Solving Hard Problems Requires Search. The Brain Already Knows How.
Reinforcement learning is, at its core, amortized search. A way of compressing the results of exhaustive trial-and-error into policy parameters, so that at inference time, the right action can be retrieved without re-running the search. It works. But it has a fundamental ceiling: the quality of the compressed policy is bounded by the quality of the search that produced it.
Hard problems, genuine reasoning, novel planning, open-ended generalization, cannot be fully amortized. They require search at test time, not just at training time. This is why scaling alone is not enough.
The brain solved this problem long ago. Biological cognition doesn't brute-force search over exponential spaces; it navigates them efficiently through mechanisms like hierarchical abstraction, dynamic gating, and selective hypothesis pruning. The brain performs structured, resource-constrained search and does it on 20 watts.
At Voaige, we study the computational principles behind this efficiency and generalization. Generalization is what determines whether a system can transfer competence to new domains, adapt to distribution shift, and remain useful in the open-ended complexity of the real world. Our goal is to understand how the brain conducts search, what algorithmic strategies it employs, and where the real gains come from, then implement principled approximations of those mechanisms using the representational machinery of large language models.
This is what we call Test Time Cognition: search-capable inference that is architecturally grounded in neuroscience, not just scaled compute.