Subconscious
Intelligence
We started from a simple observation about minds. What you can recall and recite is the conscious layer: shallow, queryable, and small. Underneath it is the unconscious, the intuition built from experience. You would die if you had to read a document to learn that snakes are dangerous when you see one. The knowing is consolidated below language.
The frontier of AI is racing to make the conscious layer larger: bigger context windows, better retrieval, more connectors. That work helps, but we think the harder gap sits deeper. The valuable thing is rarely the artifact alone. It was the derivation that produced it, and the sense of what mattered that selected it.
A mind doesn’t consolidate everything it experiences into intuition. It consolidates what mattered, and that signal fires before the outcome is known. We build the signal for how much a moment mattered from behavior itself so the system remembers like a person does, not like a log file.
Retrieval is the conscious layer: fast, shallow, and forgotten the moment the window closes. The deeper move is internalization, turning experience into a standing capability the model has, not a document it looks up. We see the structured record of how an organization works as the substrate from which that internalized layer is eventually built.
We’re building the intelligence layer beneath the one everyone else is building. It’s a contrarian bet, that the durable value in AI is not a larger working memory but a deeper one, and it touches hard problems in cognition, privacy, and machine learning at once. If the gap between what a company writes down and what it actually knows is the most interesting problem you can think of, we’re looking for you.