Specific before sweeping.
A broad idea becomes useful only after it is reduced to a claim that could be disproven.
Sea Salt Labs is a research and incubation studio created to give long-running investigations, experimental systems, and public technical work a durable home.
Read the studio positionSea Salt Labs is not organized around one arithmetic experiment, one architecture, or one product. It exists to give technically ambitious work enough time and structure to become specific.
A question may become a controlled experiment. An experimental method may become a public system. A useful system may become a larger body of work. The studio exists so those transitions do not have to be forced into a product roadmap or an academic incentive structure.
The current focus is artificial intelligence research: internal neural computation, learning and generalization, experimental evaluation, and the data systems needed to study those subjects rigorously.
A broad idea becomes useful only after it is reduced to a claim that could be disproven.
The model, evaluator, and analysis pipeline must work before their output becomes evidence.
Failed gates and amended protocols are part of the research record, not debris to be hidden.
Interesting measurements are reported without quietly turning them into stronger claims.
RESEARCH DISPATCH
Field notes, protocols, datasets, and experimental releases as they become defensible.
Follow the lab