuse? Why does "nothing about us without us," the phrase the disability rights movement fought for, belong at the center of AI policy and not at the end of a consultation? What does it actually feel like to be one of the users a safety team is arguing about?

Where this comes from

I grew up Modern Orthodox, Jewish in a tradition shaped by a phrase: Torah im Derech Eretz, roughly, Torah together with the way of the world. The idea is that sacred study and worldly knowledge aren't rivals. Engaging seriously with science, culture, and the modern world is part of a religious life, not a detour from it. That upbringing is a good part of why AI pulls at me. Judaism trained me to read closely, to argue with a text rather than simply accept it, to treat interpretation itself as serious, even sacred, work. A frontier AI model is, among other things, a strange new kind of text to read closely and argue with, one that argues back. The name of this site is a small joke in that spirit: two by two, like the animals boarding my namesake's ark, and tokens, the units a language model thinks in.

What you'll find here

On Two By Two Tokens, you'll likely read essays on AI and the mind. I will probably do closer readings of how frontier models actually behave, like where their guardrails hold and where models quietly give way. I might also bring arguments and steelman about who gets a say in the systems that shape vulnerable people. For a bit of fun, I might write the occasional stranger piece that follows an idea wherever it goes. Some of my work is rigorous, some of it is playful, and I think the best thinking needs both.

A little more about me

My writing has appeared in Tech Policy Press and HuggingFace, including pieces on the regulation AI companions, the case for youth and neurodivergent advisory councils in AI design, and Europe's approach to AI and mental health. I publish when I have something worth your time, usually a few times a month, never filler. Subscribe below and new pieces will land in your inbox.

Get in touch

I welcome tips, sharp disagreement, and collaboration. You can reach me at:

For anything sensitive, Signal is best.