Rabdology

The path to AI that reasons — not just recognizes — runs through mathematics. Math is where genuine inference can be distinguished from fluent approximation, because solutions are verifiable and novelty is constructible.

We are cartographers of the math frontier of AI. This blog charts what we find — where the models stumble, where they surprise us, and what the failures reveal about the nature of AI reasoning.

Probing the frontier is one thing; advancing it is another. This work led us to found Rabdos to create mathematical reasoning data for training and evaluating AI models. Even as models improve rapidly, we develop original problems at the edge of difficulty, with verifiable solutions, at scale. Our problems are designed to require actual thought: the kind of work where getting the right answer is strong evidence you understood the problem.

Explore our essays, subscribe below, and send any comments to contact@rabdos.ai.